Application of functional near-infrared spectroscopy in the healthcare industry: A review

Functional near-infrared spectroscopy (fNIRS), a growing neuroimaging modality, has been utilized over the past few decades to understand the neuronal behavior in the brain. The technique has been used to assess the brain hemodynamics of impaired cohorts as well as able-bodied. Neuroimaging is a critical technique for patients with impaired cognitive or motor behaviors. The portable nature of the fNIRS system is suitable for frequent monitoring of the patients who exhibit impaired brain activity. This study comprehensively reviews brain-impaired patients: The studies involving patient populations and the diseases discussed in more than 10 works are included. Eleven diseases examined in this paper include autism spectrum disorder, attentionde ̄cit hyperactivity disorder, epilepsy, depressive disorders, anxiety and panic disorder, schizophrenia, mild cognitive impairment, Alzheimer's disease, Parkinson's disease, stroke, and traumatic brain injury. For each disease, the tasks used for examination, fNIRS variables, and signi ̄cant ̄ndings on the impairment are discussed. The channel con ̄gurations and the regions of interest are also outlined. Detecting the occurrence of symptoms at an earlier stage is vital for better rehabilitation and faster recovery. This paper illustrates the usability of fNIRS for early detection of impairment and the usefulness in monitoring the rehabilitation process. Finally, the limitations of the current fNIRS systems (i.e., nonexistence of a standard method and the lack of well-established features for classi ̄cation) and future research directions are discussed. The authors hope that the ̄ndings in this paper would lead to advanced breakthrough discoveries in the fNIRS ̄eld in the future.


Introduction
The purpose of this paper is to review the applications of functional near-infrared spectroscopy (fNIRS) for diseased populations in the healthcare industry. The aging people of the world currently have various psychiatric and neurological impairments. Further, the brain functions of these patients are profoundly impaired, thereby restricting their independence in daily life. The aggravated state of these patients results in the constant involvement of caregivers to live their lives. The fatality rate is very high in the case of brain diseases. The impairments a®ecting this population include various types of dementias that are associated with memory loss and impaired executive functioning. The common forms of dementia are Alzheimer's disease (AD), vascular dementia, Lewy body dementia, medication-induced dementia, and frontotemporal disorder. AD is the most widespread form of dementia, accounting for almost 60% of all dementia-related cases. 1 Stroke is a type of permanent impairments that are caused by either a blockage in a brain vessel or by its bursting; thereby resulting in the death of the brain cells that are associated with the distribution of blood oxygen through that vessel. Stroke is treated as a medical emergency and can be highly fatal. Parkinson's disease (PD) is the most common form of movement impairments that are known as Parkinsonian syndromes. PD is associated with trembling and experiencing hardship while walking and during movements and coordination. Epilepsy involves recurring, impulsive seizures, or disturbed brain activity that causes changes in the attention span or behavior of a patient. Psychiatric impairments muddle a patient's thoughts, perceptions, characteristics, and their ability to relate to others. Common psychiatric impairments include anxiety disorders, bipolar disorder (BD), depression, schizophrenia (SZ), eating disorders, impulse control and addiction disorders, and personality disorders. Impairments other than psychiatric ones are mostly irreversible and progressive. Several medications and rehabilitation techniques are utilized to reduce the gradual decline and to improve the quality of life of these patients. The degradation starts much earlier in the brain compared to when the symptoms¯rst appear. Therefore, the early detection of these impairments is vital. The advancements made in various neuroimaging technologies and the researches that have used them have paved the way for studying as well as detecting brain impairments.
The neuroimaging modalities include eventrelated potentials measurement using electroencephalography (EEG), magnetic¯eld measurement using magnetoencephalography (MEG), radioactive tracer-based positron emission tomography (PET), gamma emission-based single-photon emission computed tomography (SPECT), and functional magnetic resonance imaging (fMRI). These modalities have allowed for valuable advancements made in the understanding of many of the neurological impairments. Examinations can only be conducted in restricted environments using fMRI, MEG, PET, and SPECT due to the large size of the machines involved and their lack of mobility, which limits the design of the study. Moreover, these systems are highly vulnerable to motion artifacts, costly, and invasive due to the insertion of radioactive tracers. They have low temporal resolution, which makes them inappropriate for conducting repeated measurements. 2 EEG has a high temporal resolution, but it lacks spatial resolution and is vulnerable to motion artifacts. 3 The rehabilitation of patients with impaired brain functions is essential; however, these modalities cannot be used simultaneously with rehabilitating techniques, such as electric stimulation, as these techniques are a®ected by electric and magnetic¯elds. 4 In contrast to established neuroimaging methods, fNIRS has proven its worthiness during the last decade. Most of the human tissues are comparatively more translucent than oxygenated (HbO) and deoxygenated hemoglobin (HbR) in a spectrum between 650 nm and 1000 nm. Therefore, optical wavelengths in this range are used to measure temporal transformations of HbO and HbR. 5 The photons, emitted by the light sources attached to the head, moving through the di®erent layers in the brain are either absorbed or scattered. Photodetectors are placed on the skin to receive these photons that travel in an expected banana-shaped photonic°ow to reach the surface. Conventionally, fNIRS uses two wavelengths; however, introducing additional wavelengths also helps in achieving better neuronal activation. 6 fNIRS has been utilized in studies to classify the sensory responses and motor cortex activation levels of di®erent¯ngers. 7,8 fNIRS can reveal the underlying neuronal networks and their complex connections in the form of functional connectivity. 4,9 Various algorithms and techniques have been developed and explored using fNIRS to improve the brain-computer interfaces (BCIs) to help physically disabled persons. [10][11][12][13][14][15] This paper reviews the research works conducted to advance the understanding of the e®ects of various diseases on our brain using fNIRS. These studies mostly involve patients with degenerative brain or psychiatric disorders. Each section of this paper is devoted to a single disease to summarize the associated research works and their¯ndings. For every disease, we created di®erent subsections based on the performed task during the fNIRS recording.

Autism Spectrum Disorder
Autism spectrum disorder (ASD), also known as pervasive development disorder (PDD) or Asperger's disease, is a mental disorder that a®ects communication. ASD is known as a developmental disorder, and it begins during childhood or even during infancy. Once an individual develops ASD, it usually remains throughout his/her life. However, several treatments and medications can improve the quality of life or even completely cure all the related symptoms. ASD patients may exhibit symptoms such as problems in talking and interacting with others, displaying repetitive behaviors, a lack of interest, or mismatched facial expressions. It is critical to diagnose ASD during early childhood because the treatment at a young age results in a much better outcome. Several studies have been conducted using fNIRS to elucidate the neuronal mechanism involved in ASD: The task-wise distribution of ASD papers is presented in Fig. 1, and all the studies are outlined in Table 1.

Visual task
For ASD children, a lower level of activation in the right inferior frontal gyrus was related to the inability to recognize his/her face showing impairment in that region. 16 The PDD patients showed a signi¯cantly lower HbO response in the prefrontal cortex (PFC) while watching fearful facial expressions when compared to healthy persons. 17 The children with PDD showed a lower HbO response in the bilateral temporal regions and especially in the right hemisphere while watching and imitating tasks when compared to healthy persons. 18 During the tasks showing familiar and unfamiliar faces, the children with ASD were differentiated from those with attention-de¯cit hyperactivity disorder (ADHD) using a support vector machine (SVM)-based classi¯cation. 19 The children with ASD exhibited a lower HbO response and abnormal connections in the PFC during the joint attention situation in joint and nonjoint attention tasks. 20 While watching human faces, the ASD patients showed bilateral temporal-occipital activation as compared to healthy individuals having right hemisphere activation. 21 The fNIRS-based neurofeedback was provided to the ASD patients during the facial identity recognition training to achieve better outcomes. 22 The infants who exhibited lower activation in the inferior frontal and posterior temporal regions in response to the social video clips were diagnosed with ASD in their early childhood. 23

Verbal°uency task
The poor performance of PDD patients in this task was related to the low HbO level in the bilateral frontal region and speci¯cally in the right hemisphere when compared to healthy persons. 24 The HbO levels of both ASD and healthy children were similar, but the adults with ASD showed lower cognitive activation as compared to healthy persons in the bilateral PFC. 25 The HbO levels of the patients with Asperger's disease in the PFC were signi¯cantly lower than those in healthy persons during the task period, thereby exhibiting the taskrelated impairment. 26 The HbO response in the left ventrolateral and dorsolateral PFCs of ASD patients was observed to be lower than that of healthy persons, but was not di®erentiable from that of ADHD patients. 27 Compared to healthy persons, the patients with ASD showed decreased HbO activity in the bilateral frontotemporal region, which was also a di®erent response from that of SZ patients. 28 In addition to the lateral frontopolar cortex activation that is observed in healthy persons, the medial frontopolar cortex of highfunctioning ASD patients also exhibited activation, thereby demonstrating the compensation mechanism of an impaired brain. 29

Cartoon watching
A functional connectivity analysis of the young children with ASD, compared to typically developing children, revealed lower network e±ciency in the prefrontal, temporal, and occipital regions. 30 The global and local network e±ciencies, based on a functional connectivity measure using HbR and total hemoglobin (HbT) levels, decreased as the age of the ASD children increased: Their HbO-based network e±ciency was reduced as well. 31 The spatial complexity analysis of functional connectivity revealed impaired information exchange in the right hemisphere of the ASD children compared to that of healthy children. 32 The long-range temporal correlation values measured using HbO levels were lower in the left temporal regions, and exponents obtained through the detrended°uctuation analysis were inversely linked with the severity of ASD. 33 The PFC of ASD children was largely responsible for the deteriorated functional connectivity. 34

Go/No-go task
When compared to healthy children, highfunctioning children with ASD showed lower HbO activation levels in the right PFC during response inhibition tasks. 35 In a frontotemporal examination, the ASD patients showed impaired cortical activation in the inferior frontal gyrus and middle frontal gyrus. 36 The administration of methylphenidate in ASD-comorbid ADHD children was revealed to suppress the hemodynamic response. 37

Resting state
The resting state functional connectivity (RSFC) calculated based on low-frequency spontaneous°u ctuations in the anterior PFC was higher in children with ASD than in healthy children, and it was associated with the Autism Diagnostic Observation Schedule scores. 38 The children with ASD exhibited lower interhemispheric RSFC in the temporal cortex and altered local connections in both their temporal cortices. 39 The°uctuations of the HbO and HbR levels at a resting state were higher in children with ASD than in healthy children and were used in the SVM-based classi¯cation along with RSFC. 40

Audio stimulus
Compared to healthy children, the children with ASD exhibited weaker cortical activity in the left temporal cortex in response to phonemic words and the right temporal cortex in response to prosodic ones. 41 The bilateral auditory cortex lesions in the ASD patients exhibited similar responses to those of healthy persons during attentive listening; however, this di®ered in the PFC that exhibited an attention impairment instead of an impaired auditory cortex. 42 The infants who showed reduced and leftlateralized temporal cortical activation in response to vocal and nonvocal sounds were diagnosed with ASD a few years later when compared with infants who remained healthy. 23

Stroop task
The study that conducted a Stroop task did not reveal any di®erences in the hemodynamic responses among ASD, ADHD, and healthy children. 35

Expression task
The ASD children showed reduced PFC activation and were more expressive for nonemotional pictures as compared to healthy children while describing their mental state in response to viewing a black and white picture depicting human eyes. 43

Imitation task
The children with ASD showed enhanced neuronal activation while performing imitation tasks after undergoing imitation training when compared to the low activation levels before undergoing the training. 44

Stop-signal
Compared to healthy persons and ADHD patients, the ASD patients showed a reduced HbO response in the ventrolateral PFC and, compared to ADHD patients, they exhibited impaired activation in the PFC during inhibitory control tasks. 27 2.11. Color and shape span task A weighted separability index based on the HbO levels was utilized to reveal signi¯cant di®erences between the left dorsolateral PFCs of ASD patients and healthy persons during a working memory task. 45 2.12. Emotional facial recognition task ASD patients in general and speci¯cally those who paid a higher level of attention to details exhibited impaired cortical activity in the left frontotemporal region. 28

Attention-De¯cit Hyperactivity Disorder
ADHD is a brain impairment that a®ects patients by causing lack of attention, excessive activity, and hastiness. This impairment is observed during childhood, and can remain throughout one's life. Most ADHD patients are diagnosed at an elementary school age when they are identi¯ed to be different from other children of the same age. A su®ering child mostly overlooks details while working, makes careless mistakes,¯dgets or squirms while sitting, talks without listening to others, or is unable to wait. The underlying reasons for ADHD are still unknown, and hence, it can neither be prevented nor fully cured. However, various therapies and medications can improve the quality of life of ADHD patients by reducing or managing the symptoms. Many research studies have been conducted using fNIRS to uncover the neuronal behavior that causes the symptoms. The task-wise distribution of ADHD papers is presented in Fig. 2, and the details are summarized in Table 2.

Go/No-go tasks
The children with ADHD showed little cortical activation in the right PFC during the inhibitory control in a no-go situation when compared to the higher activation observed in healthy children. 35,46 The administration of methylphenidate to children with ADHD resulted in an improved HbO response in the right lateral PFC that was related to a better performance during inhibitory response. 47 The reduced inferior and middle frontal gyri showed better hemodynamic activation due to methylphenidate administration, but this e®ect was not witnessed during the placebo-based activation. 48 The atomoxetine-administered children with ADHD exhibited similar improvements in cortical activation to those administered with methylphenidate during inhibitory control. 49 In a classi¯cation study, the reduced activation patterns in the region of the right PFC were useful for better distinguishing between children with ADHD and the healthy ones by resulting in high area-under-the-curve values and sensitivity levels. 50 The children with ADHD showed an overall reduced left frontopolar cortex activation, especially during response inhibition. 51 Methylphenidate improved activation levels in children with ADHD, and this medicated response has been utilized e±ciently for di®erentiating between ASD and ADHD. 37

N-back task
Compared to healthy individuals, the ADHD patients showed a decreased activation in the ventrolateral PFC during working memory tasks, especially in the case of high load conditions, such as a two-back task. 52 A reduced HbO response was witnessed in ADHD patients during a working memory two-back task, which was unrelated to the reduced HbO response due to response inhibition in the stop-signal task. 53 A complexity analysis via the permutation entropy value revealed its inverse correlation with hemodynamic activation in the PFC whereas its values of the right dorsolateral PFC of children with ADHD were higher than those of healthy children, and the entropy value was correlated with disease severity. 54 A machine learning-based classi¯cation study using multidomain measures including blood fatty acid pro¯les, psychological parameters, and fNIRS performed e±ciently in di®erentiating children with ADHD from healthy children, and utilization of HbR levels generated better results than HbO levels. 55 A multivariate pattern analysis-based classi¯cation showed 86% accuracy in di®erentiating between healthy and ADHD children and identi¯ed highly useful brain regions. 56

Stroop task
The boys with ADHD showed impaired dorsolateral PFC activation and higher brain activity in the right side as a compensation mechanism when compared to the activation in healthy boys. 57 The HbO responses were signi¯cantly increased in the bilateral inferior-PFC and especially in the inferior lateral region of healthy individuals as compared to those of ADHD patients during a Stroop color-word task. 58 The polymorphism of synaptosomal-associated protein 25 gene was associated with methylphenidate-related HbO and HbR changes in ADHD patients. 59 The go/no-go task produced signi¯cant di®erentiable changes between the HbO levels in the PFCs of healthy and ADHD patients while the Stroop task did not. 35 Compared to the Stroop task, the reverse Stroop task showed signi¯cant di®erentiable HbO responses among healthy, ASD, and ADHD children in the right lateral PFC. 60

Stop-signal task
Compared to those of healthy persons, the HbO and HbR responses in ADHD patients were weakened during the inhibitory process. 53 Further, during response inhibition, the cortical activity in the left ventrolateral PFC signi¯cantly di®ered in ADHD patients when compared to ASD patients and healthy persons. 27 A longitudinal study showed that ADHD patients exhibited improved prefrontal responses after a single dose of methylphenidate,  and its long-term use yielded an activation equivalent to that of a healthy person. 61

Visual task
The children with ADHD did not exhibit cortical activity while watching angry faces, which illustrates the impairment of ADHD children to recognize an angry face. 62 Using a¯ve-fold crossvalidation in an SVM-based classi¯cation of 24 channels of fNIRS data proved to be fruitful in achieving a maximum accuracy of 84% in di®erentiating between patients with ADHD and those with ASD. 19 Neurofeedback training using a visual display yielded good results as it reportedly reduced the ADHD symptoms in children. 63

Oddball task
Compared to healthy children, the ADHD children exhibited a lack of activation in the right prefrontal and inferior parietal cortices, which was normalized after atomoxetine administration. 64 Administration of methylphenidate to ADHD children resulted in a normalization of activity in the right PFC but not in the inferior parietal lobe. 65

Verbal°uency task
The patients with ADHD showed a lower cortical activation in the inferior frontal region compared to healthy persons, and the activation was inversely related to task performance. 66 The hemodynamic responses of ADHD patients were di®erentiable from those of healthy persons but were similar to those of ASD patients, thereby restricting the use of verbal°uency tasks (VFTs) in the multicategory classi¯cation. 27

Olfactory task
In the temporal, somatosensory, and inferior frontal cortices, the cortical activation of ADHD patients was reduced compared to that of healthy persons. 67 Administering methylphenidate to ADHD children improved the HbO responses in the temporal cortex while the cessation of the medication resulted in the recurrence of diminished activation. 68 3.9. Trail-making task The children with ADHD showed an increase in HbO and cerebral blood volume (CBV) levels during short-attention tasks, while healthy children only showed increased activity during long-attention tasks. 69 3.10. Object and spatial working memory task The cortical activations during object working memory tasks were higher than those during spatial working memory tasks for ADHD and healthy children, which showed no signi¯cantly di®erent patterns. 70 3.11. Visuospatial task Compared to healthy children, the ADHD children showed higher activation in the PFC in response to distraction during the task owing to the impairment in inhibition control. 71

Continuous performance task
The long-term usage of atomoxetine medicine sig-ni¯cantly improved the HbO and HbR responses in the right dorsolateral PFC of children with ADHD. 72

Audio task
Compared to healthy individuals, the ADHD patients exhibited lower activation levels in the superior temporal gyrus in response to angry prosody and supramarginal gyrus activation due to the compensatory mechanism. 73

Epilepsy
Epilepsy is a disease owing to which a patient suffers from seizures. The seizures can a®ect patients in many ways and can range from simply staring into space to experiencing their full-body shaking or even falling on the ground. In some cases, the symptoms are visible in the whole body, yet the cause of epilepsy is from brain impairment. To study epilepsy, various studies were conducted in di®erent environments using fNIRS. The task-wise distribution is shown in Fig. 3, and the details are outlined in Table 3.

Ictal or seizure recording
In this type of video-EEG experiment, the patient is continuously monitored using video recording, and the exact timing of each seizure is matched with that of the recording. While this is happening, the subject can be in a resting or moving state. Initially, the onset of seizure was related to an increase in CBV. 74 In a study performed after this, a contradictory result showing a decrease in CBV at the time of seizure onset appeared. 75 It was later revealed that the increase or decrease in HbO levels in patients is associated with the type of seizure. 76 Cerebral oxygenation was utilized to distinguish between complex partial seizures (CPS) and rapidly secondarily generalized CPS. 77 The regional CBV was increased via some channels, thereby facilitating the identi¯cation of the focal aspect of seizures. 78,79 Another study on children yielded similar results showing di®erent CBV changes in di®erent seizure types. 80 In a detailed study on absence seizures, the HbO level decreased while the HbR level increased. 81 Further, another study proved fNIRS to be e®ective in drug management as an anticonvulsant medication administered to an infant resulted in a reduction in seizure frequency. 82 Generalized spike-and-wave discharges are associated with absence epilepsy and exhibit oxygenation before the onset followed by deoxygenation, which is again followed by oxygenation and then returning to the baseline level in the frontal cortex. 83 An initial decrease in HbO level, known as the initial dip, that precedes the increase in HbO and HbT levels was found at the onset of ictal seizures. 84 In a study focusing on temporal lobe seizures, the HbO and HbR changes were seen in the focal point (i.e., temporal region) as well as in the remote areas such as in the frontal or parietal cortices. 85 A recording of supplementary motor area (SMA) seizures in a nineyear-old girl revealed an increase in cerebral blood°o w (CBF), which started in the SMA and extended to the premotor and sensorimotor cortices. 86 fNIRS was found to be helpful and, in some cases, it performed better than EEG in detecting frontal lobe seizures that show increased HbO and HbT levels and variable HbR responses in the focal as well as in the contralateral regions. 87 The sensitivity and speci¯city estimates resulting from a decrease in the HbR were higher than those resulting from an increase in HbO and HbT levels. 88 In the preictal and postictal periods, the regional cerebral oxygenation was increased while it was decreased near the onset time of ictal seizures, as shown in previous studies. 89 These seizure studies on epilepsy are largely a®ected by the issue of motion artifacts, which was resolved by using collodion-¯xed prismbased optical¯bers. 90 The decrease in HbR was more signi¯cant than the increase in HbO and HbT levels in a study on posterior epilepsies. 91 The increase in oxygen saturation was associated not only with ictal but also with epileptiform discharges without seizures. 92 The HbO values were observed to increase in both hemispheres, but the increase was more pronounced in one hemisphere, which allows for the localization of the epilepsy-a®ected region using fNIRS. 93 Utilizing a wireless fNIRS device to detect a seizure achieved a very low accuracy in seizure detection and contradicting results when utilized with a generic algorithm. 94 Among epilepsy studies, an EEG-fNIRS study provided better results in detecting interictal epileptic discharges than an EEG-fMRI one. 95

Verbal°uency task
The determination of language lateralization was achieved with a higher accuracy by displaying clear activation in the language areas of the brain in children as well as in adults. 96 The activation in the left Broca's area was higher than that in the right hemisphere, thereby showing the left-hemisphere dominance in children. 84,97 The damage in the brain results in a reorganization, as shown in a six-yearold child, that the left-hemisphere dominance was

Word generation task
One of the earliest researches on epilepsy using fNIRS noninvasively accessed the language dominance areas in the brain by showing activations on the same side as determined by the Wada test. 99 Patients exhibited better language lateralization results before surgery than after surgery. 100 The fNIRS was used in combination with fMRI and MEG to improve the detection of language lateralization. 101

Resting state
When comparing methods to detect seizure activity e±ciently, the Bayesian general linear model was more accurate and reliable than the wavelet generalized least-square algorithm. 102 In terms of estimating the hemodynamic response, the Volterra kernel expansion method showed better results in the cases where the conventional methods failed. 103 A model based on long short-term memory in the recurrent neural networks demonstrated an e±cient performance in seizure detection in a hybrid EEG-fNIRS study. 104

Receptive language task
While listening to storytelling, the Wernicke's and Broca's areas in both hemispheres were activated in a nine-year-old Yiddish boy. 97 The left hemisphere was dominant at¯rst in an epilepsy patient who was six-year-old while the right hemisphere was dominant when the patient turned 10-year-old owing to the functional brain reorganization due to the damage caused by epilepsy. 98

Passive motor task
Once a portion of the brain is surgically removed to treat epilepsy, the functional loss is recovered by the healthy portion of the brain via reorganization. The movement of right arm was impaired due to a surgery performed in the left hemisphere, but rehabilitation through passive movement therapy showed activation in the right hemisphere. 105

Direct cortical stimulation
The cortical stimulation of the left temporal region resulted in a rise in the HbO and HbR levels in the temporal as well as in the frontal regions, thereby displaying the possible functional connectivity of the language area. 106

Motor task
The epileptic patients showed an activation of a comparably smaller amplitude than the activation exhibited by healthy persons while gripping a soft item using their right hand. 107

Reading task
In a patient with reading epilepsy, the seizure activity was located in the left precentral gyrus covering the motor, premotor, and supplementary motor cortex areas while reading aloud or silently. 108

Depressive Disorders
Patients a®ected by various types of depressive disorders su®er from feelings of sorrow and hollowness. This condition intensely a®ects their thinking ability, emotions, and personality inclination. The patients can su®er from a lack of appetite, tiredness, loss of motivation in their daily life, frustration, or anger. The underlying reasons for depression can be a single factor or a combination of various factors like losing someone, trauma, or social, hormonal, or genetic issues. The symptoms can occur at any age, yet in most of the known cases, they developed in adults. The fNIRS has been used by several researchers to study depressive states. The taskwise distribution is presented in Fig. 4, and all studies are outlined in Table 4.

Verbal°uency task
This task has been widely used to test depressed populations. In the earliest studies conducted to discover the neuronal activation in patients with major depressive disorder (MDD) and BD, while monitoring only the left PFC, the increase in HbO levels was signi¯cantly lower in patients than in healthy individuals. 109 This result was also observed while monitoring the bilateral PFC in a study on BD patients. 110 In a study on late-onset MDD patients, a smaller area in the PFC of the MDD patients was activated when compared to the healthy persons. 111 The patients with late-onset depression exhibited impaired community interaction that was positively correlated with a reduced frontopolar HbO response. 112 Compared to healthy persons, the MDD patients showed a lower level of activation in the PFC as well as in the temporal regions. 113,114 The area-under-the-curve and weighted-center values extracted from the timeseries signal associated with the HbO responses showed signi¯cant di®erences between patients with unipolar disorder (UD) and those with BD. 115 The changes in HbO levels in the right dorsolateral PFC were inversely linked with the severity of the disease in MDD patients. 116 Rehabilitating patients with mood disorder (MD) using animal-assisted therapy resulted in signi¯cant improvements in cognitive activation in the PFC. 117 The HbO changes, in general, were steeper in MDD patients than SZ patients, and in the dorsolateral and ventrolateral PFCs were correlated with the Global Assessment of Functioning scores of MDD patients. 118 A higher ratio of positive thoughts versus negative thoughts in MDD patients was related to a higher HbO response in the left dorsolateral PFC and a lower HbO response in the right superior temporal gyrus. 119 Children with depressive disorder showed improved frontopolar activation after receiving psychodynamic therapy for six months. 120 Patients with late-life depression showed a reduced and yet statistically non-signi¯cant activation when compared to AD patients. 121 The increase in depression in MDD patients was associated with increased HbO levels during cognitive activation in the frontopolar PFC and right dorsolateral PFC. 122 A cognitive analysis of BD patients revealed that better social performance was linked with higher activation in the right PFC. 123 The HbO variations of hypomanic BD patients were signi¯cantly higher than those of depressed BD patients in the left dorsolateral PFC. 124 The depressive and euthymic states in BD patients were di®erentiated based on HbO levels in the left temporal region, whereas the intensity of the HbO change revealed the severity of the symptoms. 125 The social functioning of patients with depression during later stages in their lives was correlated with the activation levels in the frontopolar and dorsolateral PFCs while the right ventrolateral PFC predicted the e®ect of rehabilitation. 126 In a detailed study on MDD, patients with melancholic depression exhibited a signi¯cantly lower HbO response in the frontotemporal region when compared to the patients with nonmelancholic depression. 127 Depressed patients who exhibited nonsuppressive e®ects in response to the administration of dexamethasone and corticotropin-releasing hormone showed signi¯cant di®erences in fNIRS responses when compared to the patients who exhibited suppressive behavior. 128 BD and MDD patients with family histories of psychiatric diseases showed highly impaired activation in the PFC compared to those without family histories of psychiatric diseases. 129 MDD patients with positive responsiveness to selective serotonin reuptake inhibitors showed a signi¯cantly higher HbO response compared to nonresponsive MDD patients. 130 Patients with menopausal depression and those with MDD showed lower activations in the right and left dorsolateral PFCs, which di®erentiated them from each other as well as from healthy persons. 131 MDD patients who attempted suicide showed a smaller HbO response in the left precentral gyrus compared to those who did not attempt suicide and to healthy persons, and the HbO response was negatively correlated with impulsivity, hopelessness, and aggression levels. 132 The dorsolateral PFC in BD patients showed higher activation than in SZ patients, demonstrating the less severe verbal memory impairment associated with BD compared to SZ. 133 The HbO response in the right dorsolateral PFC in BD patients with psychotic symptoms was negatively associated with the extent of disease and was lower compared to   134 The depressive patients with the mandatory symptoms showed signi¯cantly lower activation in the left dorsolateral PFC compared to depressive patients without mandatory symptoms, which illustrated that a higher level of impairment in the left dorsolateral PFC is associated with mandatory symptoms. 135 In°ammatory bowel disease is linked with depression, and the patients with the disease also showed reduced cognitive activation in the PFC as observed via various studies on depressive disease. 136 BD patients showed lower activations in the right ventrolateral and dorsolateral PFCs and the bilateral PFC when compared to healthy persons. 137 The changes in HbO activation in MDD patients in the right inferior frontal gyrus and bilateral middle frontal gyri were associated with the extent of the disease and can be observed to distinguish di®erent impairments. 138

Hyperventilation task
The patients with MDD and BD showed a signi¯cantly small reduction in HbO levels during hyperventilation as compared to healthy persons. 109-111

N-back task
The patients with MDD showed a lower level of activation in the lateral PFC and superior temporal region during a two-back task. 139 The late-onset disorder patients also showed reduced activation in the prefrontal and temporal regions during a twoback task, which was signi¯cantly related to lower scores on the Social Adaptation Self-Evaluation Scale. 140 Observing the reduced HbO response in the left frontopolar region and Broca's area was conclusive in di®erentiating between UD and BD patients. 141

Stroop task
During an emotional Stroop task, the patients with BD exhibited similar HbO and HbR responses in the frontal regions to those of MDD patients in response to sad stimuli and di®erent responses in response to happy stimuli. 142 During a Stroop color-word task to measure inhibitory control, the BD patients exhibited lower activation in the frontopolar PFC. 133

Resting state
The RSFC in medicated patients with a®ective disorders was reduced compared to healthy persons in terms of the intrahemispheric, interhemispheric, and intraregional connections but was higher when compared to patients who were not medicated. 143 Thresholding the regional functional connectivity in a resting state facilitated the di®erentiation of patients from healthy persons. 144

Electroconvulsive therapy
Bilateral electroconvulsive therapy administered to MD patients resulted in reduced regional CBF in the frontal region that increased during the ictal onset and was maintained at that level during the postictal period. 145

Conversation task
PFC activation was reduced in MDD and BD patients, but the continuous activation and brisk°u ctuations could di®erentiate the impairment characteristics. 146

Digit span task
A study on posttraumatic stress disorder patients revealed activations during the retention phase and deactivations during the forward or backward recall phases, thereby illustrating the inhibition in the PFC. 147

Visuospatial task
In a comparative study on AD and late-life depression patients, the AD patients showed higher activation in the parietal cortex during the Benton Judgment of Line Orientation task. 121

Stop-signal task
The reaction time in BD patients was inversely associated with their HbO responses in the right inferior frontal gyrus. 148

Image recall task
Compared to healthy persons, the patients with MDD showed lower HbO responses in the bilateral PFC during unpleasant image recalls, and the HbO response in the left PFC was inversely associated with the depression score. 149

Tooth clenching task
Migraine patients exhibited higher HbR and HbT values in the right PFC compared to healthy individuals, thereby displaying a microvascular oxygen delivery and utilization impairment. 150

Tower of London task
The BD patients revealed signi¯cantly smaller changes in the bilateral dorsolateral PFC compared to healthy persons, indicating impaired planning and problem-solving capabilities. 137

Anxiety and Panic Disorder
Anxiety is often perceived as a healthy emotion and is considered normal unless a person regularly feels inconsistent levels of this emotion, following which it may transform into a medical disorder. This disorder may lead to feelings of fear, worry, and uneasiness. Another associated state of this condition is called panic disorder that is characterized by sudden panic attacks accompanied by perspiration, wobbling, and dyspnea. 151 Due to the portability of the fNIRS system, extensive research on patients with anxiety, panic disorder, stress, and many other types of mental health disorders has been performed and is currently underway. Here, we brie°y review the fNIRS studies on patients with anxiety and panic disorders. Figure 5 shows the task-wise distribution of the studies, and Table 5 summarizes all fNIRS studies on patients with anxiety/fear and panic disorders.

Verbal°uency task
In a study involving a word-°uency cognitive task, the left inferior frontal lobe was signi¯cantly less activated (HbO) in patients with panic disorder when compared to healthy persons. 152 This pilot study suggests that there is a dysfunction in the left frontal lobe of patients with panic disorder. Another subsequent study on these patients reported that the occurrence of panic attacks was signi¯cantly related to HbO changes in the left inferior PFC while the severity of symptoms was associated with the HbR changes in the right PFC. 153 A later study conducted with the same protocol outlines the relationship between frontal lobe function and the catechol-O-methyltransferase (COMT) genotype. 154 This study reported that the increase in HbO levels in the right lateral PFC is associated with the COMT gene of patients with panic disorder. Two studies were conducted using repetitive transcranial magnetic stimulation (rTMS) along with cognitive and additional emotional Stroop tasks, and the results associated with PFC activations/deactivations were compared with those of healthy persons. 155,156 At the baseline (without rTMS), the fNIRS measurements associated with the VFT revealed hypofrontality in the dorsolateral PFC, in panic disorder patients, which signi¯cantly di®ered from the activations observed in healthy persons. However, after sham rTMS, a signi¯cant increase in activation was reported in the left inferior frontal gyrus. While performing the VFT, patients with social anxiety disorder (SAD) showed smaller changes in their HbO responses in the ventrolateral PFC as compared to healthy persons. 157 In another fNIRS study, hyperactivity was also reported in the left frontal area of SAD patients compared to healthy persons. 158

Visual task
When emotional or fearful facial expressions were displayed as stimuli, women exhibited increased HbO responses in the right ventrolateral PFC compared to men. 159 Fearful stimuli were presented to healthy persons and pregnant women in another fNIRS study that revealed signi¯cant activation relative to the resting state in both groups. 160 However, in the group consisting of pregnant women, greater PFC activation was reported during the second trimester compared to during the third trimester, which was related to anxiety. Another interesting fNIRS study was conducted on patients with dental phobia and healthy persons. 161 Compared to the healthy persons, the patients showed an increased HbO response in the supplementary motor cortex while listening to the sound of dental drilling; however, comparable activation was exhibited in a neutral condition. The e®ects of a genetic variant of the neuropeptide S receptor gene (NPSR1) combined with fear-relevant stimuli were assessed using fNIRS. 162 Activations in the dorsolateral and medial PFCs were increased in response to the NPSR1 gene accompanied by fear-speci¯c stimuli.

Walking task
Relative to healthy persons, participants with fear of fall (FOF) exhibited reduced HbO activation in the PFC from the¯rst to the second trial while performing a dual-task walk. 163 No signi¯cant differences in PFC activation were reported in both the FOF patients and healthy persons while performing repeated single-task walks.

Social-cognitive task
The patients with borderline personality disorder (BPD) were compared with healthy persons in an fNIRS study during a social-cognitive task (playing of cards) in the presence of two associates. 164 During the task, BPD patients displayed left medial PFC hyperactivation that most likely resulted from an abnormality in the frontolimbic circuitry.

Cognitive-behavior treatment
In an fNIRS study, interventions, including cognitive-behavioral treatment and acceptance-based behavioral treatment, were administered to the participants with public-speaking anxiety. 165 Individuals treated with the latter treatment showed a decrease in the blood volume in the left dorsolateral PFC in comparison to those treated with the former treatment.

Resting state
Strong RSFC and interhemispheric correlation were observed in the orbitofrontal cortex of heroin users relative to healthy persons. 166 Small-world network properties, which correlate with the predictors of the risk of developing psychopathology in young children, were also calculated in this study. 167

Exposure therapy
In an fNIRS study conducted on patients with acrophobia, during the¯rst exposure therapy, the decreased HbO concentration changes were observed in the dorsolateral and medial PFCs; however, this activation improved towards normal levels over two more sessions. 168

Acupuncture therapy
The altered PFC HbO changes suggested a positive e®ect of acupuncture on decreasing the anxiety levels of anxiety patients. 169 6.9. Mental arithmetic task Arithmetic tasks were performed by participants with low and high levels of anxiety traits in stress and experimental conditions. 170 Overall, while performing the stress arithmetic task, reduced PFC activity was reported in participants with high levels of anxiety traits compared to those with low anxiety levels.

Schizophrenia
SZ is a disease due to which patients appear to stray from reality. It has e®ects on the thinking, feeling, and behavior of the patient. SZ patients usually create supernatural beliefs, su®er from hallucinations, live in delusions, report hearing nonexistent sounds, have cognitive impairment, and/or experience limited motivation. The symptoms of this disease typically start occurring at a young age and do not often develop in children. The causes of SZ are still not clear; however, it is linked with genetic factors, an imbalance in neurotransmitter levels, or tense relationships. Therefore, to treat SZ patients, symptom management is employed via medication or psychiatric counseling. The fNIRS has been used in various settings to reveal the impairing processes in an SZ brain. The task-wise distribution of SZ papers is shown in Fig. 6, and the corresponding studies are outlined in Table 6.

Verbal°uency task
In the earliest¯ndings on SZ via fNIRS, HbO activation was reduced in patients compared to that in healthy individuals. Among patients, typically medicated persons exhibited even lower excitation levels compared to atypically medicated patients. 171 In healthy individuals, the HbO response was higher during letter VFTs compared to semantic VFTs, whereas in SZ patients, higher activation in the PFC was observed during semantic VFTs compared to letter VFTs. 172 The patients with SZ showed lower activations in the frontopolar regions compared to healthy persons, which was associated with poor scores in psychiatric and social evaluations. 173 Genetic polymorphisms were explored in SZ patients, and genotypes associated with poor cortical activations were identi¯ed in several studies. 174-177 Impaired social functioning due to divergent thinking was linked with the ventral region of the frontopolar area in these patients. 178 Further, SZ patients exhibited a decrease in activation in line with an increase in disease severity. 179 The reduced HbO activation response had lower variations compared to those of healthy individuals. 180 In multiple studies conducted on SZ patients along with other depressive patients, the hemodynamic responses di®ered, thereby allowing for the di®erentiation of SZ patients and their levels of depression. 28,113,118,133,[181][182][183] Clinically stable SZ patients exhibited a correlation between the activation in the right ventrolateral prefrontal and temporal areas and the cognitive insight, and that between the activation in the frontopolar, left ventrolateral, and bilateral dorsolateral prefrontal areas and their subjective well-being. 184,185 Studying SZ patients revealed that their impaired thinking was associated with abnormal activation patterns in the left ventrolateral prefrontal area. 186 The early detection and treatment of symptoms are critical as patients with SZ who were untreated for psychosis for more than six months exhibited worse cortical activations in the frontotemporal regions compared to patients who were untreated for less than six months. 187,188 The Chinese speaking SZ patients also showed lower hemodynamic responses in the PFC and superior temporal regions compared to healthy individuals. 189 SZ patients showed lower activation as well as lower functional connectivity in the prefrontal and temporal regions compared to healthy individuals, thereby revealing impaired neural connections. 190 Via a principal component analysis (PCA)-based feature selection and SVMbased classi¯cation, the HbO signal was utilized to di®erentiate the SZ patients from healthy individuals. 191 Older SZ patients showed lower cortical activation compared to younger patients, which is similar in the case of healthy individuals. 192 The reduced HbO response in SZ patients compared to healthy individuals was associated with their selfreported social abilities. 193 In a multimodal study that utilized fNIRS and fMRI, the association between hemodynamic activation and gray matter volume in the left pars triangularis was linked with the onset of SZ. 194 Using positive and negative syndrome scales, the level of impaired activation in the frontotemporal region of SZ patients was associated with their level of depression. 195 In SZ patients, the cognitive ability involved in performing routine tasks was linked with activation in the  dorsolateral PFC and the frontopolar cortex. 196 A study on the association between family history and SZ found that patients with a family history of SZ exhibited an even lower hemodynamic response compared to patients without a family history of SZ, thereby revealing the e®ects of genetics on this condition. 129 Patients with SZ showed a posttask increase in HbO levels, revealing the impairment in their working memory. 197 The rehabilitation of the impairments of SZ patients did not result in any signi¯cant behavioral or neuronal activity after four weeks of therapy, emphasizing that their rehabilitation requires a longer period. 198 Rehabilitating SZ patients by administering transcranial direct current stimulation (tDCS) resulted in improved symptoms associated with positive and negative psychoses. 199

N-back task
The localization of activation in SZ patients was di®erent from that in healthy individuals, and in the patient population, no changes were observed when the load of working memory tasks was increased. 200 In a longitudinal study on SZ patients, positive e®ects were observed in response to the neuropsychological educational approach to cognitive remediation because the cortical activation was improved bilaterally. 201 In SZ patients, the right dorsolateral and bilateral PFCs and the right frontopolar region collectively showed a relationship with the impaired cognitive ability measured via a brief assessment on cognition in SZ. 202 The lateral PFC HbO response in SZ patients was directly associated with theory-of-mind scores. 203

Random number generation task
During a random number generation task, compared to healthy individuals, SZ patients showed signi¯cantly lower activation based on the HbO, HbR, and HbT levels. 204 Overall, the location of HbO activation in SZ patients was similar to that in healthy persons, and patients who developed SZ at a younger age showed high activation impairment in the right dorsolateral PFC. 205

Resting state
The resting state HbT levels in SZ patients were lower than those in healthy persons, and they were also associated with the age of disease onset. 206 The spontaneous activation levels in the medial PFC during resting state were reduced in SZ patients when compared to healthy persons. 207

Multiple cognitive tasks
Among several cognitive tasks, the verbal°uency and Tower of Hanoi tasks resulted in signi¯cant di®erences in the HbO responses, thereby facilitating good classi¯cation accuracy between healthy individuals and SZ patients. 208,209 7.6. Stop-signal task The SZ patients di®ered from the patients a®ected by methamphetamine-associated psychosis as they exhibited better activation responses in the frontopolar area and distinct activation in the premotor region, which is related to impulsivity. 210 The impaired inferior frontal region in SZ patients was responsible for the de¯ciency in the inhibitory control mechanism, whereas the superior temporal region di®erentiated SZ patients from the BD patients. 148

Continuous performance task
The healthy persons exhibited right hemispheric lateralization while the patients with SZ did not show any lateralization during activation, which was possibly due to their left-hemispheric impairment. 211 7.8. Delayed response task SZ patients exhibited bilateral activation due to compensatory reorganization as activation was observed only on the right side in healthy persons. 212

Tower of London task
The HbO and HbR responses in the PFC of SZ patients during a planning task were reduced. 213

Go/No-go task
The SZ patients with excitement symptoms exhibited an impaired HbO pattern during a response inhibition task. 214

Electroconvulsive therapy
The patients with SZ exhibited asymmetric changes in HbO responses in the bilateral PFC after therapy, which di®erentiated SZ and MD patients. 145

Conversation task
During a face-to-face conversation task, the bilateral temporal regions and the right inferior frontal gyrus were responsible for disorganized thinking, owing to which SZ patients face di±culties in conversations. 215 7.13. Language comprehension SZ patients exhibited a de¯ciency in understanding complex language as they displayed incomplete and delayed comprehension, which leads to impaired activation patterns. 216

Stroop
The severity of subclinical psychosis in SZ patients was inversely related to the activations in the dorsolateral PFC and middle temporal gyrus. 217

Video game task
Rehabilitation using interactive sports video games resulted in positive e®ects on SZ patients as their HbO response in the PFC was improved along with their quality of life. 218

Drawing task
The analysis of HbO signals showed that the activation in SZ patients during a tree-drawing task was lower than that in healthy persons. 219

Rock-paper-scissor task
HbO responses were impaired when a patient lost in this task, yet they were associated with scores on the Global Assessment of Functioning and the Negative Syndrome scales. 220

Mild Cognitive Impairment
Mild cognitive impairment (MCI) is a state that causes simple/small problems associated with human memory or thinking. MCI patients do not normally incur any alarming situations that interfere with their routine lives; yet their cognitive standing is low based on memory or thinking when compared to that of age-matched healthy individuals. This impairment is not classi¯ed as dementia but could be a high-risk situation for developing any kind of dementia. The underlying mechanism that prevents MCI from transforming into dementia is still unclear. Therefore, detecting the condition when it is at an earlier stage is important. Several fNIRS studies have been conducted to investigate the physiology of MCI patients. The task-wise distribution is shown in Fig. 7, and the key related studies are outlined in Table 7.

Verbal°uency task
In an early study that examined MCI and AD patients, the overall HbO response of MCI patients was between those of healthy persons and AD patients, but the right parietal area in MCI patients exhibited the most degradation. 221 During a dualtask that involved walking during the VFT, the PFC activation was increased when compared to during simple walking, and this increased activation was directly associated with cognitive ability. 222   The activation was distributed in the bilateral PFC compared to the concentrated activation in the left PFC in healthy persons, illustrating that the impairment in the left hemisphere of MCI patients was being compensated by their right hemisphere. 223 Compared to the HbO responses of healthy persons and AD patients, MCI patients exhibited a steeper slope during activation in the right PFC owing to the hyperactivation process. 224 The HbO response was reduced in the inferior frontotemporal cortex in MCI patients compared to that in healthy persons. 225

Resting state
Compared to healthy persons, from the amnestic MCI patients, reduced tissue oxygen saturation was found in the bilateral temporal-parietal cortex. 226 In a multimodal study that used color-coded duplex ultrasonography, fNIRS, and fMRI, the amnestic MCI patients showed neurovascular decoupling. 227 The tissue oxygenation index (TOI) that was computed using HbO and HbT levels that were derived via fNIRS provided e®ective results that can be considered as a biomarker for amnestic MCI patients as compared to the already established biomarkers obtained via transcranial Doppler sonography. 228 An entropy-based analysis revealed that the complexity of brain signals in amnestic MCI patients was higher than AD patients but was lower than healthy persons, and this reduction in complexity was associated with the clinical scores. 229 During the resting state, fewer low-frequency oscillations in the PFC were observed in MCI patients compared to healthy young persons, while in the parietal cortex, the number of oscillations was low when compared to that observed in healthy older persons. 230

N-back task
Compared to healthy persons, MCI patients exhibited a reduced HbO response in the left dorsolateral PFC, right supplementary motor area, and left superior temporal regions. 231 The working memory activations of MCI patients were comparable with those of healthy persons during low-load tasks: However, they degraded when the load was increased, thereby exhibiting an impaired working memory capability. 232 In the case of MCI patients, working memory training resulted in an improved behavioral performance, but such corresponding improvement was not observed in the PFC activation. 233

Hypercapnia
In amnestic MCI patients and healthy persons, no di®erences were found in vasomotor reactivity before, during, and after inhaling CO 2 . 234

Delayed recall task
The activation levels were similar in healthy individuals and MCI patients in response to the phase of memorizing words in this task, but compared to healthy persons, the MCI patients showed a reduced HbO response approximately in Brodmann area (BA) 9 during the retrieval of words phase. 235

Digit span task
The mean and slope of the HbO responses were correlated with the clinical scores, and the scores of the MCI patients stayed below those of the mild and moderate-to-severe AD patients. 236

Alzheimer's Disease
It is a form of dementia with the most rapidly increasing prevalence rate. Patients with AD can only be treated to manage their symptoms as there is still no known cure for this disease. It slowly sabotages the memory and takes away the capacity to do routine tasks. Various tasks have been utilized to get insights into the brain activity of these patients. The task-wise distribution of the studies is presented in Fig. 8, and Table 8 outlines the details of the works.

Verbal°uency task
The VFT is the most commonly used task in the studies associated with AD. The earlier¯ndings revealed that AD patients had reduced HbO and HbT levels in the parietal areas. 237,238 The HbO levels in the prefrontal area decreased in some patients 237 while they increased in most others. 238 These contradicting¯ndings were attributed to variability caused due to subject characteristics or the location of the fNIRS channel. A better performance in the VFT is associated with the left prefrontal hemisphere, but this physiological asymmetry is missing in AD patients. 239 AD can be di®erentiated from MCI by revealing degraded global activation when measurements of most of the brain areas are taken simultaneously. 221 In the pursuit to enhance the quality of life of AD patients by improving their symptoms, the administration of an oral drug called memantine was bene¯cial when compared to not using this drug. 240 The activation was slightly higher compared to that of patients struggling with late-life depression. 121 The patients medicated with a cholinesterase inhibitor showed improved activation in the speech-related areas of the brain as a higher concentration of HbO was measured. 241 The activation region in AD patients is di®erent compared to the patients with frontotemporal dementia, as AD patients exhibited activation in the frontoparietal areas. 242 The mean activation pattern of AD patients was lower and slower than those of MCI patients. 224

Benton judgment of line orientation test
The parietal cortex is linked with visuospatial tasks, and analyses of activation occurring in this region can be used for early detection of AD since AD patients exhibit only marginal activation when compared to the explicit activation in healthy subjects. 243 The parietal region showed considerably higher HbO activations in AD patients compared to depression patients. 121

Clock drawing test
The clock drawing test (CDT) scores can adequately be used to di®erentiate between healthy individuals and AD patients; however, the entropy analysis conducted on fNIRS recordings while the subjects were performing the CDT resulted in signi¯cant di®erences between the results of AD patients and healthy subjects. 244

Digit span test
An entropy-based fNIRS signal complexity analysis demonstrated that the digit span task (DST) could help classify AD. 244 A time-series analysis revealed that the reduction or decline in HbO levels becomes steeper when the intensity of AD progresses. 236

Corsi block-tapping test
Compared to the CDT and DST results, the results of the Corsi block-tapping test were the most effective in di®erentiating AD patients from healthy subjects via entropy analysis. 244 9.6. Driving task AD patients exhibited lower HbO values than those of healthy individuals, and the act of applying the brake during the task and HbO changes were negatively related while these were positively related in healthy persons. 245

Shiritori tasks
While performing Shiritori tasks, the area and maximum value of the fNIRS signals from the dorsolateral PFC and frontal pole cortex regions of AD patients were signi¯cantly lower. 246

Olfactory task
An interesting study conducted on AD patients using an active vanilla smell and a sham one revealed that brain activation occurs in the temporal region of healthy individuals while performing the olfactory task whereas the AD patients did not show any activation at all. 247

Free and cued selective reminding test
During the delayed free recall phase of the task, higher entropy values were observed in AD patients compared to those in healthy participants in BAs 9 and 46. 248

Rey Auditory Verbal Learning Test
AD patients exhibit higher levels of tissue oxygen saturation in the frontal cortex during the Rey Auditory Verbal Learning Test after receiving brain reperfusion rehabilitation therapy. 249

Resting state
An entropy analysis of the fNIRS signals from all the brain areas revealed that the signal complexity in the brain networks of AD patients was reduced compared to those of healthy individuals as well as MCI patients. 229

Parkinson's Disease
The early signs of PD are tremors in the hands that a®ect movement and balance. Due to a reduced sense of coordination, people with PD often drop items and are more likely to fall. Further, the posture of their bodies are slightly altered. The¯rst problem that PD patients and their caregivers face is disorder in their gait and balance. Neuroimaging techniques are now able to provide more insights into the neural mechanisms of the pathophysiology associated with the gait disorders in PD patients that can cause freezing of gait (FOG). In this section, a few older fNIRS studies and recent investigations are reviewed. The task-wise distribution of studies on PD patients is presented in Fig. 9, while Table 9 outlines all the fNIRS studies.

Deep brain stimulation
We found two fNIRS studies conducted between 1999 and 2000 which investigated the cortical changes in the frontal area in the brain of PD patients by invasively (deep brain) stimulating the thalamic nucleus ventralis intermedius (VIM) and globus pallidus internus (GPi). 250,251 At di®erent frequency ranges, various patterns of cerebral blood oxygenation were observed. Stimulating the GPi at higher frequencies resulted in an increase in HbO and a decrease in HbR. In contrast, in the VIM, the cerebral oxygenation changes were opposite to those seen via GPi stimulation. Another pilot study on PD patients was conducted to examine the motor associated cortical activity changes in response to deep brain stimulation (DBS). 252 Compared to prestimulation, after DBS, the cortical activity was higher in the PFC of PD patients. This indicates the therapeutic bene¯ts of DBS in patients with PD.

Walking tasks without training
A pilot study on PD patients who were a®ected by FOG demonstrated the feasibility of an fNIRS assessment of the locomotor task during real-life conditions. 253 During turns, in PD patients, this study reported an increase in HbO activation in the frontal lobe before and while experiencing FOG while no changes in HbO activation were observed in healthy persons. Another comparative study on PD patients showed di®erent activation patterns in the frontal lobe during complex walking tasks and concluded that the activation in PD patients depends on the nature of the task. 254 During normal walking and obstacle avoidance, the PD patients showed an increase in HbO levels while in healthy persons, no activation was observed during a dual   255 The decrease in the activation in BA 10 was observed while the patients were turning while an increase was observed while they were walking. Comparing groups of PD patients with worse and better ambulations revealed that a decrease in prefrontal activation was observed in the latter group during turning.

Walking tasks with training
In a randomized controlled trial conducted with fNIRS, the e®ects of treadmill training in a virtual reality environment on prefrontal activation in PD patients during normal, dual-task, and obstaclenegotiation walking were studied. 256 Decreased prefrontal activation was observed after gait training, thereby indicating an improvement in walking. It indicates that PD patients exhibit less reliance on cognitive resources during normal walking. Thesē ndings were further supported by the researchers' recent study in which a decrease in HbO levels was observed while the patients were walking on a treadmill compared to while they were walking on the ground. 257 Improvements in gait were also reported. Another comparative fNIRS study was conducted on the motor cortex and PFC of PD patients and healthy persons while they were walking on a treadmill at a user-de¯ned speed and an experimenter-de¯ned faster speed. 258 The increase in HbO responses of the PD patients was higher in the left and right motor cortices while walking in both conditions compared to those of the healthy persons.

Postural control task
In an fNIRS study, compared to healthy persons, PD patients showed signi¯cantly increased prefrontal activation while maintaining postural stability. 259 However, patients with mild PD demonstrated a similar activation pattern to healthy persons.

Iowa gambling task
The performances in the Iowa gambling task were assessed using fNIRS to establish the relationship between personality traits and prefrontal activity in PD patients who were pathological gamblers and those who were not. 260 The patients with active gambling behavior showed signi¯cantly increased activity in the dorsolateral PFC in response to highrisk and more rewarding options, which indicates a notable involvement in the frontal area in both emotional and cognitive processes.

Stroke
Stroke is a disease due to which a patient's brain does not receive su±cient blood based on its requirements. This condition occurs due to problems in the arteries that are responsible for the supply of blood to the brain. Stroke is broadly classi¯ed into two types: Ischemic stroke occurs when the blood supply is reduced or blocked due to clotting, whereas hemorrhagic stroke occurs when the blood vessels burst open. In both cases, the supply of blood is compromised in a part of the brain, which results in the death of brain cells within a short period. Every brain cell is linked with some function that our brain has to perform; therefore, the dying cells result in the loss of their associated functions. Therefore, it is best to prevent the occurrence of strokes by making changes in our lifestyles that control the cholesterol and fat levels in our body. Due to advanced treatments, the death rates due to strokes have reduced compared to those in the past. There are rehabilitation therapies and drugs that stroke patients can use to regain lost functions. Several researchers have used fNIRS to understand the impairment levels and types in stroke patients using various paradigms. The taskwise distribution of stroke papers is presented in Fig. 10, and these studies have been outlined in Table 10.

Resting state
Owing to the patients' conditions, most studies on stroke patients have been performed using the resting state data. In an fNIRS study, the interhemispheric connectivity of ischemic stroke patients was signi¯cantly di®erent from healthy persons by examining the low-frequency cardiac and respiratory oscillations, thereby proving the e±ciency of this modality. 261 In ischemic and hemorrhage stroke patients, the frontal cerebral oxygenation was directly correlated with the CBF measured via traditional CT perfusion imaging, illustrating the e±cacy of the technique. 262 By examining the HbO signal of symptomatic carotid occlusion and hypoperfusion patients, the interhemispheric amplitude ratio was impaired when compared with that of healthy persons. 263 Rehabilitation via the application of anodal tDCS induced neuronal activity by resulting in changes in the HbO and HbR values in stroke patients. 264 The improvements in the analysis techniques conducted on interhemispheric connectivity were critical in identifying the basis of the physiological di®erences responsible for this condition by eliminating motion artifacts in stroke patients. 265 The frequency-domain system allowed the calculations of the absolute values of HbT and hemoglobin oxygen saturation, which facilitated the identi¯cation of the impaired site in stroke patients. 266 Restless leg syndrome patients with periodic limb movements (PLM) during sleep, who may be at a high risk of developing stroke, showed increased HbO and HbR levels while sleeping when compared to healthy persons. 267 The cerebral autoregulation measured via coupling between HbO levels and average arterial pressure illustrated the impairment in poststroke patients as compared to healthy persons. 268 The optical path length was di®erent due to impairments in ischemic stroke patients, thereby illustrating the change in tissue characteristics. 269 A wireless and mobile fNIRS device facilitated the early detection of stroke symptoms by revealing reduced cerebral oxygenation in the a®ected hemisphere, as measured conventionally using perfusion computed tomography and perfusion-weighted magnetic resonance imaging. 270 The e®ective connectivity in multiple frequency bands detected by examining HbO signals was reduced or diminished in patients with cerebral infarction as compared to healthy persons. 271 The time-domain system was able to measure signi¯cant di®erences in HbO and HbR values in large vessel occlusion stroke patients as compared to healthy persons, and this di®erence was correlated with the impairment condition. 272

Walking task
Walking on a treadmill with body weight support resulted in a higher HbO response in the sensorimotor region of stroke patients. 273 During the motor rehabilitation of stroke patients, instead of a simple walking task, a dual-task that involved walking while counting backward was an e±cient technique as it resulted in a better HbO response. 274 While comparing the e®ects of dual-task walking to those of cognitive or motor tasks, both were e®ective for cortical activation, but they attenuated the gait performance in poststroke patients. 275 The poststroke patients exhibited hyperactivation in the PFC during a dual-task involving walking along with a cognitive task, and the HbO levels may become saturated while walking over obstacles demonstrating the full utilization of resources. 276 A dual-task involving walking during calculation revealed that HbO activation in the PFC was linked with physical performance in stroke patients while it was linked with cognitive performance in healthy persons, thereby revealing a di®erent prioritization trend between patients and healthy persons. 277

Hand/¯nger movement task
Electromyography-triggered functional electrical stimulation accompanied by voluntary movements of¯ngers and wrists resulted in better activation as compared to that of voluntary movements or of electric stimulation individually. 278 The cortical activation in the precuneus region in stroke patients was linked to mirror therapy, and it could be used to determine the e±cacy of the therapy. 279 In a longitudinal study, revascularization surgeries performed on stroke patients resulted in improvements as compared to the levels before surgery, the CBF   280 The right-handed stroke patients with an impairment in the right hemisphere and vice versa exhibited better motor recovery owing to a combination therapy of lowfrequency repetitive transcranial magnetic stimulation and intensive occupational therapy in the left (una®ected) hemisphere. 281

Cycling task
The rehabilitation of stroke patients via cycling resulted in better cortical activation in the premotor cortex and in better physical performance in response to providing them feedback on their speed. 282 Comparing the electrical stimulation intensity during the rehabilitation of stroke patients while they were performing the cycling task, an intensity of 10 mA resulted in better cortical excitations compared to a higher intensity of 30 mA. 283

Robot-assisted elbow movement
In poststroke patients, the combination of robotassisted rehabilitation therapy and botulinum toxin A injections was e®ective as the HbO response in the primary sensorimotor region was improved when examined after two weeks and again after four months. 284 The task-related cortical activity was signi¯cantly improved on providing biofeedback to the subjects as the robot changed the color of light based on the patient's performance. 285 11.6. Balancing task During a balancing task, stroke patients exhibited cortical activation in the bilateral prefrontal, premotor, and parietal regions similar to that of healthy persons, thereby illustrating no functional reorganization in the brain, yet the activation was smaller in the a®ected areas of the brain. 286

Oxygen inhalation task
The patients with cerebral ischemia showed reduced weights via PCA of the HbO responses in impaired regions as compared to those in normal regions. 287

Lower limb movement task
The HbT responses utilized in a linear discriminant analysis (LDA) revealed a signi¯cant discrimination in the movements between the paretic and nonparetic limbs of stroke patients. 288

Swallowing task
Compared to the observed responses in healthy persons, the stroke patients showed prolonged HbO and HbR responses while actively swallowing saliva or imagining swallow. 289 11.10. Word repetition task During a language task, stroke patients received repetitive transcranial magnetic stimulations on the opposite hemisphere from the activated hemisphere, and poststroke patients received intensive speech therapy to improve the cortical excitations and language function. 290

Design and verbal°uency task
During two case studies on stroke patients, due to visuospatial and language functions, marginal cortical activations were exhibited only in the una®ected hemisphere. 291

Tilt-table task
The poststroke patients with right-lateralized PFC activation at rest exhibited increased HbO levels in the PFC during a tilting task whereas the patients with left-lateralized HbO responses exhibited a decrease in HbO levels during this task. 292

Traumatic Brain Injury
Traumatic brain injury (TBI) results from accidents that subject the brain to sudden damage due to an injury to the head. The most common causes of TBI are tra±c accidents, falls, and sports injuries. A TBI patient can su®er from a wide range of physiological and psychological symptoms based on the a®ected location of the brain, and the impairments can last for short, long, or even life-long periods. The symptoms can appear immediately, or in some cases, they may appear after some days or weeks. The treatments for TBI involve rest, medication, and/or surgery in some cases. The fNIRS has been used to study the hemodynamic responses associated with the various types of symptoms caused by TBI. Figure 11 presents the task-wise distribution of the works, and Table 11 outlines them.

Hand/¯nger movement task
The patients with TBI showed lower cerebral oxygenation, but a similar blood volume in the left PFC during a right-hand gripping task when compared to healthy persons. 293 Children with concussion exhibited a reduced HbT and HbO coherence exhibiting impaired interhemispheric connectivity when compared to healthy children during a¯ngertapping task. 294 Compared to healthy persons, the patients with mild TBI also showed lower functional connectivity that was inversely linked with impairment intensity, and the di®erence in connectivity was more pronounced during the task period as compared to the resting state. 295

Visual task
The TBI patients exhibited a reduced HbO response in the bilateral dorsolateral PFC during an attention task when compared to healthy persons, thereby revealing the impaired intentional networks. 296 The patients with sports-related concussions showed a higher hemodynamic response in the frontal regions and a strong interhemispheric correlation in the occipital cortex when compared to healthy persons. 297

Cognitive rehabilitation
While undergoing training involving nine cognitive tasks, the TBI patients showed similar HbO responses in the lateral frontal regions and a higher HbO response in the medial frontal regions when compared to healthy persons. 298

Neurocognitive test battery
The patients with sport-related concussions showed reduced cortical activations in the a®ected areas during a computerized test involving various working memory tasks compared to healthy persons. 299

N-back task
During a working memory task, the analysis of HbO, HbR, and HbT responses in TBI patients revealed signi¯cant di®erences compared to those of healthy persons even though the behavioral performance was similar. 300

Paced auditory serial addition test
The cortical activation regions were di®erent in TBI patients compared to healthy persons during the task with or without distraction, thereby revealing poor inhibitory control. 301

Complexity judgment task
The oxygenation variability index measured via the HbO and HbR values resulted in high sensitivity in di®erentiating TBI patients from healthy persons during various levels of complexity judgment. 302

Stroop task
An increase in neural activation was observed in healthy persons by increasing the cognitive demand while the TBI patients achieved higher activations while performing more straightforward tasks, thereby revealing the impaired frontal lobe e±ciency. 303

Music listening task
A vector phase analysis conducted during music identi¯cation with or without distraction revealed Fig. 11. Task-based distribution of studies on traumatic brain injury (total studies: 13).

Spinal cord stimulation
The patients with disorders in consciousness due to TBI exhibited that a shorter interstimulus interval of spinal cord stimulation resulted in higher HbT levels in the PFC, implying a higher level of awareness of the patients. 305 13. Discussion and Future Implications In this paper, we summarized the studies conducted on notable diseases using fNIRS as a neuroimaging tool. Such notable diseases were examined/included only when we could¯nd more than 10 studies involving a patient population.

Preprocessing of fNIRS signals
Compared to fMRI, a well-established modality, fNIRS, is still a growing modality for understanding neuronal activities. The methods adopted to examine hemodynamic changes via fNIRS are diverse, and it has been eagerly proposed that a standard procedure should be followed. 306 If studies follow a standard data processing pipeline, they can be compared, and a veri¯able knowledge database can be established. The acquired raw fNIRS data are a®ected by various noise sources like physiological (respiratory, cardiac, Mayer waves, etc.), environmental (ambient light, subject movement, source/ detector attached to the scalp, etc.), and instrumentational ones (sensor noise, communication noise, line noise, etc.). The details of the noises and their properties can be found in the literature. 307 These noises reduce the signal-to-noise ratio of the desired signal, and they can override the neuronal activation for the task performed following an experimental paradigm not carefully designed. 306,308 Therefore, the removal of these noises to obtain a clean fNIRS signal is a pivotal step. Various techniques are employed to remove them as they are identi¯ed by their approximate frequencies like cardiac (1 Hz), respiratory (0.3 Hz), and Mayer waves (0.1 Hz). Mostly digital¯lters are used to remove these frequency bands from the raw signals. Two types of¯ltering are commonly used: A band-pass¯lter (used to retain a frequency range from the signal while discarding the remaining part) and a low-pass¯lter (used to remove the high-frequency part beyond a certain frequency). Some researchers prefer to use a¯lter on the light-intensity signals while others use on the hemodynamic signals. 306 In the recent past, the use of short-separation channels to remove the extracerebral e®ects is gaining attention. The short-separation channels are con¯gured by making the source-detector separation lower than 10 mm. 309 The NIR light in these channels does not penetrate deep enough and is considered to carry information only from the super¯cial layer. 310 The maximal source-detector separation was found to be 8.4 mm for a typical adult brain, and 2.15 mm was most suitable for an infant brain. 311 If the distance for short-separation channels is not carefully con¯gured, the information from the gray matter is also included. 312 In some research, the information from short-separation channels were included in a regressor to clean the fNIRS signal. But, its utilization globally across the surface of a head is critically argued due to the heterogeneous response of scalp. 313,314 One notable idea is to use two short-separation channels in the regression equation; one at the source side and the other at the detector side. 315 The experiments conducted on patient populations are more critical and yet are more prone to motion artifacts due to patient conditions. 90,265 The removal of motion artifacts from the raw signals is necessary along with physiological noises for further processing. 316 The signal is largely a®ected by motion artifacts in the case of newborns as they are more prone to movement during the experiment resulting in data loss. 317,318 Threshold levels were de¯ned for signal changes to discard motion artifacts due to infant head movements. 319,320 Many research works are being carried out on infants to understand the developing brain. The hemodynamic response of infants has been reported to alter from adults, which can be due to the e®ect of various variables like stimulus complexity and experimental designs. 321 If the duration of an experimental study is long, the patients are more likely to move during the test. The experimental design should also be planned critically to hold the patient's attention while performing the tasks without burdening or boring them, which result in mind wandering-based activations. 322 The duration of the initial baseline, task duration, and the rest period between multiple tasks should be considered carefully as the hemodynamic response is a slow process that takes time to revert to the baseline after activation. 323 The positioning of patients is also an important aspect during the experiment and ensuring that the patients are in similar postures improves the fNIRS data. 229,264,324

Processing of fNIRS signals
There have been di®erent analyses performed on hemodynamic variables. The fNIRS systems have the capability to provide HbO, HbR, and HbT values instead of only HbR values that are acquired via fMRI. Many researchers employ HbO values to conduct their analyses with the justi¯cation that these values have a more direct relationship with cortical activations and can facilitate understanding them better. [18][19][20][21][22]25,121,169,170,263 The fNIRS variables have multiple data embedded in them that can be extracted by various signal processing techniques, which range from various¯ltering adaptivē ltering methods to signal complexity analyses, such as entropy analyses. 228,229,325,326 Researchers have utilized fNIRS variables to extract several biomarkers for the classi¯cation and identi¯cation of diseases, such as low-frequency oscillations, heart rate, CBV, CBF, TOI, Cytox, and cerebral oxygen exchange. 75,230,[261][262][263]267,302 The features of fNIRS variables that consist of, but are not limited to, peak, mean, skewness, variance, slope, kurtosis, standard deviation, number of peaks, sum of peaks, root mean square, and median are frequently utilized in classi¯cation algorithms like LDA, SVM, extreme machine learning, Bayes classi¯ers, and neural networks. 13,327,328 The processing of fNIRS signals is usually done by the user's choice by mostly utilizing the software provided by the device manufacturer without having a deep understanding of the underlying methods. 329 The results are largely a®ected by the choice of procedure employed. The recommended procedure is to use a standardized preprocessing pipeline and do personalized processing to get the required information. In most studies, the authors write their codes/ routines for their own purposes. Instead, to facilitate the processing of fNIRS signals, various tools have been developed. HomER and NIRS-SPM are the most commonly utilized software packages in the fNIRS community that allow device-independent analyses of the signals. 330,331 Other important tools being used are fOSA, NAP, FC-NIRS, NinPy, NIRS brain AnalyzIR, ICNNA, and GRETNA, which have allowed fNIRS practitioners to explore many aspects of brain development, behavior, and pathologies. [332][333][334][335][336][337][338] The utilization of advanced signal processing and adaptive control algorithms in the future can be helpful to achieve earlier detection of the hemodynamic response not to mention the accuracy. 11,339-343

Channel localization
Patients are usually classi¯ed based on the impaired hemodynamic responses caused due to some disease. [128][129][130][131][132][133] The localization of an impairment in the brain is a vital step to evaluate the intensity and type of the disease. Therefore, the placement of fNIRS sources, detectors, or optodes on a patient's head based on the task involved is important as most tasks are associated with known brain regions. 8,344 Unlike fMRI, fNIRS does not allow for structural imaging, which makes it di±cult to compare studies using di®erent channel con¯gurations and placements. Therefore, a standardized placement system should be followed like the EEG electrode placement such as the 10-20, 10-10, or 10-5 systems so that the¯ndings are comparable and reproducible among studies and subjects. 9,345,346 Another method for standardized locations is to involve the brain's structural information by utilizing an fMRI scanner to select the locations for fNIRS channels initially. 347 The involvement of an fMRI scanner diminishes the advantages of utilizing fNIRS for neuroimaging and adds extra burden to the subject under study. 348 A useful approach for identifying the channel locations that are similar to fMRI using the Montreal Neurological Institute coordinate system is the utilization of a 3D digitizer to cast the fNIRS channels to a brain atlas. 349 A recently developed toolbox recommends the placement of optodes based on the desired location of the brain. 350 Most studies reviewed in this paper utilized an EEG location system, while the utilization of fMRI and a digitizer together was rare. 98,103,107,212 The most probable reason for this is the involvement of the extra cost of equipment and the extra time consumption for the subject as well as for the experiment conductor. Forming a channel for fNIRS is also associated with the distance between the light source and photodetector. The achieved depth of an fNIRS channel can vary with variations in the source-detector separation. While most of the studies measured the fNIRS signal with a 3-cm separation, variations of 2-5 cm were observed. 96,97,109 The headgears/caps are also provided by the system manufacturers, therefore, in most cases, these headgears/caps are used to place and hold the optodes at a¯xed distance of 3 cm. [28][29][30][31][32][33][34]121,[194][195][196][197][198][199]224,241,242 It is recommended to use a channel separation of 2 cm for children and infants due to smaller head size and the resulting reduced width between super¯cial layers. 23,42,167,294

Channel resolution and limitation
The limited number of optodes available on fNIRS machines and the resulting number of channels are still not su±cient to study the changes in the brain. Few studies could examine most of the brain regions. 85,91,104,108,229 Neuronal activations occurring in response to a single task are not linked to a single brain location. 351 Therefore, the outcomes of the studies that focus on a speci¯c narrow location in the brain might not be su±cient for understanding the brain functions. Brain functional reorganization happens when a certain portion of the brain is impaired. 98,286 Therefore, studying multiple brain areas is important to understand the underlying changes occurring in patients' brains. [30][31][32][33][34]79,121,236,266,272 Functional and e®ective connectivity analyses are useful to understand the processing that takes place in the human brain, and covering more brain areas will allow for better understanding based on the experimental conditions. 4,271,[295][296][297]352 The bundled optode approach that creates hundreds of channels at di®erent brain depths with high spatial resolution can assuredly assist in the creation of 3D images via fNIRS, which can be compared with those created via fMRI. 7 This technique involves spatially resolved spectroscopy that utilizes multidistance channel formation among groups of sources and detectors placed as close to each other as possible. 353 The resultant number of channels will be in the thousands compared to the maximum of a few hundred that is possible using the present state-of-the-art systems. This technique will open the gates to machine learning-based signal processing algorithms that are commonly used for large datasets like neural networks and many others. 354,355 It will facilitate the quick expansion of the research scope via fNIRS, which will help in further revealing the currently hidden patterns and properties of hemodynamics. Although physical constraints do not allow fNIRS to penetrate beyond a speci¯c depth, however, it can be used as an alternative technology for limited brain depth imaging. Therefore, new fNIRS systems with an extensive number of optodes should be developed to assist in examining the full brain with high spatial and temporal resolutions. Most of the available commercial device manufacturers do not allow the con¯guration of short-separation channels in their¯xed optode holders. 329 The integration of short-separation channels by manufacturers is recommended as it is one of the important methods for getting a clean fNIRS signal.

The fNIRS-based brain-computer interface
The utilization of fNIRS in diagnosing and classi¯cation of various diseases has been established and covered in this paper. The importance of BCI in the healthcare industry is critical, especially for aggravated conditioned patients with physical disabilities. fNIRS has been used as a neuroimaging modality in developing BCIs due to its various bene¯ts. 356 As the hemodynamics signal has inherently slow nature, the light intensity signals of fNIRS known as fast optical signals were explored in comparison to the event-related potentials of EEG. 357 The quality and reliability of fast optical signals are still very low, and further research with stable results is required for practical applications. The classi¯cation of fNIRS signals has been a challenging task and normally averaged samples are used for classi¯cation, yet researchers have showed promising results for single-trial as well as online classi¯cation of VFT, Stroop task, and resting state. [358][359][360][361] Promising research has been done to reduce the time delay in fNIRS activity detection. 12,362 The BCI usually involves imagining of tasks to generate brain activity, which can be utilized for command generation. Imagining \yes" or \no" can be a very basic imagining task that could be suitable for a wide range of patients and the classi¯cation accuracy for this task was reported to be signi¯cant for most users. 363 Motor imagery is popularly used in BCIs as imagination of movement of limbs produces reliable hemodynamic response, which is comparable to the hemodynamic response of actual limb movement. 364 The neuronal activity generation due to motor imagery was found to be enhanced with visual feedback during a robot movement control using motor imagery. 365 The acquisition of data from a realistic environment and its processing are vital for the development of BCI. 366,367 Research on the mental states and neuroergonomics of pilots during actual as well as simulated environment was conducted using fNIRS. 368,369 Satisfactory results from these studies strengthened the use of fNIRS in BCI applications. The understanding of encoding and decoding of neuronal activation is important in developing BCI applications to complement the neural encoding and decoding being used in medical robotics research which use nerve-machine interfaces. [370][371][372][373][374] 13.6. Tasks for fNIRS signals Various tasks have been used to understand the de¯cits related to the PFC. The covered studies in this paper reveal that the VFT was the most widely used task in understanding the impaired activation of the diseased population. The VFT has demonstrated e±cient performance levels in distinguishing healthy participants from a mixed population of patients and healthy persons. 96,191,208,209 The cortical activations in response to VFTs in a hybrid population comprising of patients a®ected by multiple diseases were used to di®erentiate patients of one disease from another as well as to generate information of the disease severity. 28,113,116,118,125,179,[181][182][183] Di®erent working memory tasks have been used to perceive the neuronal activations associated with various diseases using multiple task loads. 52,200,232 Various studies involving electric and magnetic stimulations have been conducted and have described the e®ects of stimulation during rehabilitation therapy on hemodynamic responses, thereby demonstrating the ben-e¯cial nature of fNIRS. 9,105,199,[250][251][252]264,290,375,376 The fNIRS captures optical intensity signals via photodetectors, which are unaltered by electric and magnetic¯elds. Therefore, fNIRS is a more suitable neuroimaging modality for evaluating the e®ects of rehabilitation in the brain compared to fMRI and EEG. This paper presents widely utilized tasks associated with each disease that can serve as a guideline for future classi¯cation studies. Also, it suggests new possible directions for research on a speci¯c disease that may have been followed for another disease.

Conclusions
In this paper, we reviewed studies involving patient populations that used fNIRS to examine mental/ physical impairments. The fNIRS is a portable neuroimaging modality that has been extensively employed to evaluate and classify various diseases. By the broad utilization of fNIRS, it is evident that this technology is appropriate to examine neuronal behavior of healthy subjects as well as patients. This paper described brie°y the signi¯cant¯ndings associated with impaired neuronal activations that were speci¯c to tasks and mental disorders. We indicated the tasks that could be used to show sig-ni¯cant cortical activations in diseased populations. Distinct patterns of activation or low-frequency oscillations were associated with speci¯c diseases and were used for classi¯cation. Although the intensity and disease classi¯cations were achieved, yet haphazardness in pre-and post-processing schemes and parameter reporting exists in the literature, which needs to be standardized.
We described the studies that used various channel con¯gurations and signi¯cant variations in the resultant number of channels. Most studies restricted their scope to a single brain area while a few studies covered multiple lobes and rarely examined the full head. The constraint on the available number of optodes in currently available commercial fNIRS systems restricts the number of channels with an intermediate spatial resolution. The dense placement of optodes covering the entire surface of the head will allow for superior spatial resolution up to a limited brain depth. The resulting massive number of channels will require new methodologies for processing big fNIRS data. The frequent application of machine learning algorithms on fNIRS data, which is currently not possible due to limited datasets, will be interesting to observe in the future.