Neurophysiological and imaging biomarkers of lower motor neuron dysfunction in motor neuron diseases/amyotrophic lateral sclerosis: IFCN handbook chapter

(cid:1) Conventional nerve conduction studies, H-reﬂex studies, and needle electromyography are important tools for the diagnosis of Motor Neuron Disease (MND) (cid:1) High-density surface EMG, nerve excitability testing, electrical impedance myography, MUNIX, and MScanFit are potential biomarkers for MND. (cid:1) Neuromuscular ultrasonography can detect changes in nerve and muscle architecture including dynamic changes such as fasciculations.


Introduction
Neurophysiological testing can provide critical information for the diagnosis and monitoring of amyotrophic lateral sclerosis (ALS), identifying pathology that may not yet be apparent in a clinical examination or other investigations.

Early diagnosis
The early diagnosis of motor neuron disease (MND)/ALS is critical for clinical trials of the response to therapy.It is likely that early treatment has a greater chance of benefit: if the disease is too widespread at the time of diagnosis, one could merely be ''shutting the door after the horse has bolted."While specialists experienced with managing patients with ALS now focus on early diagnosis, adherence to older criteria by general physicians is not necessarily helpful, particularly for clinical trials.It is a concern that including patients with ''possible" ALS in clinical trials could confound outcomes, but Geevasinga et al. (Geevasinga et al., 2016) found that ''inclusion of possible as a positive finding enhanced sensitivity. ....while maintaining specificity. ...." In any case, the rapidly progressive nature of the disease would resolve most diagnostic uncertainties during the duration of a clinical trial.It can be argued that the aim should be to make a diagnosis, albeit without absolute certainty, when only a single area is involved, and that clinical trials should focus on such patients.
In establishing an early diagnosis, one should question how current testing aids diagnosis, and whether other neurophysiological procedures might provide greater insights than the current gold standard of careful needle electromyography (EMG)-.Is there a possibility that the rigid diagnostic criteria have in fact impeded early diagnosis?

Diagnostic criteria
The El-Escorial (Airlie House) criteria for definite ALS required the involvement of three of four regions (bulbar, upper limbs, trunk, lower limbs), making it certain that when a ''definite" diagnosis was made, the degenerative process involved so much of the neuraxis that the disease could not be arrested, let alone reversed.To improve the diagnostic sensitivity, the Awaji criteria (de Carvalho et al., 2008) were introduced to supplement the El-Escorial criteria, which had been revised but still required stringent criteria for diagnosis.
The focus of the Awaji recommendations was on the value of the needle EMG study.First, EMG findings were afforded the same weight as a physical examination.Electrophysiological evidence for chronic neurogenic change was considered equivalent to clinical evidence.This change is unremarkable.All electromyographers are accustomed to finding evidence of subclinical neurogenic involvement, i.e. abnormalities of the lower motor neuron (LMN) in a variety of diseases, not just ALS, and indeed it is amazing that EMG evidence of denervation had not been given primacy of place previously.Secondly, the type of EMG data that can indicate ALS was clarified.Fasciculation potentials were considered equivalent to fibrillation and positive sharp waves in identifying denervation.Instability of fasciculation potentials and in motor unit potentials was considered an important feature in ALS.However, Johnsen et al. (Johnsen et al., 2019) showed that the old and new criteria were complex to apply and prone to error.A consensus meeting of international experts was convened to take a fresh approach to diagnostic criteria, rather than to tweak, yet again, criteria that were not fit-for-purpose.Among other things, it was noted, as in previous criteria, that absence of progression excluded a diagnosis of ALS, and the grades ''possible", ''probable" and ''definite" were abandoned because they were commonly used to define the extent of involvement rather than diagnostic certainty.The utility of the Gold Coast criteria has since been demonstrated (de Jongh et al., 2022;Hannaford et al., 2021;Jewett et al., 2022).The interested reader is referred to the original report (Shefner et al., 2020).It should be mentioned that the motivation for the Consensus Meeting that led to these diagnostic criteria was to ensure early treatment and inclusion in clinical trials and less for confirmation of a definite clinical diagnosis of ALS.However, even for the early clinical diagnosis, the Gold Coast criteria seem to be appropriate.

Monitoring progression
It is critical for the management of patients to establish reliable biomarkers of disease, so that clinicians can monitor progression and, in clinical trials, demonstrate any benefit of therapy.
In this chapter, we provide an overview of conventional and emerging neurophysiological techniques in the evaluation of LMN dysfunction in ALS.This includes techniques that investigate the number of surviving motor neurons in limb muscles, i.e., motor unit number index (MUNIX), and motor unit number estimation (MUNE) methods, the impact on the overall muscle composition via electrical impedance myography, and the complementary role of neuromuscular imaging in ALS.Techniques that address abnormalities of upper motor neuron dysfunction including cortical excitability are considered in other chapters.

Needle EMG
Needle EMG provides information about denervation activity, i.e., fibrillation potentials, positive sharp waves, and spontaneous motor unit activity, i.e., fasciculation potentials.Accordingly, information about the size, morphology, and stability of motor unit potentials (MUPs) can be extracted and the presence of reinnervation can be detected.However, the degree of denervation and reinnervation does not correlate with the magnitude of motor unit loss.Moreover, needle EMG, particularly quantitative motor unit potential analysis is time-consuming and invasive and, even though it is performed by experts, variability from one session to the next is high because one cannot record from the same motor units with certainty in successive studies.Thus, while needle EMG is important in diagnosing ALS, it is not a particularly useful biomarker in clinical trials, and it has been used as an outcome measure in only a few drug studies.Artificial intelligence-based analysis may help overcome the above-mentioned limitations of needle EMG, and in future clinical trials, needle EMG could be a satisfactory outcome measure.
The only study in which needle EMG showed significant changes was a pilot study of autologous adult stem cells in three patients with ALS where the MUP parameters demonstrated improvements (Bansal et al., 2016).Desai and co-workers (Desai et al., 1998) did not find any change in MUP characteristics before and after riluzole therapy in a randomized, controlled, blinded, and single-centre study on 15 patients with ALS.In another study (Bromberg et al., 2001), needle EMG was used as an outcome measure along with MUNE and compound muscle action potential (CMAP) amplitude on 95 patients with ALS in a negative randomised blinded drug trial of branched chain amino acids therapy.Perhaps not surprisingly, the needle EMG did not show any changes.In four other negative studies in smaller cohorts, autologous peripheral blood mononuclear cells (Li et al., 2017), bone marrow mononuclear cells (Geijo-Barrientos et al., 2020), thyrotropin-releasing hormones (Mitsumoto et al., 1986), dextromethorphan, an N-methyl-D-aspartate receptor antagonist (Askmark et al., 1993) and granulocyte-colony stimulating factor (G-CSF) (Cashman et al., 2008) failed to show treatment-induced changes in the needle EMG findings in patients with ALS -again an unsurprising result in negative studies.

Nerve conduction studies (NCS)
Much as with needle EMG, NCS are important in refining the differential diagnosis, by excluding, e.g., demyelinating neuropathies with or without features of conduction block, important in the differentiation of ALS from multifocal motor neuropathies (MMN).However, NCS in ALS may also show prolonged latencies and decreased conduction velocities due to the loss of fastconducting axons.As an outcome measure in clinical trials, maximal CMAP amplitude is the most often used NCS parameter.CMAP amplitude can reflect the number of functioning axons, but it is not always an accurate measure of axonal loss for two reasons (i) the precise placement of the surface electrodes affects reproducibility of the CMAP, and (ii) the healthier axons have the capacity to innervate muscles supplied by their neighbouring dying motor units, a phenomenon referred to as collateral sprouting.As a result, CMAP amplitude may remain normal until 50% or more of motor units are lost.Nevertheless, CMAP amplitude has been used as an outcome measure in several clinical trials of ALS.In some of these clinical trials, CMAP amplitude showed positive results (Geijo-Barrientos et al., 2020;Gold et al., 2019;Kovalchuk et al., 2018;Park et al., 2015) whereas in others, no changes have been shown (Bromberg et al., 2001;Li et al., 2017;Mitsumoto et al., 1986;Nefussy et al., 2010).
Diaphragmatic motor response to phrenic nerve stimulation is another important NCS measure.Diaphragmatic CMAP amplitude has been shown to predict functional decline (Miranda et al., 2020) and survival (Pinto et al., 2012).

CMAP indices
The CMAP amplitude change is investigated by calculating indices of measurements made during routine NCS.The 'cumulative muscle index (CMI)' summated the CMAP amplitude from 6 different muscles: APB, ADM, biceps, tibialis anterior, abductor hallucis and extensor digitorum brevis.The CMI had good reproducibility and decreased faster than the revised ALS Functional Rating Scale (ALSFRS-R) in longitudinal studies (Nandedkar et al., 2015).
The split hand index (SHI) is based on clinical observations of significant weakness and/or atrophy of the first dorsal interosseous (FDI) and APB muscles compared to that of the ADM (Wilbourn, 2000).The index is calculated as follows: (Menon et al., 2013) SHI ¼ ðFDI CMAP amplitude Ã APB CMAP amplitudeÞ = ðADM CMAP amplitudeÞ The SHI is reduced most prominently in ALS patients with upper limb onset (Hannaford et al., 2021).CMAP amplitude will decrease with MU loss.With fewer MUs to generate F waves, the persistence of F waves is reduced.The CMAP onset latency also tends to be higher when its amplitude is significantly reduced.These observations were combined into a single result called the neurophysiologic index (NI) (de Carvalho et al., 2005a;de Carvalho and Swash, 2000;Swash and de Carvalho, 2004).

NI ¼ ðCMAP amplitude
Â F wave persistenceÞ = CMAP onset latency NI has been both a positive (Gold et al., 2019;Park et al., 2015) and a negative (de Carvalho et al., 2010;Mazzini et al., 2010;Nefussy et al., 2010) outcome measure in drug trials on patients with ALS.It has been shown that NI declines faster than CMAP amplitude (de Carvalho et al., 2005b) and it seems to be a predictor of survival (Cao et al., 2019).

Reflex studies in MND/ALS
As part of the upper motor neuron (UMN) syndrome, cutaneomuscular reflexes are often clinically abnormal in ALS.The super-ficial abdominal and cremasteric reflexes are commonly absent, but the more generalised UMN signs such as the Babinski response can be difficult to elicit clinically (Swash, 2012;Swash et al., 2020).Neurophysiological tests that indicate UMN weakness, hyperreflexia and cerebral involvement can therefore be of great value in diagnosing ALS.

UMN contribution to weakness
When a patient can have both an UMN and a LMN component to their weakness, as in MND/ALS, clinical evaluation may not provide a clear answer.For example, with a footdrop and denervation in tibialis anterior, it may be difficult to establish whether there is also an UMN component to the weakness.Here the patient can be asked to perform a maximal voluntary dorsiflexion against resistance, recording the surface EMG over tibialis anterior.A noxious stimulus is then delivered to the sole of the foot during the maximal effort.If the subject is contracting maximally and the noxious reflex input produces a significant increase in EMG and greater dorsiflexion, there is likely to be an UMN contribution to the weakness.

H reflex studies
H reflexes are valuable diagnostic tools because of their sensitivity to pathology that decreases or disperses the afferent input.The reflex has a threshold and then increases in proportion to the size of the afferent volley.It is easily lost in pathology because it requires a sufficiently synchronized volley of sufficient intensity to reach the motor neuron pool, i.e., the H reflex is very sensitive to loss of fast conducting axons or minor degrees of conduction slowing.The physiology and practical use of the H reflex and neurophysiological testing in MND/ALS are the subjects of previous reports (Burke, 2016(Burke, , 2022)).
2.3.2.1.Test procedures.At rest: Place the cathode over nerve, anode remote.This is important for quadriceps, extensor carpi radialis and biceps brachii to minimize discomfort (because the nerves are deep), but it is less important elsewhere.Alternatively, one can use a bipolar orientation, with the cathode proximal.A stimulus width of 1 ms exploits the longer strength-duration properties of large afferents over motor axons.Start at 1 Hz, define H reflex or M wave, then drop the stimulus rate to once/3 sec or once/5 sec to minimise homosynaptic depression.(A slower rate is theoretically better but too tedious in clinical practice.)If the H wave cannot be recorded, return to M wave threshold, ask the patient to perform steady contraction and increase the stimulus rate to 2-3 Hz.
During voluntary contraction: A contraction minimizes homosynaptic depression, and rate is then not as important (though 2-3 Hz might be uncomfortable, particularly for deep nerves (innervating quadriceps, extensor carpi radialis, biceps brachii [at Erb's point]): Average unrectified surface EMG against the stimulus.Average as many trials as required to define the H reflex clearly (usually x32-64), superimposing repeated averages.
In subjects at rest, H reflexes can be recorded for soleus, quadriceps (vastus lateralis) and flexor carpi radialis.In < 10% of subjects they can also be recorded for extensor carpi radialis.For most other muscles, they can only be demonstrated during a weak voluntary contraction of the test muscle.A weak voluntary contraction is the best method of reflex reinforcement: it raises the motor neuron pool to firing threshold and it turns off Ib inhibition from Golgi tendon organs in the test muscle (and it minimises homosynaptic depression).

H reflex studies in MND/ALS
In patients suspected of having MND/ALS, it may not be particularly helpful to test the H reflexes of muscles that have an H reflex at rest.It may be more convenient to test tibialis anterior, abductor pollicis brevis and extensor carpi radialis bilaterally.If the reflex is obtained at rest for these muscles, the patient is hyperreflexic for that reflex arc (Fig. 1).If there are LMN findings on EMG for that muscle, this is evidence of UMN and LMN abnormalities for the one muscle.In the patient of Fig. 1 The H reflex can be useful in MND mimics.If the reflex is obtainable only during contraction, but appears at normal latency for age and height, one can conclude that conduction values for sensory and motor axons over the long afferent and efferent pathways are within normal limits, i.e., that the studies have not revealed a proximal lesion affecting sensory or motor axons.If MMN is a possibility, H reflex testing of multiple muscles (with and without a demonstrable reflex at rest) can be valuable screening.

Conclusion
Needle EMG is still essential for diagnosis.The importance of NCS is largely the exclusion of ALS mimics, though H reflex studies can also be useful in some patients in demonstrating hyperreflexia.

Introduction
Although needle EMG is an invaluable tool for the diagnosis of neuromuscular disorders, it is not convenient for disease monitoring due to its discomfort and the variability in needle positioning Fig. 1.H reflexes recorded at rest in a 49-year old male patient, with clear lower motor neuron (LMN) changes in both legs on EMG.There were multiple fasciculation potentials in right and left tibialis anterior, right medial gastrocnemius and left vastus lateralis, and the fasciculation potentials were often complex and occasionally unstable.Fibrillation was noted in the right tibialis anterior and the tongue.There were excessive polyphasic motor units in left tibialis anterior and left extensor carpi radialis, with motor units that were enlarged in duration and amplitude in right and left tibialis anterior, right medial gastrocnemius and left vastus lateralis.In each panel, the stimulus intensity was increased progressively from top to bottom.In the second trace for the left tibialis anterior, the H reflex is larger than the M wave.In the right tibialis anterior, the H reflex grows as stimulus intensity increases, larger in each trace than the M wave.In left extensor carpi radialis, the H reflex appears below threshold for the M wave in the first and second traces and is larger than the small M wave in the third trace.These reflex studies demonstrated hyperreflexia, in that H reflexes could be recorded at rest in right and left tibialis anterior (and left extensor carpi radialis (ECR)-though that can occur normally in < 10% subjects).
C. Shin-Yi Lin, J. Howells, S. Rutkove et al. Clinical Neurophysiology 162 (2024) 91-120 between recordings.Therefore, sEMG has long been of interest as a monitoring biomarker for ALS due to its advantages of being practical, non-invasive, and not unpleasant (Bashford et al., 2020a).However, sEMG has limitations in the detection of insertional activity, spontaneous activity, motor unit potential analysis, and interference pattern evaluation.Thus, in contrast to its potential value as a biomarker in disease monitoring, sEMG may only be considered as a supplementary test to needle EMG for diagnosis.
Single-channel and high-density surface EMG (HDSEMG) are the two main applications of sEMG.CMAP amplitude, MUNE, and MUNIX are also considered single-channel sEMG methods.These are discussed in the related Section 4 on MUNE and MUNIX.
In this section, the utility of single-channel sEMG and HDSEMG in the detection of fasciculation potentials, evaluation of MUP characteristics, and some of their other aspects will be summarized.

Detection of fasciculation potentials
In one study, 3-min sEMG recordings from the forearm muscles were compared with clinical observations in 26 ALS patients (Mateen et al., 2008).Fasciculation potentials measured by sEMG and clinical observation increased as strength deteriorated.However, the fasciculation potential frequency was not correlated with disease duration.The conclusion of this study was that fasciculation potentials have diagnostic but not prognostic utility in ALS.
de Carvalho and Swash (de Carvalho and Swash, 2016) could not show any change in fasciculation potential frequency or amplitude during up to 12 months follow-up period in 34 ALS patients and 19 ALS-mimic disorders although the neurophysiological index declined suggesting loss of motor units.Analysis of a single muscle and the use of single-channel sEMG may be the cause of these negative results.

MUP analysis
In early studies, sEMG showed comparable measurements to needle EMG of MUP characteristics.In MND, large sEMG amplitudes but normal-size twitch tensions were observed (Milner-Brown et al., 1974a) while the orderly pattern of recruitment was not disrupted (Milner-Brown et al., 1974b).

Other applications
Hallett (Hallett, 1979) recorded sEMG patterns from the biceps and triceps muscles during stereotyped ballistic elbow flexion movements, i.e., contractions performed as rapidly as possible.The first agonist burst, and the first antagonist burst, which are normally limited in duration, were prolonged in ALS patients with clinical signs of UMN involvement.It was suggested that prolongation of these components was necessary for the muscles to generate sufficient force to accomplish the movements.
Zhang and Zhou (Zhang and Zhou, 2014) performed sEMG to detect hidden muscle activity and found that multi-scale entropy analysis was more sensitive than power spectral density analysis in distinguishing ALS patients from healthy controls.
In one study, an electrical stimulus (20 Hz, 600 stimulations) to cutaneous afferents in the radial nerve demonstrated an increase in resting fasciculation potentials in weak first dorsal interosseous muscles in ALS patients whereas there was no change in patients with benign fasciculation syndrome (de Carvalho et al., 2019).This study suggests that in ALS but not benign fasciculation syndrome, some fasciculation potentials arise at motor neuron level rather than peripherally at, e.g., the neuromuscular junction, and can be triggered by afferent inputs.
A recent study investigated the use of a set of dynamic features extracted from sEMG to study UMN degeneration.Classification accuracies of up to 94% could be achieved for ALS using features based on detrended fluctuation analysis and peak frequency, and classifiers such as decision trees, random forest, and Adaboost (Quintao et al., 2021).In another recent study, multidimensional facial sEMG analysis for objective assessment of bulbar involvement could identify bulbar muscular changes in ALS, related to both UMN and LMN pathologies including decreased motor unit recruitment and synchronization for jaw antagonists, a potential neuromuscular adaptation for temporalis muscle (Rong and Pattee, 2022).These studies provide potential novel applications of sEMG, but they suffer from small sample sizes: both studies included only 13 ALS patients and 20 healthy controls.

High-density surface EMG (HDSEMG)
For HDSEMG, closely spaced two-dimensional multi-electrode arrays in a fixed grid formation located on the surface of the skin are used (Fig. 2).Advanced processing and recording of different C. Shin-Yi Lin, J. Howells, S. Rutkove et al. Clinical Neurophysiology 162 (2024) 91-120 temporal and spatial signal EMG amplitudes through multichannel electrodes have provided automated detection and accurate characterization of fasciculation potentials (Drost et al., 2007) and MUP characteristics (Chen et al., 2018).The non-invasive character and the unique spatiotemporal information gained from HDSEMG enables long stable recording times and provides quantitative information.The recordings may be done simultaneously with needle EMG (Fig. 2).
3.3.1.Detection of fasciculation potentials sEMG is a sensitive method for detecting fasciculation potentials in patients with ALS and was shown to be superior to clinical evaluation and needle EMG in 1973 using multi-lead surface electrodes in 14 patients with ALS (Hjorth et al., 1973).Later, these results were confirmed in a larger cohort of 43 patients with MND using eight channels of a conventional electroencephalograph with plate electrode recordings from upper and lower extremity proximal and distal muscles for 20 minutes recordings (Howard and Murray, 1992).Kleine et al. (Kleine et al., 2008) recorded two different firing patterns and refractory periods of fasciculation potentials recorded by HDSEMG for 15 minutes in 10 ALS patients, and proposed that these two different types represent axonal and spinal motor neuron origins.Sleutjes et al., (2016) (Sleutjes et al., 2016a) proposed a novel approach based on 5-10 minutes HDSEMG recordings of the thenar muscles at rest in 30 patients with ALS and 14 healthy controls.Fasciculation potentials were characterized as ''isolated" when the inter-spike intervals before and after were > 250 ms and ''continual" when both inter-spike intervals were 250 ms.Firing patterns in ALS were more frequently dominated by isolated discharges and less frequently by continual discharges compared with controls.Furthermore, in patients, more fasciculation potentials were simultaneously active.
Recently, an automated algorithm, surface potential quantification engine (SPiQE) could detect a median fasciculation potential frequency of 65/min from biceps and gastrocnemius muscles (Bashford et al., 2019).The group then went on to use an automated Active Voluntary IDentification (AVID) strategy to exclude voluntary muscle activity from HDSEMG recordings (Bashford et al., 2020b).Later, the same group showed that fasciculation potentials that were analysed using the SPiQE at a single timepoint could enhance prognostic models in ALS, where a higher rate of change of fasciculation potential frequency values indicated shorter survival (Wannop et al., 2021).Recently, HDSEMG using SPiQE and a Gaussian mixture model algorithms were combined with neuromuscular ultrasound simultaneously to explore the electromechanical properties of fasciculation potentials.Electromechanical latency was prolonged in ALS but did not correlate with clinical measures or fasciculation potential frequency.In conclusion, prolonged fasciculation electromechanical latency indicated impairment of the excitation-contraction coupling mechanism, warranting further exploration as a potential novel biomarker in ALS (Planinc et al., 2023).
Electrode configuration and filtering may also play a role in detection of fasciculation potentials.Jahanmiri-Nezhad et al. showed that monopolar sEMG electrode configuration was superior to bipolar configuration (Jahanmiri-Nezhad et al., 2014a), and that the spatial filtering of sEMG dramatically reduced the detection sensitivity and thus was not recommended (Jahanmiri-Nezhad et al., 2014b).Another important aspect is the duration of the recordings.In one study, up to 70 s recordings had a probability approaching unity and increasing the recording time to two minutes increased the probability of recording five fasciculation potentials to 95% (Zhou et al., 2012).Crook-Rumsey et al. (Crook-Rumsey et al., 2022) showed that for the detection and quantifica-tion of fasciculation potentials in patients with ALS and benign fasciculation syndrome, SPiQE HDSEMG recordings could be halved from 30 to 15 min without significantly compromising the primary outputs.
Using a machine learning approach, the combination of three novel HDSEMG markers including the clustering index, the kurtosis of EMG amplitude histogram, and the kurtosis of EMG crossing-rate expansion during different levels of voluntary muscle contraction could distinguish ALS patients and healthy controls with 90% sensitivity and 100% specificity (Zhang et al., 2014).These measures were based on characteristic changes in voluntary MUP features.Given the high diagnostic yield, the proposed surface EMG analysis was suggested as a supplement to needle EMG examination in supporting the diagnosis of ALS.It should be noted that the specificity was for ALS versus health controls, not disease mimics.
A recent study compared MUP parameters in 16 patients with ALS and 16 control subjects.The participants performed ramp-up and sustained contractions at 30% of their maximal voluntary power.HDSEMG signals were decomposed into individual MU firing behaviour using a convolution blind source separation method, and abnormal MU firing behaviour was suggested as an important physiological index for understanding the pathophysiology of ALS (Nishikawa et al., 2022).Patients with ALS had higher motor unit firing rates than age-matched control subjects, and motor unit firing rate at recruitment could differentiate people with ALS from controls.Whether these findings are specific to ALS or merely reflect motor neuron fallout can be debated.

Other applications
Multiplet discharges are defined as a train of identical or nearidentical MUPs with inter-spike intervals < 30 ms in response to proximal electrical stimulation, and can be characterized as doublets, triplets, or quadruplets in ALS.Multiplet discharges are proposed to originate solely from the distal parts of the axon because the ectopic activity is triggered by electrical stimulation, while fasciculation potentials can originate from the terminal axons to the soma and supraspinal regions, and more central, presumably at or near the motor neuron, and therefore may be a more specific biomarker for ALS.Maathuis and co-workers (Maathuis et al., 2012) performed HDSEMG of the thenar muscles in 10 ALS/PMA patients, five recordings per patient over a three-month period, and found multiplet discharges in 94% of patients.The interspike interval ranges of the multiplet discharges were 2.9-6.5 ms, which were compatible only with a distal origin providing evidence for pathophysiological excitability changes in the most distal part of the motor neuron.In this study, multiplet discharges were associated with clinical deterioration, and it was suggested that multiplet discharges may be a diagnostic and monitoring biomarker for ALS.However, in a later study, only 43% of 21 ALS patients demonstrated multiplet discharges in the thenar muscles (Sleutjes et al., 2015) which may be due to methodological differences between these two studies.In a later study, stimulated HDSEMG data were collected from 31 MND patients (22 ALS and 9 PMA) at baseline and at a maximum of 16 weeks follow-up (Sleutjes et al., 2016b).The presence of multiplet discharges was associated with a more marked clinical deterioration suggesting the potential utility of multiplet discharges as a prognostic biomarker.
Neuromuscular architecture has been evaluated in several studies.In one study, surface muscle fibre conduction velocity record-ings were performed in 22 ALS patients.Muscle fibre conduction velocities in biceps brachii were faster in ALS than healthy controls (Van der Hoeven et al., 1993), presumably reflecting larger muscle fibres in ALS, hypertrophied to compensate for the weakness.Later, HDSEMG was used to examine the so-called innervation zone of motor units supplying an individual muscle motor unit with the hypothesis that innervation zone characteristics may change and therefore provide a useful marker of ALS through changes due to denervation and reinnervation processes.In one study, the length of the motor unit innervation zone was increased in 12 ALS patients compared to 7 healthy controls (Jahanmiri-Nezhad et al., 2015).These two studies suggest that changes in the neuromuscular architecture in ALS patients may be potential biomarkers.
A search for F waves associated with fasciculation potentials was undertaken in one study which compared fasciculation potentials from 15-min HDSEMG recordings in 7 ALS patients and 7 patients with benign fasciculation syndrome.Pairs of fasciculation potentials with an inter-fasciculation potential interval between 10-110 ms were searched using template matching with the hypothesis that the inter-fasciculation interval might indicate that the second discharge was an F-wave and therefore that the first fasciculation potential originated from distal terminal axons.However, F-waves of fasciculation potentials occurred in both conditions without any group difference (Kleine et al., 2012).

Conclusion
sEMG has the advantage of being non-invasive, and it does not require electrical stimulation.Thus, while sEMG has a limited role in diagnosis, it is a promising biomarker for monitoring MND/ALS.Technological developments, fast and automated methodologies and recordings from multiple muscles will make sEMG an attractive tool for studying disease progression particularly in drug trials.However, a main limitation of HDSEMG is the need for a dedicated device and software.Technological developments may also help this limitation.

Introduction
In principle, the number of functioning motor units (MUs) in a muscle should provide a useful measure of denervation and for assessing potential treatments of MND.Unfortunately, not all limb muscles can be tested with currently available methods, so that their value for diagnosis may be limited in early disease with greater involvement of other limbs, e.g.testing upper limb muscles in lower-limb onset disease.Then abnormalities would indicate more widespread involvement than apparent clinically.However, measures of motor unit number should be of value in following the progression of ALS/MND, and this has led to the development of a variety of methods of MUNE.MUNE was pioneered by McComas and co-workers (McComas et al., 1971) more than five decades ago, using a method now known as incremental stimulation MUNE.They recorded the CMAP from extensor digitorum brevis muscle using standard nerve conduction methodology (Fig. 3A).The CMAP is the sum of surface recorded motor unit potentials (SMUPs) of all MUs in the muscle.After measuring the baseline to peak amplitude, the stimulus is reduced to zero, and then gradually increased.Initially, the intensity is too low to stimulate any axon and the signal baseline is recorded (Fig. 3B).A stage is then reached when the intensity is just enough to stimulate a single axon, and the SMUP from that MU is recorded.The ''All-or-None" pattern of this response confirms the stimulation of a single axon.With further increase in intensity, a stepwise increase in response occurs when a second axon is stimulated.The signal is then the sum of two SMUPs.Further increments in amplitude occur as more axons are stimulated with further increases in intensity.Thus, Fig. 3B shows eight steps, with several trials superimposed, and appears to show sequential stimulation of eight axons.If the summated signal amplitude from eight SMUPs is 1200 lV, giving an average SMUP amplitude of 150 lV, then dividing this into the CMAP amplitude of 15 mV (=15000 lV) yields a motor unit number estimate, or MUNE, of 100.This was the beginning of MUNE and a very exciting development in electrodiagnostic studies.C. Shin-Yi Lin, J. Howells, S. Rutkove et al. Clinical Neurophysiology 162 (2024) 91-120 Unfortunately, not all steps are due to the recruitment of separate motor units.Axon thresholds are variable, as documented by Bergmans (Bergmans, 1970) and when two thresholds (e.g., A and B) overlap, as they soon do as more units are recruited, they give rise to more than two steps (A, B and A + B).Similarly, three overlapping thresholds can give rise to seven steps (A, B, C, A + B, A + C, B + C, A + B + C) etc.This phenomenon of alternation confounds a simple interpretation of incremental MUNE recordings beyond the first few units.In addition, the method is based on the insecure assumption that the size of the first-recruited units is similar to that of higher-threshold units.Therefore, alternative methods of recording SMUP amplitudes have been developed.In the multi-point stimulation method (Doherty and Brown, 1993) the nerve is stimulated at different sites along the nerve.Later, adapted multi-point stimulation MUNE was introduced by stimulation of the nerve at three different sites, and at each site, incremental stimulation method is used to reach the third step.Thus, nine different SMUPs are sampled (Fig. 3C).The amplitude of the third step from each site is added and the sum divided by nine as average SMUP amplitude (110 lV).Dividing this into the CMAP amplitude (9637 lV) gives the MUNE (=88).Another method of SMUP extraction uses the F wave recording technique (Stashuk et al., 1994).Repeating F waves represent individual SMUP.
In the spike triggered averaging method an intramuscular needle was used to identify activity from a single MU (Brown et al., 1988;Nandedkar and Barkhaus, 1987).The time-locked surface EMG signals were averaged to obtain the SMUP.The needle position was changed to record several different SMUPs.This is similar to recording Macro EMG MUPs (Stålberg, 1980).The Macro EMG needle has also been used for MUNE (de Koning et al., 1988).
None of these early MUNE methods, based on estimating a mean SMUP amplitude, have been used widely for several reasons.They are all labour-intensive and unsuited to automation.Secondly, the low-amplitude signals used for analysis are often difficult to record due to ambient noise and interference.For these reasons the methods are time-consuming.A further problem is that they are based on the assumption that the limited number of units sampled are representative, but whether they are recruited voluntarily or by electrical stimulation, this may not be the case.Indeed, in voluntary contractions the first-recruited motor units generally have small slow axons and small SMUPs, but with electrical stimulation the bias is, in general, to activate large axons preferentially, although proximity to the skin surface may be more important.Extrapolation from a small sample of units also assumes that SMUPs do not vary much in size.This is reasonable for a healthy nerve but is unrealistic when neurodegeneration occurs.

CMAP scan analysis
A characteristic feature of degenerative MND is that denervated muscle fibres attract collateral reinnervation by surviving motor axons, which thereby enlarge their SMUPs.It is because of this process that CMAP amplitude is a less sensitive indicator of denervation, and MUNE methods have been developed.Blok and coworkers (2007) (Blok et al., 2007) extended the concept of incremental stimulation to scan the entire CMAP.The stimulus intensities that give a small response and the maximal response are measured.Next, the nerve is stimulated by decreasing the intensity in 500 even steps from maximum to minimal levels (Fig. 3D).In a normal muscle, the plot of amplitude versus stimulus intensity has a smooth sigmoidal shape (Fig. 3E).With loss of some motor units and enlargement of others, the curve shows discontinuities.Large discontinuities are called ''steps" and provide a measure of the extent of collateral reinnervation.The scan is a visually appealing method to demonstrate MU loss, and the following parameters have been used for semi-quantitative analyses of CMAP Scans (Blok et al., 2005): 1) Stimulus intensities (SIs) that elicited 5%, 50%, and 95% of the maximal CMAP (S5, S50, and S95, respectively), 2) Absolute SI range (S95-S5), 3) Relative SI range (S95-S5)/S5) and 4) Step percentage (step%).C. Shin-Yi Lin, J. Howells, S. Rutkove et al. Clinical Neurophysiology 162 (2024) 91-120 The CMAP scan does not directly yield a MUNE value, but in addition to the step percentage, it has given rise to other measures that provide indirect indices of motor unit loss.

Amplitude analysis
Sleutjes and co-workers (Sleutjes et al., 2014) developed a measurement called D50 for CMAP scan analysis.The baseline to negative peak amplitude of each response in the scan is measured (Fig. 4A).In this example, the stimulus intensity is reduced from supramaximal to minimal during the scan.We expect the amplitude to decrease monotonically during the scan.However, such a pattern is not seen.In some stimuli, the signal amplitude increased slightly although the intensity was slightly reduced.This is inherent due to excitability changes in the axon during successive stimulations (Sirin et al., 2019), and may also occur if there is phasecancellation between adjacent MUPs in the scan.The amplitude values are sorted and plotted in ascending order.The curve shows monotonic increase in amplitude.Note the difference in successive amplitudes (steps) (Fig. 4B) and the variation of step amplitude during the scan (Fig. 4C).There are a few large steps and many with low amplitude.The steps are arranged in descending order of amplitude and added to reconstruct the CMAP amplitude (Fig. 4D).The number of largest steps that make up half the CMAP amplitude is called D50.When there is MU loss, D50 will be less.
Nandedkar and Barkhaus (Nandedkar et al., 2022) plotted the step amplitudes in a descending order.It shows a low amplitude tail mainly due to noise in recording.By extrapolating the distribution of large amplitude steps, the number of steps due to physiological response was estimated.The result is called the Step index (STEPIX) (Fig. 4E).Dividing STEPIX into the CMAP amplitude gives amplitude index (AMPIX).These measurements are based on a regression model.Hence the results are considered indices, not estimates of the number of MUs.When the number of MUs is reduced STEPIX will decrease (Fig. 4F).Similarly, AMPIX will increase when MUs are larger due to reinnervation.
More sophisticated analyses of the CMAP scan, i.e., Bayesian MUNE and MScanFit MUNE, which take explicit account of the probabilistic nature of MU thresholds, are best considered as types of statistical MUNE.

Daube's statistical MUNE
The first method designed to simplify MUNE by measuring CMAPs rather than individual MUPs was introduced by Daube (Daube, 1995).This method assumed that the variations in amplitude occurring with a submaximal stimulus would follow a Poisson distribution, in which the mean amplitude of the components is equal to the variance.This method had been used successfully by Hubbard and others (Henderson et al., 2003;Hubbard et al., 1969) to estimate the number of quanta of transmitter released from a nerve terminal at the neuromuscular junction.In Daube's method a crude CMAP scan is first recorded, and on the basis of that 30 stimuli are delivered at $ 10 trials at four different current levels, with up to 500 submaximal stimuli recorded in five minutes.The analysis of amplitude variability, based on Poisson statistics, gave an estimate for the SMUP amplitude and MUNE.The approach was received with enthusiasm.However, the Poisson statistics as used by Hubbard were, like the single MU methods described above, based on the assumption that each of the MUs is of similar size, which CMAP scans demonstrate is far from being the case in neurodegenerative disease.Amplitude variability can also arise from secondary neuromuscular transmission abnormalities in MND (Alanazy et al., 2017).Various suggestions have been made to improve Daube's method (Bromberg et al., 2003;Lomen-Hoerth and Slawnych, 2003;Shefner et al., 1999), including the replacement of Poisson by binomial statistics (Blok et al., 2005), but use of these simple statistical methods has declined in favour of other methods that are equally quick to perform.

Bayesian statistical MUNE
An alternative and much more sophisticated statistical method, which takes into account every response to every stimulus in a CMAP scan was proposed by Ridall, Henderson and colleagues (Henderson et al., 2007;Ridall et al., 2006;Simon et al., 2015).They generated a model that incorporates the variability of the threshold of each MU, the variability between and within single motor unit action potentials and baseline variability.Their method, involving Markov chain Monte Carlo simulation, and a Bayesian approach to estimate the most likely model that could generate the CMAP scan, is unfortunately time-consuming.Although a 500-point CMAP scan can be generated within minutes, the numerical methods used in the analysis are complex and 'extremely long runs, taking 5 to 8 hours on a modern laptop computer', are required to estimate the number of units with reasonable accuracy (Drovandi and Pettitt, 2013).For this reason, and for lack of easily available software, this method has had limited use.

MScanFit MUNE
To overcome these limitations of the Bayesian analysis, while retaining the advantages of their probabilistic model of motor unit firing, a much faster method of CMAP scan analysis was proposed by Bostock (Bostock, 2016) and the software made freely available.The method was designed to generate a model, with the same variables as the Bayesian MUNE, that would generate a CMAP scan as similar as possible to the recorded one, and is illustrated in Fig. 5.To compare the two scans (modelled and recorded), they are smoothed in the two directions of stimulus and amplitude and converted to contour maps, with contour 'height' corresponding to the probability of a particular stimulus generating a particular amplitude of response (e.g., Fig. 5B).The 'error' of a model is then scored as the percentage difference in contour heights averaged over the full stimulus range.Because of the probabilistic nature of motor unit thresholds, the same model can generate an infinite number of different model scans with different error scores, so to test each putative model, several scans are generated, and the model judged on its lowest error score.
The generation of an MScanFit model takes place in two stages.From the CMAP scan (e.g., Fig. 5A), a preliminary model is first generated from the slopes and variances of different segments of the CMAP scan, using the relationship: MU amplitude ¼ Variance Â p p = ðslope Â stimulus Â qÞ where q is the 'relative spread' of thresholds, or standard deviation of threshold as a percentage of the mean.(A starting value for q of 2% was found to work well, although a study of threshold variability in single human axons had found an average of 1.65% (Hales et al., 2004)).Fig. 5C illustrates the preliminary model generated in this way, with each unit represented by its threshold distribution, peaking at the unit amplitude.
The preliminary model generates a scan (Fig. 5D) with a similar overall shape to the recorded scan, but it is inaccurate in detail.The inaccuracies are measured from the difference in contours generated by the original scan and the model (Fig. 5E).To improve the model, various changes in the model units are tested, including increasing and decreasing its three parameters (amplitude, mean threshold and q) and also splitting one unit into two to increase the number of units, and combining two adjacent units to reduce the number.For each change, up to 20 simulated scans are generated to test for a reduction in error score.This refinement process is continued until no more improvements in the model are obtained, or a time limit (default six minutes) is reached.The refined model is illustrated in Fig. 5F, and a scan generated from it in Fig. 5G.This corresponds closely to the original scan in Fig. 5A, as indicated by the small contour differences in Fig. 5H.CMAP Scans may be recorded by any conventional EMG machine and then can be analyzed using a freeware MScanFit analysis program or Qtrac software (Ó University College London, UK, distributed by Digitimer Ltd.).Qtrac can also be used for recordings that enable the recording of exponential CMAP Scans.Exponential scans (with stimuli decreased in 0.2% steps) are preferred to the linear scans used by Blok et al. (Blok et al., 2007), since they enable several stimuli to be delivered within the threshold range of each unit.
There are several parameters that can be extracted from the MScanFit MUNE analyses as listed below: 1) NUnits: the number of motor units, 2) MeanUnitAmp (lV): the mean size of the motor units in lV, 3) MedianAmp (mV): the median size of the motor units in mV, 4) MedianAmp (%): the median size of the motor units in %, 5) LargestAmp (mV): the size of the largest motor unit in mV, 6) LargestAmp (%): te size of the largest motor unit in %.Additionally, the following parameters can be extracted after plotting the cumulative amplitude vs unit amplitude: 1) NLarge (N50): the number of the largest motor units that make up half of the CMAP, 2) HalfAmpAmp (mV) (A50): the size of the motor unit at the 50% mark of the cumulative amplitude (ie the smallest of the N50'large' units), 3) HalfAmpAmp (%): the size of the motor unit at the 50% mark (A50), expressed as a percentage of the CMAP.
All existing MScanFit studies in MND are summarized in Table 1.
MScanFit MUNE has been shown to be reproducible in different muscles in healthy individuals (Jacobsen et al., 2017;Tankisi et al., 2022).The only reproducibility study in patients has been performed in ALS patients which showed high reproducibility.Similarly, MScanFit MUNE has shown high sensitivity, and good ability to determine the disease progression (Jacobsen et al., 2017;Jacobsen et al., 2019).In several studies in patients with ALS, MScanFit showed higher sensitivities in detecting motor unit loss than conventional electrophysiological tests and clinical scores (Gunes et al., 2021;Jacobsen et al., 2018;Oguz Akarsu et al., 2020;Sirin et al., 2019).In a multi-centre study, MScanFit MUNE reduced the sample size by 19.1% for a 6-month study compared to the ALSFRS-R (Sleutjes et al., 2021).In a recent study (Stikvoort Garcia et al., 2023), the combined diagnostic yield of nerve excitability and MScanFit MUNE showed high diagnostic accuracy in ALS.The authors suggested the combination of these two methodologies to improve early diagnosis in clinically challenging patients with suspected ALS and to generate baseline measures for clinical trials.In two studies in SMA, one in children (Kariyawasam et al., 2020) and the other in adults (Schneider et al., 2021), the utility of MScanFit in monitoring disease progression during nusinersen treatment has been shown.Overall, the existing studies suggest MScanFit as a potential biomarker in monitoring ALS and SMA in drug trials.In an ongoing multicentre study with the participation of 15 centres, the sensitivity of MScanFit in the diagnosis of ALS and monitoring disease progression are being tested.
A recent ''Round Robin" study on applying MScanFit to 3 different muscles in healthy subjects has shown dependence on CMAP amplitudes, which were recorded differently by different operators, especially for tibialis anterior muscle (Sørensen et al., 2021).A modification to allow for such variations in CMAP amplitude measurement has been released as MScanFit-2, and is recommended for multi-centre studies (Sørensen et al., 2023).The origi-nal MScanFit (MScanFit-1) options are suitable when electrode placement is consistent, and normal CMAP amplitude $10 mV (measured from baseline to negative peak).

MUNIX
Most MUNE methods use electrical stimulation to assess the SMUP characteristics.MUNIX uses surface EMG activity for that purpose (Nandedkar et al., 2010;Nandedkar et al., 2004).The surface EMG interference pattern (SIP) signal is made of SMUPS.This strategy reduces the time of examination.With practice, it takes less than 5 minutes to complete the recording including the time for patient preparation, electrode application, etc.This allows one to study multiple muscles, including large muscles like tibialis anterior.Stimuli are used only for the CMAP recording.Thus, it is better tolerated by and less uncomfortable for the patient (de Carvalho et al., 2018).
MUNIX is a three-step process (Fig. 6).First, the CMAP is recorded using standard nerve conduction procedures with one additional manoeuvre.The position of active or E1 electrode is varied to record the CMAP with highest amplitude.The CMAP amplitude is measured from baseline to negative (upward) peak.The negative phase is used to calculate area and power.The area is calculated by summing the absolute voltage of the signal.Power is calculated by adding the squared voltage values.
Next the SIP is recorded when the subject exerts a constant isometric contraction.An epoch of 500 ms duration is used for analysis.Recordings are made at 5 to 8 different force levels ranging from slight to maximal effort.After a short period of rest, the procedure is repeated.For each SIP, its power and area are used to calculate the 'Ideal Case Motor Unit Count (ICMUC)'.
ICMUC ¼ ðCMAP power Â SIP AreaÞ = ðCMAP area Â SIP PowerÞ Finally, a plot of ICMUC versus SIP area is constructed.A power regression defines the relationship between these measurements.
MUNIX is the ICMUC value when SIP area is 20 mV x ms.
The motor unit size index (MUSIX) is calculated as

MUSIX ¼ CMAP amplitude = MUNIX
There are several quality controls applied to avoid spurious measurements.Recordings with SIP Area < CMAP area, or SIP area < 20 mV x ms, or ICMUC > 100 are excluded from the regression.MUNIX analysis is not recommended when the CMAP amplitude is less than 0.5 mV.A low-amplitude CMAP can be due to farfield recordings from other muscles (Nandedkar and Barkhaus, 2007).MUNIX is also underestimated in patients if tremor is present or develops in the tested limb/muscle.The automated version of MUNIX does all the above calculations and provides an on-line update of results as more SIP signals are recorded.
Fig. 6 shows MUNIX analysis from the abductor pollicis brevis muscle (APB) of a normal subject and a patient with ALS.Both have very normal and similar CMAP amplitude.However, the SIP signals are markedly different.In the ALS patient distinct SMUPs with high amplitude are seen at low effort.The ICMUC values are also low.This difference is easily seen at low to moderate effort and when the SIP area is less than 150 mV x ms.Therefore, it is critical to make many SIP recordings with area < 150 mV x ms.This is emphasized in recent guidelines (Nandedkar et al., 2018).
MUNIX has a strong correlation with CMAP amplitude.Therefore, one must maximize the CMAP amplitude by moving the active electrode to different sites.We believe this is necessary for all MUNE methods.Suboptimal CMAP recordings with low amplitude will reduce the number and/or size estimates of the MUs regardless of the method.Strong correlation does not imply that one measurement is unnecessary.The relationship between CMAP amplitude and MUNIX, i.e.MUSIX, is quite different in patients.In Fig. 6, the patient MUSIX is significantly higher than in the normal subject although they have similar CMAP amplitudes.The combination of low MUNIX and high MUSIX indicate MU loss and increased size of surviving MUs due to collateral innervation, such that the amplitude of the CMAP tends to normalize.In this instance, the MUNIX measurements have demonstrated disease activity even though the CMAP appears normal.
MUSIX is useful to assess changes in MU size due to reinnervation.Fig. 7 shows recordings from APB muscle of two patients with similar and reduced CMAP amplitude.In one patient (Fig. 7A), large amplitude SMUPs are easily seen.They are also firing at high rates (>20 Hz) which is the typical pattern seen on needle EMG examination.It is interesting that the same pattern can be also seen on surface EMG.For this reason, we often observe SIP signals at moderate contraction as part of NCS.MUNIX is significantly reduced indicating moderate to severe loss of MUs.Increased MUSIX demonstrates reinnervation.The large SMUPs seen in SIP correspond to the large steps in CMAP scan.This is also seen in patient recordings in Fig. 6.In contrast, the study from the second patient (Fig. 7B) has reduced CMAP and MUNIX, but normal MUSIX.The SIP signals do not show any large amplitude potentials.This indicates failure and/or lack of reinnervation (Chan et al., 2022).The CMAP scan study is consistent with MUNIX and shows a normal sigmoidal pattern with no large steps.
Figs. 6 and 7 show studies of number of MUs using 4 different methods: D50, MFitScan, STEPIX and MUNIX.The numerical values of results from these analyses are different due to the differences in underlying models for analysis.Yet they all show the same pattern, and the individual values are sufficiently reproducible that each method is suitable for clinical trials or for following progress of the disease.

MUNIX in clinical trials for MND/ALS
MUNIX has been applied in the past years in several neuromuscular diseases, especially ALS.It has to be stressed, that the idea of MUNIX was not to use it as a diagnostic biomarker in such diseases, but as a marker of disease progression and motor neuron loss, when individual subjects serve as their own controls.It was hypothesized, that tracking one fundamental pathological aspect of ALS -the loss of lower motor neurons -would be more sensitive compared to other outcome measures in clinical trials to detect potential disease modifying treatment effects.After the first description of the method by Nandedkar and Stålberg 2004, a first small proof-of-concept study was performed in an ALS trial in seven patients.It could be shown, that the change over time was more prominent for the MUNIX sum score of eight muscles compared to the ALSFRS-R or the slow vital capacity (Neuwirth et al., 2010).
After these first promising results, the next step was to evaluate the reliability of this method in multiple centres in healthy volunteers and, in consequence, its feasibility.Interestingly, the reliability was in general good and supported by different groups (Ahn et al., 2010;Delmont et al., 2020;Higashihara et al., 2020;Neuwirth et al., 2011), but was rater-dependent and varied between centres and different working groups, underlining the need for technical standards and the training of observers before proceeding into clinical trials (Neuwirth et al., 2011).The data collected in this study in healthy volunteers were also used to define normal values.However, these values ranged considerably between individuals, like the CMAP amplitude.
Longitudinal observational multi-centre studies followed in ALS patients by several working groups (Fathi et al., 2016;Neuwirth et al., 2015), confirming the higher change rate over time of a set of muscles compared to the gold-standard ALSFRS-R (Fig. 8).In consequence, MUNIX in a set of four muscles would be able to detect potential motor neuron protective effects of disease modifying therapies of 25% in less than half of the time than the ALSFRS-R (12 vs. 26 months, n = 100)).In a further large cohort with 237 ALS patients and 67 controls (healthy subjects and ALS mimics), MUNIX of 3 muscles reflected motor neuron loss and disease accumulation already in early phase 1 and 2 of the disease when applying the D50 model.In summary, the findings showed that MUNIX measurements in the D50 model can serve as a powerful biomarker for individual disease accumulation independently of individual disease aggressiveness.The D50 model assumes a non-linear, sigmoidal approach instead of the linear decline commonly assumed in most studies.The early detection of disease accumulation might be of special interest in phase 1 or 2 clinical trials for a) stratification and b) to detect early effects of potential therapeutics (Ebersbach et al., 2022).
Loss of motor neurons was detectable even before the onset of weakness, functional deterioration and considerable loss of CMAP amplitude due to LMN involvement, and this was confirmed in two independent studies (Escorcio-Bezerra et al., 2018;Neuwirth et al., 2017), suggesting that MUNIX might be able to monitor disease progression in early disease stages.An example is illustrated in Fig. 9.
In consequence, MUNIX has now been applied in clinical ALS trials to investigate treatment effects in addition to the traditional outcome measures.While the results of such studies are still pending, there is at least one report of the use of MUNIX as a primary outcome measure in ALS (Vucic et al., 2023).
Besides its use in ALS, MUNIX was also applied in other motor neuron diseases, such as spinal muscular atrophy.As the method itself is less invasive with the need of relatively few electrical nerve stimulations, it proved to be applicable in children with SMA from the age of 6 years (Mendonca et al., 2021).This offers a further possibility of tracking LMN survival in light of the available disease modifying therapies, nusinersen and risdiplam.
Also, in inflammatory peripheral nerve disorders like chronic inflammatory demyelination polyneuropathy, Anti-MAGneuropathy and MMN, MUNIX was able to objectify treatment effects with IVIG (Philibert et al. 2017, Okhovat et al. 2022), This adds the potential application in other diseases for longitudinal assessments, as MUNIX seems also to reflect nerve dysfunction and not solely lower motor neuron death.This widens the possible application of this method.
However, to harmonize the technical implementation of this method and reduce variability by avoiding typical sources of errors, guidelines have been published on the basis of a large real-world observational trial.In this study, 36 raters from 27 different centres in North America and Europe were trained in a faceto-face practical session and then had to submit test-retest result from four healthy volunteers from their home-labs (Neuwirth et al., 2018).The performance of raters varied considerably between raters and centres.Besides, interindividual differences of raters in general practical experience and routine in electrodiagnostic testing, practical reasons could be identified repeatedly as causes of lower reliability.These were mainly a) insufficient optimization of the CMAP amplitude, and b) recording of only few and/or unstable SEMG signals.These sources of error were highlighted in the guidelines.As a consequence, a practical training and qualification process of raters is recommended, before using the method in clinical trials with patients.However, this aspect is true for every diagnostic method that is applied in studies and clinical practice.

Conclusion
The techniques of estimating the number of MUs have changed significantly in the last two decades.The new methods are using mathematical models to study the CMAP and/or SIP signals to make estimations.Recordings are made under conditions of high signal to noise ratio.Automation has reduced the time of examination, making these methods more feasible in clinical routine.This permits the study of more muscles.Pooling data from multiple muscles reduces the variability and also allows assessments on several body regions.This will allow a ''body-scan", especially important in diseases like ALS with focal onset and spreading of disease.These improvements will make MUNE/MUNIX attractive as biomarkers to study disease progression, even in clinically asymptomatic muscles.The numerical results differ among the methods but demonstrate parallel change on serial studies.Inclusion of MUNE/MUNIX as an outcome measure in clinical trials may reduce the required sample size when only clinical scores are used.
All methods rely on the CMAP recordings for analysis.It is important to ensure that it is recorded with highest care.This includes monitoring all technical factors that affect the CMAP amplitude such as appropriate stimulation, filters, temperature, placement of recording electrodes, etc. Suboptimal amplitude will yield underestimated number of MUs and/or their size and will prevent proper assessment of the disease processes.

Electrical impedance myography
In electrical impedance myography (EIM), a weak, high, multifrequency electrical current is forced through a localized area of muscle and the consequent voltages measured (Rutkove and Sanchez, 2019;Sanchez and Rutkove, 2017).This provides the complex impedance (the resistance and reactance) of the tissue, which is impacted by the compositional and structural aspects of the muscle.EIM can be applied through both surface (sEIM) and needle-based approaches, the former being better established but the latter now being integrated with standard needle electromyography to create a dual-technology, impedance-EMG (iEMG) (Rutkove et al., 2022).Measurements are typically performed with the muscle in a relaxed state, though contraction-induced changes can also be assessed (Sanchez et al., 2016).Fig. 10 provides a basic overview and schematic of both techniques.Fig. 11. A. The stimulus response curve (SR) measures the (baseline-to-peak) responses of graded stimuli which decrease from a level which elicits a maximal response until the response is subliminal.B. The resultant Stimulus-Response curve is typically sigmoidal in nature.① The red circle indicates the fraction of the maximal response that is used to threshold track all subsequent protocols.C. Strength-duration properties (SD) are measured by tracking the target response using stimuli of different widths (usually 1, 0.8, 0.6, 0.4, 0.2 ms).D. When the charge (stimulus current Â width) is plotted vs stimulus width, the relationship in healthy axons is typically linear.② The intercept on the x-axis corresponds to (the negative of) the Strength-Duration Time Constant.E. Threshold electrotonus (TE) measures the excitability before, during and after long-lasting (100 ms) subthreshold currents.Depolarizing TE (red) uses a subthreshold conditioning stimulus which is + 40% of the test (unconditioned) stimulus current amplitude.Hyperpolarizing TE (green) uses a À40% conditioning stimulus current.F. When the conditioned test stimuli are plotted as threshold reductions, they resemble electrotonic responses of membrane potential to stimulus currents from conventional physiology recordings.③ Accommodation to depolarization is seen as a 'sag' following the peak threshold reduction.G.The Current-Threshold relationship (IV) is another way of measuring the rectifying properties of the axonal membrane.It tests the excitability 200ms after the onset of conditioning stimuli which are graded from À100% to + 50% of the unconditioned test stimulus.H.The resultant plot of current vs threshold reduction resembles the Current-Voltage (IV) relationship from current clamp recordings.④ Resting IV slope is the slope of the current-threshold calculated between À10% and +10% polarizations and is indicative of conductances active at resting membrane potential.I.The Recovery Cycle (RC) is recorded by testing the excitability at conditioning-test intervals following a supramaximal stimulus which are roughly logarithmically spaced from 2 to 200 ms.J.The threshold change demonstrates the typical pattern of excitability following activation: refractoriness followed by superexcitability and then late subexcitability.⑤ Superexcitability is the peak reduction in excitability following an action potential and is due to the charging and then discharging of the internodal axolemma following an action potential.
C. Shin-Yi Lin, J. Howells, S. Rutkove et al. Clinical Neurophysiology 162 (2024) 91-120 sEIM has been studied in the assessment of MND predominantly as a biomarker of disease progression or remission as it can provide a sensitive and simple measure of overall muscle state.The major alterations that are observed in MND is a reduction in the single frequency (e.g., 50 or 100 kHz) reactance, representing a reduction in myofibre membrane content within the muscle and an increase in the single frequency resistance, due to compositional changes as well as simple global muscle atrophy.These two primary impedance measures are combined in the phase angle (or simply phase), via the equation Phase = arctan (reactance/resistance); given that resistance increases in this disease and reactance decreases, the phase angle shows the best sensitivity to progression.For most MND applications, the sEIM 50 kHz phase has appeared to be the single most sensitive measure to disease progression.Approaches for capturing the entire frequency spectrum in single parameters have also been used (Rutkove et al., 2010), but more recently machine learning approaches can also be applied that capture multiple components of the frequency spectrum and all three impedance parameters at once (Kapur et al., 2018;Pandeya et al., 2022).
sEIM is advantageous in that it provides rapid quantification of individual muscles or muscle groups and can thus track localized disease progression.Several studies, including one single-centre (Rutkove et al., 2012) and two multicentre studies (Rutkove et al., 2010;Shefner et al., 2018), confirmed the potential for EIM for use in MND.By tracking muscles that are most rapidly declining in a patient with MND, EIM can be used to sensitively assess disease progression.Another potential application of EIM in MND would be for assessing drug effects in an individual patient, allowing for ''precision medicine" to be applied-namely assessing whether a specific medication is slowing progression in that patient since it can provide quantitative data from that body region or individual muscle that is progressing most rapidly.Another important aspect of sEIM for tracking MND progression is its simplicity of use.Even caregivers can be readily trained to obtain high quality data allowing for at-home assessments, with increased frequency of measurement, thus helping to average out measurement noise, and further increase the sensitivity to disease progression and potentially drug effect (Rutkove et al., 2020).sEIM in animals has been shown to closely correlate to CMAP and motor unit number estimates (Li et al., 2013) and also to CMAP in humans (Zong et al., 2018).
Whereas sEIM can be performed using commercial bioimpedance devices (those generally used for whole body bioimpedance analysis) with adhesive electrodes applied over specific muscles or muscle groups of interest, commercial systems for measurement are also being developed and, at the time of this writing, in the process of undergoing regulatory approval.
Needle EIM technologies have also been employed mainly in animals, with either four separate needle electrodes (Semple et al., 2020) or two concentric EMG electrodes used in combination (Cardoner et al., 2021).The major advantage of needle-based methods is that the impact of intervening tissues (subcutaneous fat and skin) is removed and good electrical contact with the tissue is relatively assured, thus providing a more true muscle measurement.More recently, four impedance electrodes have been added to a modified concentric needle, for a total of six electrodes (Rutkove et al., 2022).The value of this combination technology is to obtain both EMG and EIM data simultaneously, providing a yet still richer data set that assesses both the active (EMG) and passive (EIM) properties of muscle at once.While still in early development, this technology, could have value in achieving earlier diagnosis of MND and also in tracking progression more sensitively.Early work in animal models supports this potential application (Pandeya et al., 2023).
There are several challenges to impedance technologies that are important to highlight.First, all electrical impedance methods are highly sensitive to the contact between the electrodes and the surface being measured.Poor contact for any reason will greatly distort the data being obtained.In contrast, however, 50 or 60 Hz line interference, which is such a common source of noise in standard neurophysiological measurements, is rarely observed in EIM measurements.Second, the method is highly sensitive to the position of the electrodes and their orientation to the underlying muscle.Slight variations in positioning can alter the data significantly.However, if one adopts a standard set of procedures in data collection, it is relatively easy to overcome this issue.A third limitation is that peripheral oedema will impact measurements, especially for surface studies, although whole body hydration status has minimal effect on the data (Jia et al., 2014).The final challenge, which is perhaps the most significant, is the neuromuscular and neurophysiological communities having little knowledge or comfort with the concepts of electrical impedance compared to standard neurophysiology.Perhaps, over time, as the technology matures and becomes more widely accepted and employed, this final hurdle to broader acceptance will be overcome.

Introduction
Nerve excitability techniques have been used to examine many diseases of the LMN including MND.Although nerve excitability techniques technically study axonal excitability at the site of study, the axon itself is an extension of the motor neuron, and therefore changes in axonal excitability likely reflect changes in the motor neuron itself.Indeed, the axons and their excitability are an essential part of Sherrington's final common pathway.Furthermore, the distal terminals of an axon are likely to be where many fasciculations originate, particularly in healthy subjects and benign fasciculation syndrome.Changes in axonal excitability in MND reveal insight into pathophysiology and have been used as a biomarker of progression and therapeutic intervention (see later).
Nerve excitability techniques provide further insights into the health of an axon than conventional NCS.They are particularly sensitive to changes in resting membrane potential, ion channel dysfunction and the structural integrity of the axon and myelin sheath.Excitability techniques are easy to perform for the clinical neurophysiologist, although they still require specialized hardware and software.For background on the technique and its application see Bostock et al. (1998); Kiernan et al. (2020); Kiernan et al. (2000); Krishnan et al. (2009).The technique is summarized in Fig. 11.

Amyotrophic lateral sclerosis
There are no axonal excitability studies which have focused on primary lateral sclerosis even though the technique is likely to give insight into changes of the lower motor neuron downstream of the central insult.Most nerve excitability studies in MND have focused on the most common form, sporadic ALS, which will be described in the following sections.

First excitability observations
The first clinical study using modern nerve excitability techniques used the threshold electrotonus protocol in 11 subjects with ALS (Bostock et al., 1995).The recordings demonstrated a reduction in accommodation to depolarization, suggestive of reduced K + conductances.The authors simulated these changes in vitro with slow and fast K + channel blockers; and in silico by reducing slow and fast K + channel conductances in a model of a human motor axon.The authors concluded that ''nerves in ALS patients suffer a progressive imbalance between sodium and potassium conductances".Furthermore, they hypothesized that ''drugs which reduce steady state sodium currents or activate potassium channels should act to normalize threshold electrotonus in ALS, and if they do so they may also reduce the incidence of fasciculations and cell death".So, it must be particularly satisfying for these authors to see the use of nerve excitability techniques (including threshold electrotonus) to monitor the effect of the slow K + activator, ezogabine/retigabine (Wainger et al., 2021).
The strength-duration time constant (SDTC) of cutaneous afferents is longer than that of motor axons in healthy axons, and healthy sensory axons are more susceptible to ectopic activity than motor axons.Because SDTC can reflect changes in persistent Na + currents, Mogyoros et al. (1998b) sought to gain insight into the role of persistent Na + currents in ALS.They found an increase in SDTC in 23 ALS patients compared to healthy controls (but only when expressed as a percentage of SDTC in sensory axons), and no evidence for a change in resting membrane potential.They concluded that ALS axons have a greater persistent Na + conductance than normal controls, a change that could contribute to fasciculations.Kanai et al. (2003) used the then recently developed TROND protocol (Kiernan et al., 2000) to measure excitability properties in Machado-Joseph disease (MJD) and 22 patients with ALS.In MJD, they also found a significant prolongation in SDTC, and this was correlated with cramp severity, which responded to treatment with mexiletine, as do cramps in ALS (Oskarsson et al., 2018).In ALS, there was a reduction in accommodation during depolarizing threshold electrotonus, along with increased superexcitability but only a trend toward a longer SDTC.These findings suggest increased persistent Na + current in MJD, and lesser K + currents in ALS, the latter consistent with (Bostock et al., 1995).

Fasciculations and persistent Na + currents
Because fasciculations may have a distal origin, Nakata et al. (2006) compared the excitability of motor axons in 22 ALS patients at both the wrist and the motor point.They found a greater reduction in accommodation of ALS axons at the motor point than at the wrist, and concluded potassium channel function is impaired to a greater extent at the distal motor terminals.
Tamura et al. ( 2006) used the technique of latent addition, which is considered a more accurate measure of persistent Na + currents (Bostock and Rothwell, 1997), in neuropathy and 36 ALS patients.They found significantly greater changes in latent addition in neuropathy than normal controls, but the changes did not reach significance in ALS.

Disease stage and prognosis
Kanai et al. (2006) correlated excitability properties at different disease stages (according to the size of the CMAP) in 58 ALS patients.They found that SDTC was only longer in patients early in the disease (CMAP > 5 mV), and abnormal TEd and supernormality was greatest in established disease (CMAP 1-5 mV).With the use of mathematical modelling, they interpreted the evolving pattern of dysfunction as initially due to increased persistent Na + conductances, followed by impairment in K + conductances.
In agreement with these findings, in a later study of 112 ALS patients, Kanai et al. (2012) found strong correlations with shorter survival, for SDTC in early disease (CMAP > 5 mV) and superexcitability in established disease (CMAP < 5 mV).

Heterogeneity of motor units
In healthy subjects, the excitability properties of axons recruited at different target levels change in a graded manner from low to high electrical thresholds (Shibuta et al., 2010;Shimatani et al., 2023;Trevillion et al., 2010).However, this relationship appears less reliable in ALS, with increased variability of excitability of axons recruited at different target levels (10% to 60% of maximal CMAP; Shibuta et al., 2013).There is clear evidence of greater variability of threshold in ALS motor axons compared to normal controls in many studies, including (see Table 2, Figs. 2 and 3 in Garcia et al., 2023;Fig. 1B & 2 in Kanai et al., 2006; Table 1 in Nakata et al., 2006;Table 1 in Tamura et al., 2006;Figs 2 & 4 in Vucic and Kiernan, 2006;Fig. 1 in Vucic and Kiernan, 2010).

Evidence for a common mechanism?
Howells et al. ( 2018) hypothesized that the functional demands of reinnervation could impact on the axonal excitability of single motor units.When they compared the axonal excitability of single motor units from moderately and severely affected muscle of ALS patients to those from normal controls, they found a graded change in abnormal excitability which was correlated with measures of fine motor function such as the number of motor units, fine motor subscore of the ALS functional rating scale [ALSFRS], and Medical Research Council score [MRC].Modelling suggested that the most plausible explanation was a non-selective reduction of all ion channel conductances and a loosening of the attachment of the myelin sheath at the paranode.The authors suggested that this could be because the sick motor neuron was unable to maintain protein homeostasis in the distal axon membrane, resulting in reduced expression of all ion channel currents and a loosening of the paranodal seal (due to reduced supply of anchoring proteins).

Biomarkers of axonal dysfunction
There has been limited use of axonal excitability in ALS for monitoring progression of LMN dysfunction and response to treatment, perhaps because of the complexity of the disease, and that the neurodegeneration does not lend itself to the use of a single parameter.The SDTC appears to be a strong prognostic indicator -but only early in the disease.Reduced accommodation to longlasting depolarization and increased superexcitability are consistent findings in nearly all ALS studies to date.Development of biomarkers in a heterogeneous condition such as ALS will require larger datasets, the inclusion of nerves other than the median nerve and better markers of disease stage.Lugg et al. (2022) performed a systematic review and metaanalysis to identify potential biomarkers in ALS.They found the following seven candidate biomarkers from excitability studies (ranked in order of descending overall effect size): depolarizing threshold electrotonus at 90-100 ms (TEd90-100 ms), SDTC, superexcitability, TEd40-60 ms, resting IV slope (slope of the current-threshold relationship at rest), 50% depolarizing IV (threshold reduction for 50% polarization in current-threshold relationship) and subexcitability.However, when they examined the sensitivity of candidate biomarkers from studies performed in early disease, they found the following four parameters helpful (again in descending order): TEd10-20 ms, TEd90-100 ms, superexcitability and SDTC.
In the largest study of ALS axonal excitability to date, Garcia et al. (2023) combined four excitability parameters (based on the conclusions for early disease by Lugg et al., 2022) and the CMAP scan to build an 'electrophysiological risk score', which they used to assess diagnostic accuracy in 153 ALS and primary muscular atrophy patients.This risk score had an area under the receiver operating curve of 0.85, and was able to correctly identify 24 out of 33 ALS patients who could not be diagnosed clinically on the day of examination.Interestingly, this study did not find a signifi-cant difference in SDTC, but rather a 'nonlinear' measure which reflects the distance from the mean of healthy controls.

Sensory excitability
Although ALS is generally considered a purely motor disorder, there are numerous reports of subtle sensory abnormalities.Given the sensitivity of axonal excitability techniques to changes in ion channel function and membrane potential, Matamala et al. (2018) examined the excitability of cutaneous afferents at the wrist in ALS subjects using an extended excitability protocol.They found no evidence of a change in excitability of sensory axons in subjects with significant motor neuron degeneration (in agreement with Burke et al., 1997;Mogyoros et al., 1998a).They concluded that if there were any changes in sensory axon excitability, they would have to be proximal to the site of testing (the wrist).In contrast, many of the ALS mimics do have significant differences in sensory excitability.

Therapeutic agents
On the basis that SDTC is a marker of persistent Na + channels, several clinical studies have examined the effect of drugs which block Na + channels.Preclinical studies have demonstrated that riluzole blocks persistent Na + channels in murine models.However, two axonal excitability studies in ALS patients have found no evidence for a change in axonal strength-duration properties (observed at the wrist) either acutely or after 7 weeks of a therapeutic dose (Kovalchuk et al., 2018;Vucic et al., 2013).Similarly, therapeutic doses of the Na + channel blockers, flecainide and mexiletine have had no significant effect on axonal excitability properties (Park et al., 2015;Shibuya et al., 2015;Weiss et al., 2021).
However, the slow potassium channel activator, ezogabine/retigabine does demonstrate a dose-dependent reduction in accommodation to depolarization and superexcitability.Interestingly, retigabine also reduces SDTC both acutely (6 hours after 300 mg) and after 8 weeks of 900 mg/day (Kovalchuk et al., 2018;Wainger et al., 2021) presumably due to its effects on membrane potential.

Less common forms of ALS
Vucic and Kiernan (2007a) applied cortical and peripheral excitability techniques to patients with flail arm syndrome.They found evidence for increased SDTC, reduced accommodation to depolarization, but no difference in recovery cycle parameters.In combination with findings of reduced short interval intracortical inhibition they concluded that the flail arm syndrome is an unusual variant of ALS.
The SOD1 and c9orf72 mutations of familial ALS (FALS) also demonstrate a pattern of abnormal axonal excitability (Geevasinga et al., 2015;Vucic and Kiernan, 2010).As is the case in the sporadic form of ALS, SDTC is significantly increased in ALS patients with mutations in SOD1 and c9orf72 when compared to healthy controls.
However, carriers of these two familial forms of ALS did not have increased superexcitability (regardless of whether they were symptomatic or not), nor was there any evidence of reduced accommodation to depolarization in carriers of the SOD1 mutation.

ALS mimics
Many of the conditions which are considered mimics of ALS have also been examined using axonal excitability techniques.All of the ALS mimics studied to date have a pattern of excitability that differs in some way from that seen in sporadic ALS.
Below is a summary of the key differences in abnormal excitability (relative to healthy controls) between mimics and sporadic ALS.
Axonal excitability studies of ALS mimics have provided useful insights into the pathophysiology, and a technique for monitoring disease progression and response to treatment in patients that are encountered in specialist ALS clinics.

Challenges
Performing nerve excitability studies in ALS subjects is rarely challenging.The technique is well tolerated and copes well with occasional fasciculations, and indeed reinnervation often makes studying single units easier than in healthy controls.The challenge in ALS nerve excitability studies, and in particular the multiple excitability measures TROND protocol, is nearly entirely to do with interpretation.The risk of misinterpretation (or even confirmation bias) is high when individual excitability parameters are used to interpret the underlying biophysical mechanisms.The strength of the TROND protocol is that it efficiently probes multiple underlying mechanisms which have overlapping effects on the overall pattern of excitability.For example, we know that most axonal excitability parameters are sensitive to membrane potential, but membrane potential itself is dependent on the properties of many voltagegated ion channels and pump currents.

Site selection
For simplicity many studies test the excitability of median nerve axons at the wrist, measuring the response from abductor pollicis brevis, which of course is often involved early in the ALS split hand.This works well for conditions which are systemic in nature, or the pathology is generalized within the tested motor unit pool.These underlying assumptions need careful consideration in an evolving heterogenous disease like ALS.

Strength-duration time constant (SDTC)
Much emphasis has been placed on the significance of prolonged SDTC in ALS, even though it doesn't reach statistical significance in many studies.A larger SDTC is often presented as evidence for increased persistent Na + currents.However, SDTC is also dependent on other membrane properties, such as axonal depolarization and, in the single motor unit study of Howells et al. (Howells et al., 2018), Na + currents were reduced, along with all other conductances.Interestingly, the only significant druginduced changes in SDTC in ALS has come from clinical studies of K + channel activators and not drugs that block Na + channels (Kovalchuk et al., 2018;Wainger et al., 2021).

Heterogeneity
Although nerve excitability studies typically track compound responses, they essentially measure the excitability of the few axons recruited near the target response.In healthy axons, there is good evidence that these motor axons are representative of the parent motor pool, with graded changes according to recruitment order.However, the notion of a typical motor unit in ALS is problematic (as it is in some MUNE methods), with marked variability within an individual motor pool.Future studies may consider profiling a whole motor pool to get a better understanding of neurodegeneration.

Disease staging
It is obvious that there is an evolution in the excitability of motor axons, and to further understand changes in excitability there will need to be more sensitive measures of disease stage.Measures of fine motor skills (MUNE, ALSFRS-R fine motor subscore, MRC) in the tested muscle are likely to yield better correlations than the more general measures ALSFRS-R, CMAP or symptom months.

Conclusion
Nerve excitability studies are in vivo, non-invasive, easy to perform, and provide a wealth of information about the underlying pathophysiology.The same technique is applicable to preclinical studies and there is a relatively simple mathematical model which is well suited to interpretation.Furthermore, the same technique that is used for probing pathophysiological mechanisms can also be used as a biomarker for drug development and clinical trials.Nerve excitability studies are well suited to teasing out the complexity of lower motor neuron degeneration in ALS.

Introduction
Neuromuscular ultrasound is a non-invasive imaging technique that uses a high-frequency transducer to visualize the peripheral nerves and muscles.Nerve ultrasound can detect changes in nerve morphology, such as swelling and hypoechoic areas, which are indicative of nerve damage.Muscle ultrasound can detect changes in muscle morphology, such as atrophy and fasciculations.It may also detect fibrillation, though the sensitivity is low ( van Alfen et al., 2011).In recent years, there has been increasing interest in the potential for ultrasound not only as a diagnostic biomarker in ALS but also as a tool to detect disease progression (Cartwright et al., 2011;Nodera et al., 2014;Schreiber et al., 2015).

The role of nerve ultrasound in ALS
Nerve ultrasound can measure the size of peripheral nerves, either circumferential which is referred to as the cross-sectional area (CSA), or its longitudinal structure.In the evaluation of patients with ALS, nerve size including its proximal and distal ratio, has been helpful in differentiating between ALS subtypes as well as between diseases that mimic ALS.For instance, in patients with ALS, there is typically a reduction in peripheral nerve CSA, reflecting the underlying nerve atrophy due to axonal loss (Cartwright et al., 2011;Nodera et al., 2014;Schreiber et al., 2015).

Nerve CSA as a lower motor neuron biomarker
In a study by Schreiber et al., the authors investigated the role of distal ulnar and median nerve CSA as a marker of LMN involvement (Schreiber et al., 2015).They detected significant reductions only in ulnar CSA at the wrist and forearm levels in the bulbar and limb onset ALS.
Similar findings of nerve atrophy were also demonstrated by Nodera et al. in the cervical nerve roots of patients with ALS (Nodera et al., 2014) (Fig. 12) The authors detected a reduction in nerve root diameter when compared to healthy controls.The degree of reduction tended to be greater than in the peripheral nerve CSA of patients with ALS suggesting that ultrasound changes in the cervical nerve roots might be more sensitive at detecting axonal loss compared to peripheral nerves.However, there was no correlation between nerve root size and the disease severity/duration limiting its clinical utility (Mori et al., 2016;Nodera et al., 2014).
In addition to the use of nerve CSA or diameter in the diagnosis of ALS, ratios of distal and proximal peripheral nerve CSA are also sensitive biomarkers.In the study by Noto et al., the authors demonstrated that reductions in proximal median nerve CSAs were greater than the distal portions resulting in an increase in the distal-proximal CSA ratio in ALS (Noto et al., 2018a).These findings support the proposed pathophysiology that motor axons predominate in the proximal vs. distal segments of the median nerve (Watchmaker et al., 1991), leading to the characteristic pattern of non-uniform atrophy that is seen in ALS.The median nerve CSA wrist-upper arm ratio was increased across all ALS subtypes.This suggests that peripheral nerve CSA distal-proximal ratios could potentially be a sensitive diagnostic biomarker, regardless of the site of disease onset in ALS.
To study the relationship between nerve size and nerve growth factors, Schreiber et al. analysed the CSF and serum level of progranulin (PGRN), a multifunctional growth factor found in various human tissues (Schreiber et al., 2018).In response to central and peripheral nerve injury, including ALS, there is typically an upregulation of microglial PGRN to maintain neuronal structure and function.The authors demonstrated an inverse relationship between median and ulnar nerve CSAs and CSF levels of PGRN.They speculated that in ALS, the increased PRGN expression is an effort to compensate for the progressive axonal damage.

Differentiating between ALS subtypes and mimic disorders
Within the ALS disease spectrum, there are subtypes, including PMA, that have different disease trajectories as compared to classical ALS (Shahrizaila et al., 2016).Determining the disease subtype helps to better prognosticate.
In the study by Schreiber et al., (2015) the authors demonstrated that distal ulnar CSA was significantly reduced in upper motor neuron-dominant ALS but not PLS which has similar clinical characteristics.These findings suggest that distal ulnar nerve CSA has the additional benefit of differentiating between certain ALS subtypes.
There are also several conditions that might also mimic ALS such as MMN.In MMN, nerve enlargement rather than reduction is frequently detected, and ultrasound is a useful technique to distinguish MMN from ALS (Grimm et al., 2015a;Jongbloed et al., 2016;Loewenbruck et al., 2016;Noto et al., 2018a).Although there is a correlation between CSA measured by MRI and nerve ultrasound, nerve ultrasound was found to be more sensitive than MRI in distinguishing patients with MMN from those with ALS (Jongbloed et al., 2016).
In trying to differentiate between its mimics, nerve ultrasound protocol that is tailored to a patient's clinical deficits is superior in terms of diagnostic accuracy and efficiency when compared to a rigid protocol (Loewenbruck et al., 2021).

Monitoring disease progression
Studies investigating the role of nerve ultrasound in monitoring disease progression have been less promising.In one study, the authors detected a significant decline in ulnar nerve CSA at a monthly rate of À0.04 mm 2 at the forearm and À0.05 mm 2 (Schreiber et al., 2016).However, the authors concluded that the changes were too subtle to be of true clinical value, especially in treatment trials which would require approximately 332 patients to detect a 50% treatment effect on ulnar nerve CSA.
In another study, the authors constructed a predictive model investigating the probability of fast disease progression in ALS which included proximal to distal nerve CSA ratio (Toh et al., 2021).The authors found the proximal (at mid-arm) to distal (at forearm) median nerve CSA ratio < 1.22 was one of the strong predictors of fast progressors in ALS.

The role of muscle ultrasound in ALS
Over the years, several criteria have been developed to facilitate the diagnosis of ALS (Brooks, 1994;de Carvalho et al., 2008;Shefner et al., 2020).The disease hallmark of ALS remains identifying progressive dysfunction of the lower and upper motor neurons, either by clinical evaluation or through EMG.Muscle ultrasound has emerged as a tool that can detect changes in muscle architecture including dynamic changes such as fasciculations (Arts et al.,Fig. 13.Tibialis anterior muscle of a patient with lower limb onset ALS at two time-points from disease-onset: (A) 10 months: there are patchy streaks of higher echo intensity typical of neurogenic pattern of muscle changes on ultrasound.(B) At 20 months, these features are more apparent with further increase in echointensity.Clinically, there was associated reduction in muscle strength.
C. Shin-Yi Lin, J. Howells, S. Rutkove et al. Clinical Neurophysiology 162 (2024) 91-120 2012; Cartwright et al., 2011;Lee et al., 2010;Misawa et al., 2011).The proposed ''Gold Coast" criteria considered the use of muscle ultrasound in providing supportive evidence for detection of fasciculation, further adding validity to its role in ALS (Shefner et al., 2020).Several parameters, including muscle thickness, echo intensity and fasciculations can be evaluated through muscle ultrasound evaluation of bulbar, respiratory and limb muscles in ALS.

Fasciculations
To date, the most robust use and evidence for use of muscle ultrasound in ALS has been in detecting fasciculations (Grimm et al., 2015b;Johansson et al., 2017;Juan et al., 2020;Liu et al., 2021;Misawa et al., 2011;Tsuji et al., 2017).Fasciculations are random, spontaneous twitching of a group of muscle fibres that belong to a single motor unit (Dengler et al., 2020).In contrast to needle EMG, ultrasound can evaluate a larger area of muscle as well as detect fasciculations in deeper muscle sections, which are typically not clinically apparent (Supplementary video 1).There is also excellent inter-rater agreement and one study reported fasciculations could be distinguished from artifacts with a sensitivity of 90.9% and specificity of 98.5% (Kramer et al., 2014).Although there are several ultrasound protocols for detecting fasciculations, the optimal scan time has been reported as 30 s in most muscles.However, a period of 60 s was required to confirm fasciculations in the presence of two or more fasciculations (Noto et al., 2017a).
Several studies have demonstrated that muscle ultrasound is more sensitive in detecting fasciculations when compared to clinical and EMG evaluation (Grimm et al., 2015b;Johansson et al., 2017;Misawa et al., 2011).When used in combination with EMG, the diagnostic sensitivity was substantially improved (Bokuda et al., 2020;Grimm et al., 2015b;Misawa et al., 2011).Muscles for which fasciculations were more frequently detected through ultrasound compared to EMG were the tongue, biceps brachii, and tibialis anterior muscles (Misawa et al., 2011).In contrast, the paraspinal muscles had lower rates of detection.The probability of detecting fasciculations in more than one muscle was also higher in ALS when compared to non-ALS patients (Fukushima et al., 2022;Tsuji et al., 2017).There are also studies utilising ultrasound scoring systems to improve the diagnostic sensitivity and specificity in ALS through the detection of fasciculations.Tsuji et al. developed a fasciculation ultrasound score incorporating nine muscles from the different anatomical regions (Tsuji et al., 2017).A score of > 2 had a sensitivity of 92% and specificity of 100% in diagnosing ALS.
In determining ALS patients with fast progression, a predictive model utilising biceps brachii fasciculation count in its construct had good sensitivity (81%) and specificity (91%), indicating that fasciculation frequency is associated with disease progression (Toh et al., 2021).In the study by Noto et al., fasciculation intensity was also associated with disease progression and separately to markers of cortical dysfunction, specifically the advent of cortical hyperexcitability (Noto et al., 2018b).There are also differences in the distribution and firing frequency of fasciculations between patients with ALS and the mimic disorders and MMN (Tsugawa et al., 2018;Tsuji et al., 2020).In MMN, fasciculations were not detected in the tongue or truncal muscles (Tsuji et al., 2020).There was also no asymmetry in the rate of detection between the limb muscles in MMN.In contrast, patients with ALS had a significantly higher fasciculation detection rate at the site of disease onset (Tsugawa et al., 2018).
There have been other efforts to automate the detection of fasciculations through computational analysis (Harding et al., 2016).In a study that compared automated vs operator-identified twitches, the authors found that the computational technique could detect muscle twitches with a high degree of accuracy (0.8 3 < accuracy < 0.96) whereas the manual operator-identified techniques had higher inter-operator agreements when there was a higher number of twitches.

Muscle echo intensity and muscle thickness
In ALS, one of the key clinical characteristics is muscle atrophy due to progressive motor axonal degeneration.On muscle ultrasound, these features can be detected by changes in the muscle echo intensity and reduced muscle thickness, even in the early stages of the disease (Arts et al., 2008).Muscle echo intensity depicts the brightness of the ultrasound image which is low in healthy muscles.In ALS, muscle denervation results in fatty infiltration which causes muscles to appear brighter, i.e. higher echo intensity, and greater homogeneity over time (Pillen et al., 2009a) (Fig. 13).To objectively quantify muscle echo intensity, greyscale analysis has been found to be more reliable and more sensitive compared with visual evaluation of the images (Arts et al., 2008;Pillen et al., 2009b).One standard technique involves exporting image to software programmes, such as ImageJ, where a region of interest can be selected for greyscale analysis (Hobson-Webb and Simmons, 2019).
In one of the initial ultrasound studies investigating the structural muscle changes in patients with ALS, the authors found a significant increase in echo intensity, which was more marked than the decrease in muscle thickness (Arts et al., 2008).The same investigators later performed a longitudinal 6-month follow-up study, evaluating muscle thickness, echo intensity and fasciculations, together with strength and ALSFRS-R scores (Arts et al., 2011).They found that muscle ultrasound abnormalities could be detected even in muscles that had preserved strength.However, longitudinal changes of echo intensity and muscle thickness have large variations with little correlation with functional or strength scores.This limits their value as a prognostic disease biomarker.
The utility of muscle ultrasound in evaluating dissociated small hand muscle atrophy, or the split hand index (SHI) in ALS has also been investigated.Seok and colleagues studied the SHI using muscle echo intensity and compared these to CMAPs (Seok et al., 2018).They demonstrated the SHI echo intensity to be significantly higher in patients with ALS, particularly in upper limb-onset ALS, and the measure also had significantly better diagnostic accuracy than the CMAP-derived SHI.In a study by Rajabkhah et al., the association between muscle ultrasound (echo intensity and thickness), MUNIX and clinical parameters were investigated (Rajabkhah et al., 2020).They found that echo intensity was highly associated with clinical scales and MUNIX in patients with ALS, further confirming its utility as a diagnostic biomarker.
There have also been studies to evaluate automated techniques in quantifying muscle echo intensity.In one study, muscle echo intensities of the biceps brachii, rectus femoris and tibialis anterior muscles were evaluated using greyscale analysis and the 16 automatic thresholding methods of the ImageJ program (Spiliopoulos et al., 2022).Both the mean greyscale values and mean hyperechoic fractions of 8 out of the 16 automatic thresholding methods were significantly different between patients and controls in all three muscles.Four thresholding methods (Default, Li, Moments, Otsu) showed a significant correlation between hyperechoic fractions and muscle strength, and diagnostic accuracy that was comparable or superior to greyscale values.The Otsu method also had the advantage of detecting ultrasound changes in muscles of normal strength in ALS patients.These findings were promising indicators that automated echogenicity analysis could potentially be implemented in clinical studies.
There have also been studies investigating muscle echo variance rather than echo intensity in patients with ALS.In a study by Martinez-Paya, the authors evaluated the characteristics of echo variation comparing these with other muscle ultrasonography parameters (Martinez-Paya et al., 2017a).They found that although there was increased echo intensity, decreased thickness and decreased echo variation in the muscles of patients with ALS, echo variation strongly correlated with muscle strength and the ALSFRS-R scores.The authors also compared the diagnostic accuracy of utilising differences in grey-level co-occurrence matrix (GLCM) parameters, with standard muscle ultrasound parameters (echo intensity, echo variation, and muscle thickness) (Martinez-Paya et al., 2017b).They found that the GLCM parameters showed reduced granularity in the muscles of ALS patients compared to controls.A combination of GLCM, echo variation and muscle thickness provided the most promising measure to discriminate ALS.To determine if these parameters could also represent disease progression in ALS, the authors followed changes over a 5-month period (Martinez-Paya et al., 2018).There was loss of muscle thickness and increase in both echo intensity and echo variance, although these changes were small.One significant limitation in evaluating muscle echointensity is difficulties in standardising ultrasound images between machines.The way each system is set-up to depict images rely on proprietary software of the ultrasound manufacturers.As a result, comparison between centres using different ultrasound systems is challenging (Wijntjes and van Alfen, 2021).

Respiratory muscle evaluation
The electrophysiological evaluation of respiratory muscles such as phrenic nerve studies and EMG of diaphragm are technically challenging.Dynamic muscle ultrasound offers a more objective measure of evaluating respiratory muscles, specifically the diaphragm (Supplementary video 2).
Hiwatani and colleagues evaluated ultrasound measurements of diaphragm thicknesses during respiration, comparing them to conventional measurements of respiratory function in patients with ALS (Hiwatani et al., 2013).The maximal diaphragm thickness during the maximal inspiratory effort (DTmax) and minimal diaphragm thickness at the end expiratory position (DTmin) were measured.The thickening ratio (TR), defined as the ratio of DTmin to DTmax, was also calculated.DTmax, DTmin and the TR were all significantly decreased in ALS patients correlating with % vital capacity and inversely correlating with pCO 2 .There was also high inter-observer reliability, suggesting the technique could be reliably used in assessing respiratory function in patients with ALS.There have since been multiple studies confirming similar findings with diaphragmatic ultrasound in ALS (Aliberti et al., 2013;Fantini et al., 2016;Pinto et al., 2016).Correlations have also been established between diaphragmatic thickness, vital capacities and CMAPs following phrenic nerve stimulation.However, the value of diaphragmatic ultrasound in evaluating disease progression is less clear.One study found longitudinal changes in diaphragmatic thickness over a 4-month period failed to reach statistical significance (Pinto et al., 2017).Instead, phrenic nerve stimulation demonstrated significant decline of diaphragmatic CMAP amplitude, together with ALSFRS-R, maximum expiratory and sniff nasal inspiratory pressures.

Evaluation of bulbar muscles
Video fluoroscopy and tongue needle EMG are helpful in evaluating the extent of bulbar involvement in ALS but can be very uncomfortable and require patient cooperation.Muscle ultrasound detecting fasciculations in the tongue muscle has been demonstrated to be helpful in differentiating patients with ALS from mimic disorders (Grimm et al., 2015a;Misawa et al., 2011).In other studies, tongue thickness has been shown to correlate with function (Nakamori et al., 2016;Tamburrini et al., 2010) (Supplementary video 3).In a study that investigated nine patients who underwent simultaneous static and dynamic video ultrasound and video fluoroscopy, the authors found that video ultrasound had a higher sensitivity in evaluating the dynamic factors that represent the early signs of dysphagia (Tamburrini et al., 2010).Ultrasound was able to demonstrate lingual atrophy in the static phase, and reduced lingual movement in the dynamic phase was observed in 5/9 patients through ultrasound compared to fluoroscopy.Nakamori and colleagues investigated tongue thickness in 18 patients with ALS by measuring the vertical distance from the surface of the mylohyoid muscle to the tongue dorsum using ultrasonography (Nakamori et al., 2016).They found that tongue thickness was associated with body mass index and type of disease onset.On repeat studies, patients with ALS had reduced tongue thickness associated with disease progression and tongue dysfunction.
In a study to determine the pattern of change in bulbar muscles of patients with ALS, Noto et al. recorded the mylohyoid and geniohyoid muscle complex (mylohyoid-geniohyoid-muscle-complex) thickness in 30 ALS patients using MÀmode (Noto et al., 2017b).The ratio of mylohyoid-geniohyoid-muscle-complex thickness was determined by the maximum thickness during swallowing divided by thickness at rest.The authors found reduced thickness ratio in patients compared to controls and this was more pronounced in patients with bulbar-onset compared to limb-onset ALS.The thickness ratio was also negatively correlated with the severity of UMN involvement in the bulbar region, suggesting its potential value as an objective marker of UMN involvement in the bulbar region of patients with ALS.
Measures of echo intensity of the tongue muscle have also been investigated through quantitative ultrasound to determine its use in distinguishing patients with ALS from controls and to correlate with function (McIlduff et al., 2020).The authors found that echo intensity could differentiate healthy from diseased tongue muscle and demonstrated a negative correlation with the bulbar ALSFRS-R sub-scores.

Conclusion
In the right clinical context, neuromuscular ultrasound has an important role in establishing the diagnosis of ALS and distinguishing it from potential disease mimics.In the most recent ALS diagnostic guidelines, the inclusion of neuromuscular ultrasound lends further support to its use in the diagnosis of ALS.At present, muscle ultrasound shows most promise in detecting fasciculations and changes in muscle thickness and echo intensity.More research is required to provide robust evidence for its use as a biomarker of disease progression.

The utility of MRI in the diagnosis and monitoring of ALS
MR imaging has long been used to assess upper motor neuron involvement in ALS.A recent systematic review identified 116 such studies (Grolez et al., 2016), although it is fair to say this has not entered routine clinical use.More recently, attention has turned to MR imaging of peripheral nerves and skeletal muscles to detect lower motor neuron involvement (Kriss and Jenkins, 2022).
There are numerous potential advantages of MR imaging, both as a diagnostic test and as a biomarker of disease progression.In contrast to needle EMG, it is entirely non-invasive and pain free, has excellent spatial resolution and samples a large volume of muscle.Balanced against these are the limited availability and high cost of MR scanning time, and the inability of some patients with respiratory impairment to lie flat in the scanner.Despite these limitations, developing techniques capable of detecting lower motor neuron involvement in ALS is an active area of research.

Peripheral nerve assessment
MR imaging using conventional clinical sequences (T1 and T2weighted) is often performed in order to exclude alternative diagnoses.A retrospective analysis of MR imaging of the cervical and lumbar plexus in 60 patients with limb-onset ALS showed abnormalities in 85% (Staff et al., 2015).These included T2 hyperintensities within the peripheral nerves (60%), and nerve enlargement (30%).The authors speculated that these changes resulted from axonal degeneration and macrophage infiltration.However, the imaging changes did not correlate with disease severity, and did not distinguish patients with ALS from those with multi-focal motor neuropathy.Furthermore, patients were scanned at a median of 12 months after symptom onset, so it is unknown whether the early stages of denervation would have been seen using conventional MR sequences alone.Gerevini et al. also observed frequent T2 hyperintensities in the cervical nerve roots of ALS patients which correlated with faster disease progression (Gerevini et al., 2016).
Diffusion tensor imaging (DTI) measures the diffusion of water molecules in multiple directions, the total diffusion expressed as mean diffusivity (MD) and the relative diffusion in each direction expressed as the fractional anisotropy (FA).These provide reproducible and quantitative measures of tissue microstructure, with MD typically increasing (Winklewski et al., 2018) and FA decreasing with axonal degeneration (Werring et al., 1999).DTI of the peroneal and tibial nerves shows reduced FA at baseline in ALS patients compared to healthy controls (Simon et al., 2017), and at least in the tibial nerve declines further with disease progression.DTI of the sciatic nerve also shows significantly lower FA than controls (Lichtenstein et al., 2021).Although these results are promising, DTI requires much longer scanning times than standard MR sequences as well as extensive post-processing of the resulting images, and as yet this has limited its role to the research setting.

Skeletal muscle assessment
Early studies of MR imaging in ALS simply measured muscle volume as a surrogate marker of lower motor neuron loss.However, studies in the leg muscles show no difference at baseline compared to controls (Bryan et al., 1998), and no consistent change in the volume of the intrinsic hand muscles, tibialis anterior or the tongue over 12 months follow up (Jenkins et al., 2013).
A more promising approach is to use a combination of MR sequences to reveal changes in the muscle microstructure, which occur far earlier in the disease process than atrophy.T2-weighted sequences are primarily influenced by tissue water content.Denervation results in fluid shifts which in turn result in a prolonged T2 relaxation time (Wessig et al., 2004).This can be assessed semiquantitatively, e.g., with short-tau inversion recovery (STIR) imaging (Kamath et al., 2008), or quantitatively with T2-mapping.
STIR-imaging shows significant hyperintensities in the leg muscles of ALS patients, but not in the tongue (Klickovic et al., 2019).ALS patients show increased T2 relaxation times in the legs muscles at baseline compared to controls, and changes in T2 relaxation time correlate with CMAP amplitude and disease progression (Bryan et al., 1998).Advances in scanning technique make it feasible to study all 4 body regions in a single 30-40 minute scan session, and Jenkins et al. showed a significant increase in T2 signal in the tibialis anterior muscles over a 4 month follow-up (Jenkins et al., 2018).No significant changes were seen in any other body region.A subsequent study with assessments at both 4 and 12 months showed significant changes in a number of lower limb muscles, but again not consistently in the upper limbs, thoracic paraspinal muscles or tongue (Jenkins et al., 2020).Interestingly, the propensity for changes in the lower limb muscles appeared independent of the site of disease onset.
T1-weighted sequences are primarily influenced by fat content, and hence are sensitive to fatty infiltration as occurs later in the disease.The 3-point Dixon technique provides a quantitative measure of muscle fat fraction.This is higher in the calf muscles of patients than controls and correlates with the ALS-FRS score (Klickovic et al., 2019).

Assessment of the tongue
A small number of studies have focussed on the tongue.Patients with severe bulbar involvement show reduced tongue volume compared to controls (Cha and Patten, 1989;Konagaya et al., 1990), and a small-scale pilot study using DTI showed a reduced FA and increased MD in ALS patients' tongue muscles compared to controls (Lee et al., 2018).The largest study to date assessed tongue volume and T1 signal intensity.Disappointingly, there was no difference in either of these measures between 175 ALS patients and 104 healthy controls, although T1 intensity did distinguish patients with bulbar vs limb onset (Hensiek et al., 2020).
To summarise, using MR imaging to assess muscle volume is a crude and rather insensitive tool.Much like measurement of the CMAP amplitude, it is unlikely to ever detect early disease due to compensatory re-innervation, and appears unsuitable as a biomarker of disease progression.Semi-quantitative and quantitative sequences show some promise as a marker of disease progression C. Shin-Yi Lin, J. Howells, S. Rutkove et al. Clinical Neurophysiology 162 (2024) 91-120 in the lower limbs, but have so far proved disappointing in the other body regions.This may reflect the relative ease of imaging the legs compared to other body regions; for example, imaging of the upper limbs is often limited to one arm due to the difficulty of positioning the limb in the centre of the scanner bore, and imaging the tongue is prone to movement artefact.

Dynamic MRI of muscle
A fundamental limitation of current muscle MRI is that although it provides detailed information on muscle structure, the images are static and provide no indication of muscle function.Fasciculation is one of the earliest changes seen in ALS and is a promising biomarker of early disease in that it can occur in muscles before the development of motor unit remodelling (de Carvalho and Swash, 2013).
Investigators using diffusion-weighted MR to study primary muscle disease first reported the presence of sporadic signal voids within the muscles (Lemberskiy et al., 2014).At the time these were considered an inconvenient and unexplained artefact.Although primarily designed to be sensitive to diffusive (Brownian) motion, diffusion-weighted imaging sequences are sensitive to any process which re-arranges the relative spatial position of water molecules within the imaging voxel, including that which occurs during muscle contraction (Steidle and Schick, 2015).Subsequent studies demonstrated that the signal voids had a reproducible recruitment sequence, had dimensions and twitch profiles consistent with human motor units (Birkbeck et al., 2020;Heskamp et al., 2021), and their occurrence strongly correlated with surface EMG detected fasciculation potentials (Schwartz et al., 2018).Based on these observations, the technique was named motor unit MRI, or MUMRI.
MUMRI is much faster to perform than conventional DWI because it acquires images in only one direction, typically along the muscle fibre axis.By running the DWI sequence repetitively, it is possible to study the frequency and distribution of fasciculations in multiple muscles simultaneously.This approach demonstrated a significantly increased fasciculation rate in the leg muscles of ALS patients compared to healthy controls (Whittaker et al., 2019), the MUMRI acquisition time being only 3 minutes (Fig. 14).
MUMRI has been applied in multiple body regions, and a study of fasciculation rates in the tongue, biceps brachii, thoracic paraspinals and lower leg muscles in 10 ALS patients and 10 healthy controls showed increased rates in all regions except the tongue (Heskamp et al., 2022).Studies are ongoing to assess its sensitivity compared to surface EMG and ultrasound in detecting fasciculation, and a multi-centre trial is soon to start assessing its potential role as a diagnostic tool.

Conclusion
MR remains a relatively limited and expensive resource, and is unlikely to ever become as readily available as EMG and ultrasound.Some MR sequences take a considerable time to perform, which patients with respiratory impairment may find impossible to tolerate.However, the attraction of a single non-invasive test capable of detecting both early (fasciculation with MUMRI), intermediate (denervation with STIR and T2 mapping) and late (fatty infiltration with T1-weighted) disease stages is obvious.MR imaging in ALS remains primarily a research tool and its place in the diagnosis and monitoring of patients is as yet unproven.However, with the increasing use of MR in the routine work-up of patients with suspected ALS it may yet find a role in combination with other neurophysiological and imaging modalities.

Final remarks
In this century, a goal for treating physicians should be to have reasonable clinical certainty that the patient has ALS, even when the clinical features are largely confined to a single territory.This will allow therapies to be given early in the disease course when they are likely to be of greater benefit.Neurophysiological studies are an essential component of the work-up and monitoring of patients with MND/ALS, but there are other potential biomarkers of disease (e.g., neurofilaments, imaging), though they are, as yet, less well validated than many of the neurophysiological measures.
For individual patients, neurophysiological studies are important for diagnosis and monitoring, and may help answer a number of questions: 1. Can we refine the LMN findings further?We can use MUNIX and the various forms of MUNE to detect (and follow) the loss of motor neurons.We can use ultrasound (and possibly MRI) to detect fasciculations and changes of muscle atrophy.These modalities are particularly valuable for the tongue and bulbar muscles, tibialis anterior and biceps brachii, and some otherwise less-accessible muscles, such as the diaphragm.In assessing LMN disease many of the techniques discussed here (CMAP scans, MUNIX and the various forms of MUNE, EIM, nerve and muscle ultrasound) may be of use, as may be other biomarkers (e.g., neurofilaments), and it is doubtful that they will be superseded as imaging techniques become more sophisticated.5.Only further studies will establish whether these potential biomarkers are more useful than functional measures, such as forced vital capacity and ALS rating scales, though a number of studies do indicate this already.An advantage over global rating scales is that the neurophysiological measures can be applied to individual muscles or limbs, and are thus appropriate for following patients in the earliest stages of ALS.As discussed above, this principle is already being applied to a number of neurophysiological procedures.

Declaration competing of interest
Hugh Bostock and James Howells receive from UCL shares of the royalties for sales of the Qtrac software.Sanjeev Nandedkar is an employee of Natus Medical Inc.He has no conflicts to report.The other authors have no potential conflicts of interest to declare.
Seward Rutkove has equity interest in and serves as a consultant and scientific advisor to Myolex, Inc., and Haystack Diagnostics, Inc., companies that design impedance devices for clinical and research use; he is also a member of Myolex's Board of Directors.The companies also have an option to license patented impedance technology of which he is named as an inventor.
, the EMG findings detailed in the Figure legend and reflex findings for tibialis anterior bilaterally indicate LMN and UMN abnormalities.The EMG findings listed in the Legend indicate widespread disease.

Fig. 3 .
Fig. 3. Methods of motor unit number estimation (MUNE).Recordings from the abductor pollicis brevis muscle of healthy subjects are shown.(A, B) Incremental stimulation method.(C) Multiple point of stimulation (D, E) Compound muscle action potential (CMAP) scan.See text for details.

Fig. 4 .
Fig. 4. Amplitude analysis of a compound muscle action potential (CMAP) scan in a normal muscle is illustrated in A-E.(A) Amplitude recorded during scan.(B) Amplitude after arranging values in ascending order.The curve shows monotonic increase.Difference in successive amplitudes is a 'step'.(C) Distribution of step amplitude in the scan.(D) Step amplitude is arranged in descending order and summated.The sum is normalized to the CMAP amplitude.The number of largest steps generating 50% of the CMAP amplitude is D50.(E) Step amplitudes are arranged in descending order (grey line).Regression analysis (dashed line) of the large amplitude steps is used for STEPIX calculations.The traces for this scan are shown in Fig. 4A.(F) Scan from a patient with amyotrophic lateral sclerosis shows steps with high amplitude (Fig. 4B).STEPIX is reduced while AMPIX is increased.

Fig. 5 .
Fig. 5.An example of MScanFit motor unit number estimation (MUNE) is illustrated.The method was designed to generate a model that would generate a CMAP scan as similar as possible to the recorded one.See text for details.

Fig. 6 .
Fig. 6.Motor unit number index and compound muscle action potential (CMAP) scan recordings from (A) a healthy subject and (B) a patient with ALS.The CMAP waveform, surface interference pattern (SIP) signals at different force levels, regression analysis, and traces from the CMAP scan study are shown.In the regression analysis plot, the control subject and patient measurements are shown by open and filled circles, respectively.See text for details.

Fig. 7 .
Fig. 7. Motor unit number index (MUNIX) and compound muscle action potential (CMAP) scan analysis in two patients with ALS and reduced CMAP amplitudes.(A) Surface EMG interference pattern (SIP) recordings show large amplitude fast firing surface motor unit potentials (SMUPs).Motor unit size index (MUSIX) is increased.CMAP scan shows large steps.The firing rate of SMUP indicated by the arrow is > 20 Hz.Its amplitude matches a large step in the CMAP scan.(B) MUSIX is normal.SIP signal does not show large amplitude potentials.CMAP scan shows gradual amplitude change without any large steps.

Fig. 10 .
Fig. 10.Cartoons demonstrating application of electrical impedance myography (EIM) (not to scale). A. Surface EIM using a multi-electrode handheld array.Current (blue) is applied through the 2 sets of outer electrodes and the resulting voltages (yellow) measured via the pair of inner electrodes.B. Needle impedance-EMG.A modified EMG needle contains four circular impedance electrodes along the barrel proximal to the EMG tip (which itself contains a inner needle E1 electrode and barrel tip E2 electrodes).Again, the outer impedance electrodes apply current (blue) and the inner measure the resulting voltages (yellow).

Fig. 12 .
Fig. 12. Longitudinal and cross-sectional images of C6 nerve root in a control subject and an ALS patient.The diameter and cross-sectional area (CSA) of the ALS patient were smaller than those of a control subject.

Fig. 14 .
Fig. 14.Example of fasciculation maps in ALS patient compared to healthy controls.Top row -heat map of fasciculations in a patient with ALS from 5 slices of MUMRI imaging, overlaid on to structural T1 weighted images.Bottom row -heat map of fasciculations in a healthy control.Note the different scale due to the low number of fasciculations in the control.

Table 1
Existing studies in motor neuron disorders using MScanFit motor unit number estimation (MUNE).
2. Can we define UMN weakness?We can see if reflex inputs can potentiate a maximal voluntary contraction significantly.Can we define hyperreflexia?We can look for H reflexes at rest in muscles that only have a demonstrable H reflex if contracting.3. Can we demonstrate cerebral involvement?We can study single-shock TMS [CMCT] and double-shock TMS and perhaps other cerebral processes, as discussed in other chapters.4. Can we measure progression of disability accurately?What measures are sufficiently accurate to be useful in clinical trials?