Neural correlates of treatment response to ketamine for treatment-resistant depression: A systematic review of MRI-based studies

Treatment-resistant depression (TRD) is defined as patients diagnosed with depression having a history of failure with different antidepressants with an adequate dosage and treatment duration. The NMDA receptor antagonist ketamine rapidly reduces depressive symptoms in TRD. We examined neural correlates of treatment response to ketamine in TRD through a systematic review of brain magnetic resonance imaging (MRI) studies. A comprehensive search in PubMed was performed using “ketamine AND depression AND magnetic resonance.” The time span for the database queries was “Start date: 2018/01/01; End date: 2024/05/31.” Total 41 original articles comprising 1,396 TRD and 587 healthy controls (HC) were included. Diagnosis of depression was made using the Structured Clinical Interview for DSM Disorders (SCID), the Mini-International Neuropsychiatric Interview (MINI), and/or the clinical assessment by psychiatrists. Patients with affective psychotic disorders were excluded. Most studies applied ketamine [0.5mg/kg racemic ketamine and/or 0.25mg/kg S-ketamine] diluted in 60cc of normal saline via intravenous infusion over 40 minutes one time, four times, or six times spaced 2–3 days apart over 2 weeks. Clinical outcome was defined as either remission, response, and/or percentage changes of depressive symptoms. Brain MRI of the T2*- weighted imaging (resting-state or task performance), arterial spin labeling, diffusion weighted imaging, and T1-weighted imaging were acquired at baseline and mainly 1– 3days after the ketamine administration. Only the study results replicated by ≥ 2 studies and were included


Unmet needs in treatment-resistant depression
However, more than one-third of patients with major depressive disorder do not experience clinically significant improvement even after multiple courses of antidepressants and can thus be regarded as having treatment-resistant depression (TRD) (Rush et al., 2006).The presence of TRD can be defined as nonresponse to one or two antidepressant treatments with an adequate dosage and duration (Gaynes et al., 2020).TRD might correlate with higher than average rates of intentional self-harm and all-cause mortality (Lundberg et al., 2023).Diverse options for treating TRD have been proposed and examined (Marwaha et al., 2023).First, atypical antipsychotics (aripiprazole, extended-release quetiapine, and combined olanzapine-fluoxetine) were approved by the U.S. Food and Drug Administration to treat depression (Jha and Mathew, 2023).Second, repetitive transcranial magnetic stimulation targets and promotes the long-term potentiation of cortical structures such as the dorsolateral prefrontal cortex and modulates deep gray matter structures such as the subcallosal cingulate cortex and proximate white matter pathways (Idlett-Ali et al., 2023).Third, a glutamate-gated N-methyl-d-aspartate receptor antagonist called ketamine, which can be administered per oral or by intravenous infusion, has fast-acting antidepressant and anti-suicidal effects in TRD patients (Lv et al., 2023;Meshkat et al., 2023a).Its relatively fast reduction of depressive symptoms and possible reduction of suicidality make ketamine a promising treatment for TRD (Jollant et al., 2023).
As a matter of fact, both stronger functional activation of medial prefrontal cortex and posterior cingulate cortex as well as the lowered stability in the resting-state intra-network functional connectivity of default mode network in MDD than HC (Wise et al., 2017) could contribute to the ruminative thought in MDD (Perkins et al., 2015).Conversely, acute intravenous infusion of selective serotonin reuptake inhibitor (SSRI) to human healthy controls (HC) not only increases static resting-state functional connectivity between the medial prefrontal cortex versus posterior cingulate cortex and DLPFC, but also reduces the variability of connectivity between medial prefrontal cortex versus precuneus and posterior cingulate cortex (Arnone et al., 2018)..Initial treatment effect of SSRI is to alter the emotional processing of aversive stimuli by way of the dorsal raphe nuclei via inhibitory 5-HT 1A auto-receptors (Selvaraj et al., 2018).Acute intravenous infusion of SSRI in HC increase the BOLD response bilaterally in the amygdala to fearful faces and in the right amygdala to happy faces, which is proportional to the availability of the 5-HT 1A receptor in the dorsal raphe nucleus (Selvaraj et al., 2018).

Aim of this study and study hypothesis
However, although some evidence of its anti-inflammatory (Meshkat et al., 2023b) and neuroplasticity (Medeiros et al., 2022;Wilkowska and Cubała, 2022) effects has been collected, the detailed mechanisms driving ketamine's treatment effects in TRD remain elusive.Also, recent systematic reviews of ketamine-related human brain MRI studies have focused on pre-versus post-comparisons of ketamine administration and have provided little information about the brain-based markers of a treatment response, not only in terms of depressive symptom severity but also for item-level depressive symptoms (Meshkat et al., 2023b;Souza-Marques et al., 2021;Zavaliangos-Petropulu et al., 2023).Of note, structural integrity of white matter tracts connecting the prefrontal cortex and subcortical regions are decreased in MDD compared to HC (Han et al., 2019;Phillips et al., 2015).Medial prefrontal cortex, a pivotal component of the default mode network involved in decision-making, attention, and self-referential processes (Perkins et al., 2015), shares extensive connectivity with cortical and limbic regions involved in affective modulation (Arnone et al., 2018).Increased resting-state functional connectivity of dorsomedial prefrontal cortex with executive control network (having a hub region of DLPFC) (Rogers et al., 2004), default mode network (with hub regions of posterior cingulate cortex and precuneus) (Sheline et al., 2009), and affective network (including subgenual and pregenual ACC) (Price and Drevets, 2010) are reflective of changed neural dynamics of affective regulation in MDD (Sheline et al., 2010).
Therefore, we examined the neural underpinnings of a treatment response to ketamine in TRD patients using a systematic review of brain MRI studies published from January 2018 to May 2024.With the goal of finding structural and functional brain characteristics associated with a treatment response, the current study considers multiple modalities of brain MRI acquired before and/or after ketamine treatment for TRD: (1) T1-weighted images to extract regional gray matter volumes, (2) diffusion-weighted images to estimate white matter tracts and their structural integrity, (3) T2*-images acquired during task performance to calculate task-related functional activation patterns, (4) T2*-images acquired in the resting-state to calculate the resting-state functional connectivity, and (5) arterial spin-labeling images acquired in the resting-state to measure regional brain perfusion.Of note, intermediate brain markers of a response to ketamine administration for item-level depressive symptoms, including suicidality and anhedonia (Jawad et al., 2023;Jollant et al., 2023), were searched to find target depressive symptom(s) for which ketamine administration could be applied as a rescue treatment.We hypothesized that a treatment response to ketamine administration would have associations with both the baseline profile and ketamine-related changes in brain regions that compose the default mode, salience, fronto-parietal, subcortical, and affective (limbic) networks (Marwaha et al., 2023;Runia et al., 2022;Yun and Kim, 2021).

Registration, databases, and search terms
To achieve a high standard of reporting, this systematic literature review was conducted and is reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) statement (Page et al., 2021).This systematic literature review was also submitted to the PROSPERO (https://www.crd.york.ac.uk/prospero/).MRI-based studies of pharmacotherapy with ketamine for depression were identified from peer-reviewed human research articles published in English.A comprehensive search in PubMed was performed on February 29, 2024, using "ketamine AND depression AND magnetic resonance."The time span for the database queries was "Start date: 2018/01/01; End date: 2024/05/31" (initial search, April 2022; updated search: May 2024).Additional records were sought from the reference sections of included papers and neuroimaging reviews of ketamine use in depression.The database search and cross-referencing yielded 126 records.

Inclusion and exclusion criteria
After the removal of duplicates, the titles and abstracts of all retrieved articles were screened by one author (JYY).Studies eligible for this systematic review satisfied these inclusion criteria: 1) participation of human individuals diagnosed with depression; 2) brain MRI (T1WI, DWI, or T2* images) before the pharmacotherapy with ketamine, 3) results of the pharmacotherapy with ketamine provided regarding (1) longitudinal comparisons of the brain MRI profile pre-versus posttreatment, (2) associations between improvements in depressive symptoms and changes in brain MRI features after treatment, or (3) predictive features for a treatment response found in the baseline brain MRI features of depressive patients.The exclusion criteria were 1) non-human participants (animal study), 2) no inclusion of patients (only healthy volunteers), 3) patients diagnosed with affective psychotic disorders (e. g., bipolar disorder or major depressive disorder with psychotic features), 4) no pharmacotherapy with ketamine, 5) non-MRI-based methods (e.g., electroencephalography, positron emission tomography, or magnetic resonance spectroscopy), or 6) conference abstracts, case reports, letters to the editor, reviews, and meta-analyses.
The PubMed search initially yielded 126 publications.Fig. 1 presents the PRISMA flowchart.Following the removal of duplicates, conference abstracts, case reports, letters to the editor, reviews, and meta-analyses, 126 publications remained.The screening of titles and abstracts by JYY yielded 126 articles that underwent full-text assessment for eligibility.Eighty-five additional articles were further excluded because they did not meet the selection criteria.The full text of the remaining articles was assessed for eligibility, and a final decision was made regarding inclusion to this systematic review.In that way, 41 articles comprising 1,396 patients (male/female = 644/752) with TRD (patients diagnosed with depression having a history of failure with different antidepressants with an adequate dosage and treatment duration) and 587 healthy controls (male/female = 264/323) were included in this systematic review.
Diagnosis of depression was made on the diagnostic criteria of depression defined in the Diagnostic and Statistical Manual of Mental Illnesses [DSM; DSM-5 (N = 19), DSM-IV-TR (N = 8), and/or DSM-IV (N = 11)], the International Classification of Diseases [ICD; ICD-10 (N = 1)], and/or by clinical assessment of psychiatrists (N = 3), and used the diagnostic instruments of the Structured Clinical Interview for DSM Disorders (SCID; N = 20), the Mini-International Neuropsychiatric Interview (MINI; N = 4), the Kiddie Schedule for Affective Disorders and Schizophrenia, Present and Lifetime Version (K-SADS-PL; N = 1), and/or the clinical assessment by psychiatrists (N = 16).To be classified as TRD, histories of ≥ 2 (N = 25) or ≥ 1 (N = 7) treatment failures for depression with different antidepressants with an adequate dosage and treatment duration were required; some studies did not provide number of In regards of the psychiatric comorbidity, patients diagnosed with affective psychotic disorders were excluded in all studies (N = 41).Detailed information of psychiatric comorbidity was available in 9 studies [anxiety disorder (N = 1), generalized anxiety disorder (N = 4), post-traumatic stress disorder (N = 4), panic disorder (N = 2), obsessivecompulsive disorder (N = 1), bipolar disorder (N = 3), and/or history of substance abuse (N = 2)].

Protocol of ketamine administration
In terms of the effective dose of ketamine, most studies (N = 38) applied ketamine [0.5mg/kg racemic ketamine (N = 38) and/or 0.25mg/kg S-ketamine (N = 4)] diluted in 60cc of normal saline via intravenous infusion over 40 min one time (N = 24), four times (N = 9), or six times (N = 5) spaced 2-3 days apart over 2 weeks.Other 3 studies administered oral ketamine in a dosage range of 0.5-3.0mg/kgover 6 weeks (Dutton et al., 2024;Gallay et al., 2021) or a 96-hour intravenous infusion of ketamine combined with clonidine with a ketamine dosage range of 0.15-0.6mg/kg/hour(Siegel et al., 2021).Concomitant administration of psychotropic medication including antidepressants was permitted in some studies (N = 27) or not allowed in others (N = 10; only unmedicated patients participated); some studies did not provide detailed information (N = 4).

Measurement of depressive symptoms at baseline and after ketamine application
The severity of depressive symptoms at baseline and after ketamine infusion was measured using the Montgomery-Asberg Depression Rating Scale (MADRS; N = 23), Hamilton Depression Rating Scale (HDRS; N = 16), Beck Depression Inventory (BDI; N = 5), and/or Children's Depression Rating Scale-revised (CDRS-R; N = 1).The severity of depressive symptoms reflected in the mean total scores of those measures per study (N = 38) demonstrated that the TRD participants suffered moderate [MADRS mean total score of 20-34 (N = 19), HDRS-17

Table 1
Studies that examined associations of treatment response to ketamine with brain T1WI and DWI in treatment-resistant depression.The intensity of specific depressive symptoms of anhedonia [Snaith-Hamilton Pleasure Scale (SHAPS; N = 6), and Temporal Experience of Pleasure Scale (N = 2)], apathy [Apathy Evaluation Scale (N = 1)], rumination [Rumination Response Scale (N = 2)], and suicidal ideation [Beck Scale for Suicidal Ideation (N = 3), and the 10 th item of the MADRS (N = 2)] was also measured.Of note, information about suicidality [current suicidal ideation (N = 8) and/or previous suicidal attempts (N = 7)] at baseline was described in 13 studies, but only 3 studies (Chen et al., 2019;Gallay et al., 2021;Li et al., 2023) of those 13 studies compared the intensity of suicidal ideation after ketamine infusion with that at baseline.

Acquisition of brain magnetic resonance imaging
Brain MRI acquisition was generally performed both at baseline and after ketamine infusion [2-4 hours (N = 2), 1-3 days (N = 32), 5-7 days (N = 2), 10-14 days (N = 4), and/or 28 days (N = 1) of post-infusion]; but in some studies, MRI data were gathered only at baseline (N = 4) or during the ketamine infusion (N = 1).All studies (N = 41) acquired brain MRI images using a field strength value of 3 Tesla, with one study also applied a filed strength value of 7T for study sub-population (Evans et al., 2024), and diverse facets of brain features were explored (including 2 studies of multi-modal imaging approach (Siegel et al., 2021;Wade et al., 2022)): region-to-region functional connectivity networks in the resting-state (N = 19) or task-related regional functional activation (N = 9) were measured using T2*-weighted imaging; cortical-subcortical gray matter volumes and morphology were calculated from T1WI images (N = 10); the structural integrity of brain white matter tracts was estimated from DWI (N = 4); and cerebral blood flow was measured using arterial spin labeling (N = 2).

Data extraction
The following data were extracted from each included study: (1) study design, numbers of participants (patients and healthy controls); (2) demographics: age and sex; (3) clinical information: illness duration,

Table 3
Studies that examined associations of response to ketamine in treatment-resistant depression with task-based fMRI/multimodal imaging.rating scale(s) used to measure depressive symptoms, severity of depressive symptoms pre-/post-pharmacotherapy with ketamine, suicidality, pharmacotherapy status; (4) brain MRI: modality [T1WI, DWI, task-based functional MRI (fMRI), or resting-state fMRI], acquisition parameters, sequence, and field strength; and (5) MRI analysis: statistical methods, main results.In the results section followed, only the study results replicated by ≥ 2 studies regarded meaningful and were reported (Figs. 2 and 3).

Brain morphometry
Six studies among the total 10 studies that acquired T1WI examined grey matter morphological changes after the pharmacotherapy with ketamine in TRD than baseline (Table 1).One study explored whole brain by way of the voxel-based morphometry; regions-of-interest comprising the subcortical, limbic, default-mode, and frontoparietal networks were also examined in 5 studies.In Fig. 2, the results showed increased grey matter volumes in the left amygdala (Evans et al., 2024;Zhou et al., 2020a) and right hippocampus (Zhou et al., 2020a;Zhou et al., 2020b) after single or six serial intravenous infusion of ketamine in TRD.Of note, in TRD with suicidality, treatment with ketamine was associated with decreased volumes in the right dorsolateral prefrontal cortex (single intravenous infusion) (Li et al., 2023) as well as increased volumes in the left amygdala, bilateral hippocampi, nucleus accumbens, striatum, and thalamus (either six intravenous infusion/2 weeks or six oral administration/6 weeks) (Dutton et al., 2024;Gallay et al., 2021;Zhou et al., 2020a;Zhou et al., 2020b).

Structural connectivity
Four studies acquired DWI [DTI (N = 3) and neurite orientation dispersion and density imaging (NODDI (Kraguljac et al., 2023); N = 1)] at baseline and after the treatment with ketamine in TRD (Table 1).Longitudinal comparisons of MRI after ketamine administration than baseline were performed in 3 studies; structural integrity of white matter tracts were measured in multiple white matter tracts (N = 2) or only in the forceps and cingulum bundles (N = 1).In Fig. 2, the results showed that the structural integrity of inferior longitudinal fasciculus (= a backbone of the ventral visual stream that connects the ventral surface of the anterior temporal lobe to the extra-striate cortex of the occipital lobe (Herbet et al., 2018)) might be initially increased (measured using the fractional anisotropy; after single intravenous infusion) (Sydnor et al., 2020) but lowered again [evaluated with the neurite density index (= the packing density of axons or dendrites) (Kraguljac et al., 2023); after four intravenous infusion/8days] (Taraku et al., 2023) in response to the administration of ketamine in TRD.

Resting-state functional connectivity
By way of the T2* MRI (19 studies) or ASL (2 studies) acquired in TRD both at baseline and after the pharmacotherapy with ketamine, inter-regional functional connectivity and regional perfusion at restingstate were examined (Table 2).Longitudinal comparisons of restingstate functional connectivity and/or functional activation/cerebral perfusion between post-ketamine than baseline were done in 13 studies (11 studies of T2* MRI and 2 studies of ASL).In Fig. 3, intravenous infusion of ketamine to TRD resulted in the increased resting-state functional connectivity of limbic network with the components of default mode, salience, and subcortical networks as well as decreased resting-state functional connectivity with fronto-parietal networks.
First, increased rs-FC between subgenual ACC versus ventromedial prefrontal cortices (Alexander et al., 2023;Chen et al., 2019) and between right habenula (a diencephalic structure that acts as an interface between the limbic forebrain and brainstem nuclei) versus left parahippocampal gyrus (Rivas-Grajales et al., 2021), were found.Second, the rs-FC between subgenual ACC versus pregenual ACC (Alexander et al., 2023;Chen et al., 2019) and between subgenual ACC versus bilateral caudal ACC and anterior insular cortices (Siegel et al., 2021) were increased.Third, increased rs-FC was found between the striatum versus subgenual ACC (Alexander et al., 2023;Chen et al., 2019) and orbitofrontal cortex (Mkrtchian et al., 2021).On the contrary, decreased rs-FC between the fronto-parietal network of right DLPFC with right frontal pole (Abdallah et al., 2021;Chen et al., 2019) and emergence of anti-correlation between hippocampus versus left fronto-parietal network components (Vasavada et al., 2021) were found.In regards of the regional changes in the cerebral blow perfusion (measured using arterial spin labeling) and functional activation after the intravenous ketamine infusion in TRD, increased functional activation and/or cerebral blow perfusion was shown in brain regions comprising the default-mode (posterior cingulate cortex and precuneus), subcortical (thalamus), sensory-motor (paracingulate gyrus, paracentral lobule, and N motor cortex), and visual (lateral occipital cortex, fusiform gyrus, occipital pole, and cuneus) networks (Gonzalez et al., 2020;McMillan et al., 2020;Sahib et al., 2020b).

Functional brain activation during task performance
To uncover the functional brain activation in regards of the emotional regulation in the middle of higher cognitive performance (such as working memory, conflict control, and response inhibition), total 9 studies applied tasks during the acquisition of T2* MRI in TRD both at baseline after the treatment with ketamine (Table 3).In analyses of MRI images, four studies evaluated whole brain regions and other 5 studies examined selected brain regions.In Fig. 2, longitudinal comparisons of MRI between baseline versus post-ketamine were conducted in 6 studies.
After intravenous ketamine infusion to TRD, functional activation of dorsal ACC was increased for emotional processing of positive valence (Reed et al., 2019) but was reduced for the conflict control in the middle of emotional face stimuli (Reed et al., 2018).Second, functional activation of DLPFC was reduced both during the response inhibition (Sahib et al., 2020a) and conflict control in the middle of emotional face stimuli (Reed et al., 2018).Third, in the limbic network, intravenous ketamine infusion resulted in increased functional activation of NAc and OFC for positive feedback (Sterpenich et al., 2019), reduced functional activation of subgenual ACC in response to the positive feedback (Morris et al., 2020), and reduced functional activation in the right amygdala for the fearful emotional stimuli (Loureiro et al., 2020).Fourth, task-based functional activation of default mode network during the emotional processing of positive valence was increased in the posterior cingulate cortex and parahippocampal cortex (Reed et al., 2019) but was decreased in the medial PFC (Reed et al., 2018).

Brain morphometry
Among the total 10 studies that acquired T1WI at baseline and after the pharmacotherapy with ketamine in TRD, four studies examined associated features of treatment response to ketamine in the baseline T1WI (Table 1).In Fig. 2, larger pretreatment volumes in the right thalamus, right rostral ACC, and a hippocampal sub-region of the left subiculum in TRD correlated with a greater reduction in depressive symptoms after single or four serial ketamine infusions (Herrera-Melendez et al., 2021;Siegel et al., 2021;Zhou et al., 2020a).Conversely, greater improvement in depressive symptoms in TRD after a continuous 96-hour infusion of intravenous ketamine (0.15mg/kg/hr-0.6mg/kg/hr) was associated with a smaller gray matter volume in the right hippocampus at baseline (Siegel et al., 2021).In regards of the item-level depressive symptoms, more improvement of ruminative thoughts after ketamine administration could be predicted by smaller grey matter volume of putamen at baseline (Wade et al.,

Structural connectivity
Among the total 4 studies that acquired DWI at baseline and after the pharmacotherapy with ketamine in TRD, two studies examined associated features of treatment response to ketamine in the baseline DWI (Table 1).In Fig. 2, higher baseline values of the diffusion kurtosis [= a direct estimates of deviation in regards of the water diffusion from a single Gaussian component; a positive value indicates a distribution more peaked than normal (Henriques et al., 2021)] and/or fractional anisotropy (= a measure of the structural integrity of the white matter tract of cellular tissue) in the white matter tracts of superior longitudinal fasciculus [= a large associative bundle of fibers connecting the parietal, occipital and temporal lobes with ipsilateral frontal cortices (Kamali et al., 2014)] and the hippocampal and cingulate gyrus portions of left cingulum bundle [= fiber tracts forming a ring from the orbitofrontal cortex, along the dorsal surface of the corpus callosum, then down the temporal lobe towards the temporal pole (Bubb et al., 2018)] were associated with a lesser reduction in depressive symptoms in TRD after receiving the single or four serial intravenous ketamine infusions (Langhein et al., 2022;Sydnor et al., 2020;Wade et al., 2022).

Resting-state functional connectivity
Among the total 21 studies that acquired either T2* MRI (19 studies) or ASL (2 studies) at baseline and after the administration of ketamine in TRD, nine studies examined associated features of treatment response to ketamine in the baseline resting-state functional brain imaging (Table 2).In Fig. 3, the results showed that stronger rs-FC between limbic network and default mode network in TRD at baseline could predict better reduction of depressive symptoms after intravenous infusion of ketamine.
In other words, stronger rs-FC of habenula with precuneus, middle temporal gyrus, and angular gyrus (Wang et al., 2022) and of rostral hippocampus with inferior parietal lobule and angular gyrus (Zhang et al., 2023) at baseline were associated with more improvement of depressive symptoms after six serial intravenous infusion of ketamine in TRD.
Further, stronger rs-FC between the cerebellum versus left putamen and insula (Sahib et al., 2022) as well as the weaker rs-FC between the superior frontal cortices versus striatum and subgenual ACC (Chen et al., 2020;Gärtner et al., 2019) and between the subgenual ACC versus amygdala (Nakamura et al., 2021) at baseline were also predictive of better treatment response after single/four/six serial intravenous ketamine infusions in TRD.(Vasavada et al., 2021).On the other hand, lower baseline cerebral blood flow measured using arterial spin labeling (Petcharunpaisan et al., 2010) in the fusiform cortex (Sahib et al., 2020b) as well as stronger baseline global brain connectivity in the right medial PFC, rostral ACC, and right posterior insula (Wade et al., 2022) could predict a larger reduction in depressive symptoms, following four serial intravenous infusion of ketamine in TRD.
Of note, improvement of item-level depressive symptoms in TRD after ketamine infusion could be predicted using the rs-FC measured at baseline.First, better reduction of anhedonia after the completion of 4 th intravenous infusion of ketamine in TRD could be expected based on the increased rs-FC of right hippocampus with left fronto-parietal network after the 1st intravenous ketamine infusion.Second, better improvement of ruminative thoughts in TRD after four serial intravenous ketamine infusions might be associated with weaker baseline global brain connectivity values in the left insular cortex and left posterior cingulate cortex (Wade et al., 2022).

Functional brain activation during task performance
Among the total 9 studies that applied tasks during the acquisition of T2* MRI at baseline and after the administration of ketamine in TRD, four studies examined associations between the treatment response to ketamine and task-related functional brain activation at baseline (Table 3).In Fig. 2, first, stronger functional activation of pregenual ACC during emotional processing (positive or negative) in TRD at baseline was correlated with a larger reduction in depressive symptoms after a single intravenous ketamine infusion (Weigand et al., 2022).Second, stronger task-related functional connections (psychophysiological interactions) of cerebellar lobule VIIb with fronto-parietal and salience networks during response inhibition in TRD at baseline were predictive of more improvement in depressive symptoms after four intravenous infusion of ketamine (Loureiro et al., 2021).Third, weaker functional activation of dorsomedial PFC during the working memory task with emotional regulation in TRD at baseline was related to larger improvements in cognitive symptoms after a single ketamine infusion (Stippl et al., 2021).Fourth, higher functional activation of subgenual ACC in response to the positive monetary feedback following the conflict control task in TRD at baseline was associated with better improvement in anhedonia after a single intravenous ketamine infusion (Morris et al., 2020).

Brain morphometry
Among the total 10 studies that acquired T1WI at baseline and after the pharmacotherapy with ketamine in TRD, four studies examined treatment response-related changes in the grey matter morphology (Table 1).In Fig. 2, the results showed associations between the increased volumes in limbic networks and better treatment response to ketamine administration in TRD.More reduction of depressive symptoms after single or six serial intravenous ketamine infusions in TRD correlated with an increased grey matter volumes in the bilateral hippocampi and its subfields (Dutton et al., 2024;Zhou et al., 2020a), left amygdala (Zhou et al., 2020a), and opercular part of right inferior frontal gyrus (Dai et al., 2020).Conversely, more decrement of gray matter volumes in the left pallidum (Zhou et al., 2020a) and right DLPFC (Li et al., 2023) after the ketamine infusion was related to the less treatment response to single or six serial intravenous infusion of ketamine in TRD.

Structural connectivity
Among the total 4 studies that acquired DWI at baseline and after the pharmacotherapy with ketamine in TRD, four studies examined treatment response-related changes in the DWI (Table 1).In Fig. 2 most of all, larger decrease of neurite density index and/or diffusion kurtosis values in the left superior longitudinal fasciculus and left internal capsule after ketamine administration are associated with more reduction of depressive symptoms including anhedonia in TRD (Taraku et al., 2023;Wade et al., 2022).In addition, a smaller increment in the fractional anisotropy value in the corpus callosum-forceps minor [= the anterior part of the corpus callosum, which connects the homologous regions of the anterior frontal lobes between the two hemispheres (Mamiya et al., 2018)] and bilateral uncinate fasciculi [= a direct, bidirectional monosynaptic pathway connecting the orbitofrontal cortex to the anterior temporal lobes (Von Der Heide et al., 2013)] correlated with a greater reduction in depressive symptoms after a single ketamine infusion in TRD (Sydnor et al., 2020).

Resting-state functional connectivity
Among the total 21 studies that acquired either T2* MRI (19 studies) or ASL (2 studies) at baseline and after the administration of ketamine, eleven studies examined treatment response-related changes of TRD in the resting-state functional brain imaging (Table 2).In Fig. 3, first, increased intra-network rs-FC of limbic network after ketamine administration was related to more reduction of depressive symptoms in TRD.In other words, increased rs-FC between habenula versus right parahippocampal gyrus and frontal (BA 9)-temporal (BA 38) poles, between the subgenual ACC versus orbitofrontal cortex, anterior PFC (BA 10 and J.-Y.Yun and Y.-K. Kim Psychiatry Research 340 (2024) 116092 46), and right DLPFC (Gärtner et al., 2019;Rivas-Grajales et al., 2021) in addition to a reduction in the rs-FC between insula and pregenual ACC (Alexander et al., 2023) after single intravenous ketamine infusion indicated larger reduction of depressive symptoms in TRD.Also, reduction in depressive symptoms after single or six serial ketamine infusions in TRD was associated with an increase in resting state entropy [which predicts the magnitude of brain activation or deactivation at lower frequencies (Lin et al., 2022)] or increased cerebral blood flow in the nucleus accumbens, pallidum, and ventral striatum, as well as reduced cerebral blood flow or smaller changes in the resting-state functional activation in the frontal pole, posterior cingulate cortex, left precuneus, medial PFC, ACC, and insula (Gonzalez et al., 2020;McMillan et al., 2020;Roy et al., 2021).Second, improvement of anhedonia in TRD after ketamine administration was associated with increased rs-FC of subgenual ACC with other subregions of ACC and striatum.Increased rs-FC of dorsal caudate with right ventrolateral PFC and pregenual ACC, and of subgenual ACC with dorsal ACC and ventral striatum after a single ketamine infusion reflected a reduction in anhedonia in TRD (Alexander et al., 2023;Mkrtchian et al., 2021).Moreover, increased global brain connectivity in the right medial PFC and ACC was proportional to the improvement of depressive symptom including anhedonia after four serial intravenous ketamine infusions in TRD (Wade et al., 2022).Third, neural correlates of a reduction in the severity of suicidal ideation after a single ketamine infusion to TRD were increased rs-FC between the left dorsal ACC and the right subgenual ACC and reduced resting-state functional connectivity between the right DLPFC and the left superior parietal cortex (Chen et al., 2019).

Functional brain activation during task performance
Among the total 9 studies that applied tasks during the acquisition of T2* MRI at baseline and after the administration of ketamine in TRD, four studies examined associations of treatment response-related changes with changes in the task-related functional brain activation in TRD (Table 3).In Fig. 2, first, when TRD performed a facial emotion recognition and matching task after four ketamine infusions, increased functional activation for happy faces and reduced functional activation for fearful face in the right DLPFC were associated with a larger reduction in depressive symptoms and anhedonia (Loureiro et al., 2020).Second, the remission of depressive symptoms after four serial ketamine infusions in TRD correlated with reduced functional co-activation for response inhibition and inhibitory control between the left cerebellar lobule VIIb and the fronto-parietal and sensory-motor networks (Loureiro et al., 2021).Third, stronger functional activation for angry face and weaker functional activation for happy face in the left parahippocampal gyrus and amygdala for conflict control with emotional processing after a single ketamine infusion correlated positively with a larger improvement in depressive symptoms in TRD (Reed et al., 2018).

Neural correlates of depressive symptom reduction after ketamine infusion in TRD patients
This review revealed associations of the better treatment response to ketamine in regards of depressive symptom severity in TRD with the lower structural integrity in the superior longitudinal fasciculus and the cingulate gyrus-hippocampal portions of the cingulum bundle at baseline, with the stronger resting-state functional connectivity between the limbic network versus default-mode network at baseline, in addition to the increased hippocampal volumes and increased intranetwork resting-state functional connectivity of limbic network after ketamine administration than baseline.The superior longitudinal fasciculus connects nearly all cortical areas of the lateral cerebral hemisphere, including the frontal, parietal, and parts of the temporal lobes (Nakajima et al., 2020).The cingulum bundle forms a near-complete ring that covers the orbital frontal cortices along the dorsal surface of the corpus callosum and then continues down the temporal lobe toward the pole (Wu et al., 2016).The structural integrity of the hippocampal portion of the cingulum bundle at baseline was the strongest predictor of a treatment response after 8 weeks of pharmacotherapy with a selective serotonin reuptake inhibitor in patients with depression (Davis et al., 2019).Of note, gray matter volumes and glucose metabolism values in brain regions that largely overlap with the white matter tracts of the cingulum bundle might contribute to the prediction of a treatment response to deep brain stimulation targeting the subcallosal cingulate [which is the intersection of the forceps minor, uncinate fasciculus, cingulum bundle, and fronto-striatal fiber bundles (Tsolaki et al., 2021)] in TRD (Elias et al., 2022).Collectively, improvement of depressive symptom severity after ketamine administration in TRD is associated with brain features of affective limbic network, salience network, fronto-parietal network, and default-mode network.

Neural correlates of thought rumination relief after ketamine infusion in TRD patients
This review also uncovered relationship of better treatment response to ketamine in TRD in regards of the improvement in ruminative thoughts with the smaller grey matter volume of putamen and weaker resting-state global brain connectivity in the insula and posterior cingulate cortex at baseline.Thought rumination is a cognitive process that focuses on self-relevant and negative experiences and could be associated with poor sleep quality (Chang et al., 2023), increased risk of depression (Lawrence et al., 2023), and poor prognosis in major depressive disorder (Tsuchiyagaito et al., 2023)].This is in concordance with previous studies that reported associations of insula with treatment response to pharmacotherapy, electroconvulsive therapy, and mindfulness training in patients with depression.The baseline resting-state functional connectivity between the insula and other regions is a predictive marker of response to pharmacotherapy with sertraline in patients with major depressive disorder (Zhao et al., 2023).Likewise, increased functional activation of the bilateral insula after one week of pharmacotherapy with either citalopram or quetiapine, compared with baseline, predicted the treatment response after 8 weeks of pharmacotherapy for depression (Williams et al., 2021).Moreover, increased functional activation of the left insula during interoception (focusing on their own heartbeat) accompanied an increased ability to trust in their own body sensations among patients with moderate-to-severe depression who received mindfulness training (Datko et al., 2022).On the contrary, shrinkage of gray matter volume in the right insula during the first six months after index electroconvulsive therapy sessions in TRD patients correlated with the re-elevation of depressive symptoms after the index therapy (Brooks et al., 2023).Collectively, alleviation of thought rumination in response to ketamine administration in TRD is related to the brain features of default-mode network and salience network.

Neural correlates of anhedonia/apathy reduction after ketamine infusion in TRD patients
This review demonstrated associations of better reduction of anhedonia after ketamine administration in TRD with stronger baseline resting-state functional connectivity between the hippocampus versus fronto-parietal network, with stronger functional activation of subgenual ACC with the positive reward of conflict control at baseline.Treatment response for ketamine administration in TRD in regards of the improvement of anhedonia was also proportional to the decrease of neurite density and diffusional heterogeneity in the left superior longitudinal fasciculus [which connects nearly all cortical areas of the lateral cerebral hemisphere, including the frontal, parietal, and parts of the temporal lobes (Nakajima et al., 2020)], changed functional activation in the DLPFC for facial emotions, and increased resting-state functional connectivity between the subgenual ACC versus striatum and other sub-regions of ACC.The intensity of anticipatory anhedonia in adolescents with depression correlated with the summed strengths of resting-state functional connectivity between the subgenual ACC versus dorsal ACC and the caudate versus other brain regions (Ely et al., 2021).In short, affective limbic network and fronto-parietal network could be a putative brain-based marker of anhedonia reduction after ketamine administration in TRD.

Neural correlates of a treatment response in suicidal ideation after ketamine infusion in TRD
In this review, associations of reduction in the severity of suicidal ideation after ketamine administration in TRD with increased rs-FC between the left dorsal ACC versus right subgenual ACC and reduced resting-state functional connectivity between the right DLPFC versus left superior parietal cortex (dorsal attention network) were found.These are in accordance with other studies that demonstrated the importance of ACC in suicidality.First, suicide attempters with a diagnosis of bipolar disorder showed increased functional oscillations at a lower frequency (measured using the dynamic amplitude of low-frequency fluctuations) than HC in the right ACC (Tian et al., 2023).Second, stronger resting-state functional connectivity is found in patients with bipolar depression than unipolar depression not only in the whole-brain functional connectivity of prefrontal cortex, ACC, thalamus, and parietal and temporal cortices, but also in the intra-network functional connectivity of default-mode and fronto-parietal networks (Goya-Maldonado et al., 2016;Wang et al., 2017a;Wang et al., 2017b).Patients with bipolar depression had reduced white matter integrity in the anterior part of the corpus callosum and posterior cingulum, which play a critical role in communication within emotion-and cognition-processing neural circuits (Emsell et al., 2013;Fani et al., 2014), than UD (Matsuoka et al., 2017;Repple et al., 2017;Wise et al., 2016).Functional activation/deactivation of fronto-striatal regions in the default-mode network during the performance of visuospatial planning and working memory tasks could differentiate unipolar depression from bipolar depression (Rive et al., 2016;Rodríguez-Cano et al., 2017).With regard to cortical thickness differences, a thinner DLPFC (rostral middle frontal cortex) is found in bipolar depression compared to both unipolar depression and HC (Lan et al., 2014;Niu et al., 2017;Hibar et al., 2018).In bipolar disorder, carrying the BDNF Val66Met risk allele is associated with reduced hippocampal volumes compared to non-carriers (Pereira et al., 2017).Greater and weaker functional activations of amygdala in unipolar depression than bipolar depression are found in response to the negative and positive emotional facial stimuli, respectively (Fournier et al., 2013;Grotegerd et al., 2014;Grotegerd et al., 2013).Differences in functional activity of ventral or dorsal ACC between bipolar depression versus unipolar depression in response to the emotional facial stimuli are also found (Bertocci et al., 2012;Bürger et al., 2017).Collectively, treatment response of suicidal ideation after ketamine administration in TRD might have associations with possible subthreshold bipolarity and brain features of salience network, default-mode network, and fronto-parietal network.

Limitations
The current study has some limitations to be addressed.First, the heterogeneity of these 41 studies in relation to the clinical and technical variables could affect measurements and findings.Heterogeneity of TRD also has to be mentioned.While many studies required histories of ≥ 2 (N = 25) or ≥ 1 (N = 7) treatment failures for depression with different antidepressants with an adequate dosage and treatment duration to be classified as TRD; some studies did not provide number of treatment failures to be classified as TRD (N = 9).Therefore, in ranking the evidence acquired from each studies in terms of the credibility of the findings, based on the recent studies of the neurobiology in depression (Yun and Kim, 2021), we selected the default mode, salience, fronto-parietal, subcortical, and affective limbic networks as regions-of-interest, and regarded only the study results replicated by ≥ 2 studies as meaningful in this review.Second, most of the studies published to date examined the acute or subacute effects of ketamine infusion on the brain, using measurements taken 1-3 days post-treatment.Third, only a few studies reported the effects of ketamine infusions in relieving item-level depressive symptoms such as suicidality in TRD.Fourth, only a little information is available to date about the brain white matter tract-related factors in the treatment response to ketamine infusion for TRD.Therefore, further studies using multi-modal brain imaging (Kundu et al., 2021) are needed to explore long-term treatment effects (Deyama and Kaneda, 2023) of ketamine infusion (Li et al., 2017;Zimmermann et al., 2022) in TRD (Khattab et al., 2022).Also, the combined use of a clinical symptom network and neuroimaging (Yun and Kim, 2020;Yun et al., 2022) could help to uncover the multi-layered associations among brain network characteristics and item-level depressive symptoms in TRD.

Conclusions
The results of our systematic review suggest that brain gray matter volumes, the structural integrity of white matter tracts, task-related functional activation, and resting-state functional connectivity could be useful candidate brain-based features for predicting the treatment response to ketamine in patietns with TRD.Structural and functional brain features related to the affective limbic (amygdala, hippocampus, and subgenual ACC), fronto-parietal (DLPFC), salience (dorsal ACC and insula), default-mode (dorsomedial PFC and posterior cingulate cortex), striatum, superior and inferior longitudinal fasciculi, and cingulate gyrus-hippocampal portions of the cingulum bundle could be useful in finding the subgroup of TRD patients with a high probability of responding to ketamine treatment with relief of depressive symptoms, including anhedonia, thought rumination, and suicidal ideation..

Ethics approval
This article belongs to review, which does not contain any studies with human participants or animals performed by any of the authors.