Structural MRI-Based Predictions in Patients with Treatment-Refractory Depression (TRD)

Blair A. Johnston; J. Douglas Steele; Serenella Tolomeo; David Christmas; Keith Matthews

doi:10.1371/journal.pone.0132958

Abstract

The application of machine learning techniques to psychiatric neuroimaging offers the possibility to identify robust, reliable and objective disease biomarkers both within and between contemporary syndromal diagnoses that could guide routine clinical practice. The use of quantitative methods to identify psychiatric biomarkers is consequently important, particularly with a view to making predictions relevant to individual patients, rather than at a group-level. Here, we describe predictions of treatment-refractory depression (TRD) diagnosis using structural T₁-weighted brain scans obtained from twenty adult participants with TRD and 21 never depressed controls. We report 85% accuracy of individual subject diagnostic prediction. Using an automated feature selection method, the major brain regions supporting this significant classification were in the caudate, insula, habenula and periventricular grey matter. It was not, however, possible to predict the degree of ‘treatment resistance’ in individual patients, at least as quantified by the Massachusetts General Hospital (MGH-S) clinical staging method; but the insula was again identified as a region of interest. Structural brain imaging data alone can be used to predict diagnostic status, but not MGH-S staging, with a high degree of accuracy in patients with TRD.

Citation: Johnston BA, Steele JD, Tolomeo S, Christmas D, Matthews K (2015) Structural MRI-Based Predictions in Patients with Treatment-Refractory Depression (TRD). PLoS ONE 10(7): e0132958. https://doi.org/10.1371/journal.pone.0132958

Editor: Martin Walter, Leibniz Institute for Neurobiology, GERMANY

Received: February 17, 2015; Accepted: June 21, 2015; Published: July 17, 2015

Copyright: © 2015 Johnston et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited

Data Availability: Relevant data are within the paper and its Supporting Information files.

Funding: The project was supported by two awards from an Anonymous Trust administered by the University of Dundee and, in part, by staff resource from the Scottish Mental Health Research Network (SMHRN). BAJ was partially supported by a SINAPSE (www.sinapse.ac.uk) studentship, which included a SINAPSE-SPIRIT industry partnership with Siemens Medical. The funding sources played no part in the design of the study, analysis and interpretation of the data, nor manuscript preparation.

Competing interests: BAJ was partially supported by a SINAPSE (www.sinapse.ac.uk) studentship, which included a SINAPSE-SPIRIT industry partnership with Siemens Medical. JDS has received research funding via an honorarium associated with a lecture from Wyeth. ST has received research project funding from Lundbeck & Merck Serono. KM has Chaired advisory boards for studies of Deep Brain Stimulation for Obsessive-Compulsive Disorder sponsored by Medtronic. He has received educational grants from Cyberonics Inc. & Schering Plough, and he has received research project funding from Lundbeck, Merck Serono & Reckitt Benckiser and also from St Jude Medical for a multi-centre clinical trial of Deep Brain Stimulation for depression. He has received travel and accommodation support to attend meetings from Medtronic and St Jude Medical. DC reported no biomedical financial interests or potential conflicts of interest. There are no patents, products in development or marketed products to declare. This does not alter the authors' adherence to all the PLOS ONE policies on sharing data and materials.

Introduction

Based on the 2010 Global Burden of Diseases Study, Whiteford, Degenhardt (1] estimated that mental and substance use disorders were the leading causes of years lived with disability worldwide. This is due to the chronic nature of psychiatric disorders and the typical onset of symptoms at a young age. Furthermore, depressive disorders were found to be the largest contributor to disability-adjusted life years, by some margin, within this grouping [1]. Major Depressive Disorder (MDD) is defined by persistent and disabling symptoms of low mood, anhedonia, hopelessness, guilt, low self-worth, poor concentration, lack of energy, suicidal thoughts and altered appetite and sleep [2, 3]. The diagnosis is elicited by clinical interview and mental state examination in the absence of robust established pathophysiological mechanisms or biomarkers. Remission rates with standard antidepressant drug treatments are only 30–40%, even when depression is accurately diagnosed and patient adherence is excellent [4–6]. The reliance on subjective symptoms in the absence of objective biomarkers, along with under-developed understanding of depressive endophenotypes may explain these treatment gaps.

It is, therefore, necessary to have a better understanding of the underlying mechanisms of depressive illness and to be able to identify objective biomarkers that can be used to make clinically-useful predictions for individual patients. Studies involving patients with so-called ‘treatment-resistant’ or, rather, treatment-refractory depression (TRD) are important because, by definition, these are the patients who do not respond well to available treatments and who experience chronic impairment, disability and social disadvantage. We have adopted the Massachusetts General Hospital—Staging (MGH-S) protocol [7] for quantifying treatment-resistance (similar to the Antidepressant Treatment History Form approach [8]). The crucial concept is of a ‘failed adequate trial’ of a treatment intervention; usually a medication. An adequate trial requires four criteria to be simultaneously satisfied: i) prescription of a given antidepressant medication for which there is evidence for efficacy with some patients with the diagnosis, ii) at a given minimum dose, iii) for a given minimum duration of time, iv) with reasonable certainty the patient took the medication as prescribed. If all four criteria are not met then the treatment does not constitute an ‘adequate’ trial. If all criteria are satisfied and the patient does not respond to treatment, then this is conceptualised as a ‘failed adequate trial’. Essentially, the total number of failed adequate trials (with some weightings to the scoring) constitutes a numerical measure of treatment resistance for an individual patient.

Whilst patients who are highly treatment refractory tend to experience chronic and enduring illness (evidenced by e.g. high Montgomery-Åsberg Depression Rating Scale (MADRS) or Beck Depression Inventory-II (BDI) scores) and functional impairment, they are not necessarily “in episode” continuously. For example, an individual patient can have shown minimal or no response to several ‘adequate’ medication trials and psychological therapy before finally responding to a course of electroconvulsive therapy. Such a patient would still be considered to fall within the category of treatment refractory despite the improved current health status. Treatment-resistance is, therefore, regarded as an enduring propensity to be resistant to antidepressant treatment and to be at increased risk of further episodes on the basis of previous clinical features. This can be considered as analogous to, for example, epilepsy, where a patient does not cease to have a propensity to epileptic seizures despite no recent seizure activity, particularly where seizure control has been difficult to achieve. This conceptualisation has the advantage of avoiding a conflation of concepts of current illness severity (e.g. MADRS) and treatment-resistance (e.g. MGH-S).

Ultimately, improved understanding of the underlying mechanisms contributing to TRD could facilitate more effective targeting of drug discovery research, substantially impacting on chronic disability worldwide. However, there are few studies of this important clinical population [6]. In the imaging studies that have focused on TRD, reduced hippocampal volume has been consistently implicated. Studies by Shah and colleagues reported reduced grey matter density in the left temporal cortex, including the hippocampus [9, 10], in TRD patients compared both with controls and recovered MDD patients. A more recent volumetric study reported that both TRD and treatment-resistant schizophrenia patients showed reduced hippocampal volumes compared with controls [11]. The volume of the entorhinal cortex, a structure with strong connections with the hippocampus, has also been reported to be significantly reduced in female patients with TRD compared with controls, but not in the corresponding male comparison [12].

Whilst there have been only a few studies specifically investigating patients with TRD, there are many studies which have reported group-level differences in brain structure between patients with MDD (non-TRD) and healthy controls; a majority of which have reported regional reductions in grey matter volume compared with healthy controls [10, 13–15]. MDD patients, compared with healthy controls, have been reported to show decreased grey matter in one or more meta-analysis in: the bilateral rostral anterior cingulate cortex; putamen; caudate; insula; globus pallidus; thalamus; hippocampus; and diverse areas within the frontal lobes [14, 16, 17]. Kempton et al. [17] also reported that MDD subjects had larger lateral ventricular and CSF volumes compared with controls. The amygdala and thalamus were reported to show no differences between groups by Kooschijn et al. [14], but the latter was found to be decreased in MDD patients in the more recent meta-analysis by Kempton et al. [17].

Regions identified using voxel-based morphometry (VBM) provide information about group-level differences. By contrast, techniques based on machine learning, such as support vector machines (SVM), which include additional techniques such as feature selection, can determine which brain regions consistently differ between groups in order to produce an accurate individual subject classifier. Although no reported study has used machine learning-based techniques to classify individual subjects with TRD, a number of studies have used these techniques to classify MDD patients and healthy controls ranging between 68–90% accuracy [18–20]. The highest accuracy reported was obtained by Mwangi and colleagues [20], who successfully classified structural MR images between MDD patients and healthy controls using images obtained over two scanning centres. In that study, grey matter reductions were identified in MDD participants compared to controls in the dorsolateral prefrontal cortex, medial frontal cortex, orbitofrontal cortex, temporal lobe, insula, cerebellum and posterior lobe. Identification of a pattern of brain abnormalities that is able to accurately predict TRD vs. healthy controls in individual subjects using machine learning-based techniques could increase the understanding of and elucidate the mechanisms of TRD. Similarly, identification of brain regions that support prediction and correlate with, for example, the MGH-S staging method, could be used to define novel biomarkers relevant to treatment non-response.

Another aim of the study was to investigate whether it was possible to make accurate individual predictions of symptom severity scores derived from the self-rated BDI and/or the clinician-rated MADRS and 17-item Hamilton Depression Rating Scale (HDRS₁₇) using the same TRD participant structural images. Mwangi and colleagues previously used structural MR images to predict MDD illness severity [21]. In that study, they found that it was possible to predict BDI scores from whole-brain structural MRI scans [21].

Therefore, the primary aims of the present study were to:

classify accurately TRD participants and never-depressed controls on the basis of structural MRI scan data alone,
investigate whether the level of treatment resistance derived from the Massachusetts General Hospital (MGH-S) staging method [7] could be accurately predicted in the patient group,
investigate whether symptom severity scores derived from the self-rated BDI and/or the clinician-rated HDRS₁₇ and MADRS could be accurately predicted in the patient group.

Materials and Methods

Structural T₁ weighted scans and contemporaneous clinical ratings were obtained at the Clinical Research Centre, Ninewells Hospital and Medical School in Dundee, UK. The study was approved by the East of Scotland Research Ethics Service. Written confirmation of informed consent was obtained from all volunteers.

Participants

Twenty adults with a lifetime history of TRD, but who did not necessarily meet criteria for diagnosis of depressive episode at time of scanning were recruited from the Advanced Interventions Service in Dundee. All TRD participants had experienced lifetime and/or current chronic episodes of depression (>24m). Diagnoses were made by experienced clinicians according to DSM-IV criteria using the MINI PLUS (version 5.0) interview schedule [22]. The level of treatment non-response was based on detailed case note review and was scored using the MGH-S [7]; a clinical staging method for evaluating ‘treatment resistance’ that assigns points for ‘adequate’ trials (i.e. exceeding a minimum dose and duration of a given medication) of antidepressant medication. Scoring incorporates optimisation of antidepressant dose, antidepressant combinations and treatment augmentation strategies.

Study exclusion criteria included potentially confounding clinical features including: other primary psychiatric disorder (e.g. bipolar disorder, personality disorder or psychotic illness); history of current or previous substance misuse or significant head injury. Eighteen TRD participants were being treated with one or more anti-depressant medications at time of scanning (venlafaxine (6), sertraline (3), trazodone (3), citalopram (2), fluoxetine (2), isocarboxazid (2), mirtazapine (2), L-tryptophan (1), phenelzine (1), tranylcypromine (1)). In addition, 7 TRD participants were being treated with anti-psychotic medications (quetiapine (6) and chlorpromazine (1)). Three TRD participants were being treated with lithium augmentation.

Twenty-one healthy, never-depressed controls were recruited (mostly from partners, relatives and friends of TRD participants) and underwent clinical screening using the same semi-structured interview schedule and questionnaires. None of the controls had a history of current, or previous, psychiatric or neurological disorder, and all control participants were medication-free.

All participants had a predicted premorbid Full Scale Intelligence Quotient above 106 (one control was not assessed for IQ) as estimated by the National Adult Reading Test [23]. Handedness was assessed using the Edinburgh Handedness Inventory [24]. All subjects were right-handed with the exception of two left-handed participants in the control group and one and three ambidextrous participants in the control and patient groups respectively. Handedness was unknown in one participant from each group.

Image Acquisition

Structural whole-brain images were acquired using a 3T Siemens Magnetom TrioTim syngo scanner using a T1-weighted MP-RAGE (magnetisation-prepared rapid acquisition gradient echo) sequence with the following parameters: TR = 1900 ms, TE = 2.64 ms, flip angle = 9°, FOV = 200 mm, matrix = 256 x 256, 176 slices, voxel size 0.8x0.8x1 mm, slice thickness 1 mm.

Image Pre-processing

All scans were visually inspected for gross artefacts and particular care was taken to identify motion artefacts which appeared as blurring or ‘ghosting’ [25]. All scans were free from such artefacts and were included in the analysis.

Pre-processing was performed using SPM8 [26]. The procedure involved segmentation of T₁ weighted images into separate grey matter, white matter and CSF compartment images and normalisation of the grey matter segmented images towards to the default SPM8 anatomical template. The resultant images were smoothed with an 8 mm full-width at half-maximum (FWHM) Gaussian kernel.

Individual Subject Classification

Individual subject classification was implemented in Matlab (The Mathworks Inc.) using a SVM toolbox [27] with a Gaussian kernel [28] and custom Matlab scripts.

SVM analysis consisted of two stages: training the classifier, then testing the accuracy using data not used for training. Within-study replication (‘leave-one-out’ cross-validation, LOOCV) was used for training, meaning N-1 subjects are used as training data and the left out subject is predicted using the trained classifier, with all subjects used as testing data at some stage [29]. The SVM parameters were selected on the basis of training stage accuracy to avoid mixing prediction data with testing data which would result in inflated accuracy estimates.

Feature selection was used to identify brain regions supporting predictive classification. The feature selection method chosen during the SVM classification was a standard t-test, as implemented in the SPM toolbox. A t-test between the patient and control groups within each training set was performed during LOOCV (i.e. the subject being classified was not included in the t-test) with significance defined as p<0.05 at a whole brain, family-wise error corrected level [30]. The z-scores at each of the significant voxels were then ranked during LOOCV and the threshold, whereby voxels with z-scores above this threshold would be included in the classification, was optimised at the same stage as the SVM parameter selection. As the features selected differed slightly for each iteration of LOOCV, we report the sum of each feature selection map from each iteration, indicating which brain regions are most consistently selected for prediction of the ‘left out’ test dataset.

Group-level Differences

In order to compare the regions identified during the classification with a more conventional group-level analyses, we performed a standard t-test on all subjects, as implemented in the SPM toolbox. This comparison demonstrates that the feature selection method, which involves selecting a threshold during cross-validation, corresponds to a more stringent threshold than using a conventional t-test, but otherwise the results are very similar. For the conventional group-level VBM analysis, the null hypothesis of no difference in brain structure between patients and controls was tested using an unpaired t-test as implemented in SPM8. Significance was defined as p<0.05 at a whole brain, family-wise error corrected level, with the simultaneous requirements for voxel threshold and minimum cluster extent identified using a popular Monte-Carlo method [30].

Individual Subject MGH-S and Severity Score Predictions

Individual subject predictions of MGH-S and severity scores was implemented in Matlab (The Mathworks Inc.) using the Relevance Vector Regression (RVR) function within the PRoNTo toolbox [31, 32] and custom Matlab scripts. Similar to the SVM analysis, within-study replication (LOOCV) was used for training [29].

The RVR parameters were selected on the basis of a combination of three standard statistical variables: the root-mean square error (RMSE), the mean absolute error (MAE) and Pearson’s correlation coefficient (R) as calculated using the gfit2 toolbox [33]. The Shapiro-Wilk test was used to test whether each continuous variable was normally-distributed prior to using RVR (a requirement when using the Pearson correlation coefficient). The MGH-S and symptom severity scores all met this requirement.

The RVR predictions were performed on scans from TRD participants only, both with and without feature selection. The feature selection methods investigated and the corresponding results are reported in the S1 File.

Group-level Correlations

Similarly, the group-level regressions were performed as implemented in SPM8 using TRD participant data and both the MGH-S and symptom severity scores (according to the HDRS₁₇, MADRS and BDI-II, reported in the S1 File). Significance was defined as p<0.05 at a whole brain, family-wise error corrected level, with the simultaneous requirements for voxel threshold and minimum cluster extent identified using a popular Monte-Carlo method [30].

Results

Participant Characteristics

Age and estimated pre-morbid IQ (t-test, p>0.1, excluding the one control subject who failed to complete the testing) and gender distribution (as assessed by a chi-square calculation) did not differ significantly between groups. TRD participants mean ± SD age was 51.8 ± 11.2 years and mean ± SD estimated IQ was 122.8 ± 4.7. The control group mean ± SD age was 46.1 ± 14.0 years and the mean ± SD estimated IQ was 122.8 ± 5.8.

The average MGH-S score of the TRD participants was 13.3, which is consistent with a previous report of patients attending the Advanced Interventions Service (15.5) and should be noted as significantly higher than patients with depression treated in UK secondary care (5.3); and primary care (0.5) [34]. Indeed, this compares to the average MGH-S score for US patients enrolled in a speciality clinic, who had mean ± SD MGH-S scores of 1.6 ± 1.2 [35]. The average HDRS₁₇, MADRS and BDI illness severity rating scores in the TRD group were 16.1, 22.5 and 32.2 respectively, reflecting group-level depression severity at time of scanning to be in the mild-moderate range. These results are summarised in Table 1.

Download:

Table 1. Clinical descriptors for the TRD and healthy control groups in the structural MRI analysis.

Variables are shown as mean (standard deviation).

https://doi.org/10.1371/journal.pone.0132958.t001

Two TRD participants were in remission at the time of scanning (as assessed through the MINI PLUS interview schedule and clinical ratings). Using a previously described scheme for categorisation of severity [36], HDRS₁₇ scores were categorised into approximate levels of severity. Using this approach, one participant was classed in the “very severe” range, three in the “severe” range, ten in the “moderate” range, five in the “mild” range and one in remission at the time of scanning. Using alternate severity ranges as defined by Zimmerman, Martinez [37], two participants were classed in the “severe depression” range, seven in the “moderate” range, ten in the “mild” range and one in remission at the time of scanning. Using either categorisation, TRD participants had a wide range of symptom severity scores at the time of scanning.

Individual Subject Classification

A Gaussian SVM was used and feature selection was implemented using t-tests with a variable threshold that was optimised during cross-validation. The analysis using grey matter images resulted in an individual subject predictive accuracy of 85% (sensitivity 0.85, specificity 0.86, χ² = 17.7, p <0.0001).

Grey matter reductions in the caudate, insula, and periventricular grey matter supported individual prediction at an accuracy of 85%. Regions used in the classification are shown in Fig 1 and Table A in S1 File.

Download:

Fig 1. Feature selection (Gaussian SVM) identified brain regions in grey matter.

PV—periventricular grey matter; C—caudate; IN—insula.

https://doi.org/10.1371/journal.pone.0132958.g001

Group-level Differences

The brain regions identified using feature selection overlapped to a great extent with the results of the VBM group-level analysis (p<0.05, whole brain level significance) as t-tests were used in both cases (although the t-tests used during feature selection did not include the subject being classified and not all significant voxels from the t-tests were used in the prediction as it was thresholded). In the VBM analysis, mostly grey matter reductions were identified in the TRD participants, but a small number of increases were also found in the posterior cerebellum and middle frontal gyrus. As shown in Fig 2 and Table B in S1 File, the largest grey matter reductions were found in the caudate, insula, and periventricular grey matter. In addition, TRD patients also had significantly reduced grey matter volumes in a region identified by Lawson, Drevets [38] as corresponding to the habenula. There were no significant group differences detected in the hippocampi.

Download:

Fig 2. Group-level grey matter reductions in patients with TRD compared with healthy matched controls.

PV- periventricular grey matter, C—caudate reductions, H—habenula and IN—insula.

https://doi.org/10.1371/journal.pone.0132958.g002

Whilst the majority of the regions identified in the classification overlapped with the VBM results, a few voxels did not overlap either due to the thresholding of the t-tests during feature selection or differences in the t-test due to the left out subject during cross-validation. The overlap between the VBM results and the results from the feature selection is shown in Fig 3.

Download:

Fig 3. Overlapping grey matter regions between features selected during classification (purple/blue) and regions selected in the VBM analysis (red/purple).

https://doi.org/10.1371/journal.pone.0132958.g003

Individual Subject MGH-S and Severity Score Predictions

A linear kernel RVR was used to predict symptom severity scores (HDRS₁₇, MADRS and BDI) using the whole brain (no feature selection) MRI images from the TRD participants only. There was a significant correlation between the true and predicted HDRS₁₇ scores (RMSE = 4.6963, MAE = 3.6212, R = 0.50712, p = 0.02). A significant correlation was also identified between the true and predicted MADRS scores (RMSE = 6.8328, MAE = 5.441, R = 0.4822, p = 0.03). The best-fit line between true and predicted scores for both the HDRS₁₇ and MADRS predictions is shown in S1 Fig. and the highly distributed brain regions that provided the highest weightings are shown in S2 Fig. The most common regions to be used in both predictions were those along the cingulate gyri. However, the BDI scores could not be predicted in individual TRD participants using this approach. The results when feature selection was investigated is included in the S1 File.

Using the same method that was able to predict various symptom severity scores, it was not possible to predict MGH-S scores in individual subjects, irrespective of whether feature selection was used.

Group-level Correlations

Grey matter volume did not correlate strongly with previous medication exposure (as assessed through the MGH-S scores). Indeed, there were no positive correlations with MGH-S scores, but the insula and various small cortical regions showed a negative correlation (as shown in S8 and S9 Figs and Table C in S1 File). Hence, TRD participants with greater ratings of refractoriness to treatment showed reduced grey matter volume in these regions.

The insula was previously identified in the VBM analysis as one of the regions which was significantly reduced in volume in TRD participants compared with healthy controls. However, as MGH-S scores did not correlate with any of the symptom severity scores (HDRS₁₇, MADRS or BDI), it is reasonable that we did not find any overlap between brain regions that correlated with symptom severity (reported in the S1 File, S4–S7 Figs and Tables D-F in S1 File). Furthermore, adding MGH-S as a covariate in the HDRS₁₇ regression (and vice-versa) made no substantial alterations to the results.

Discussion

The high classification accuracy (85%) achieved in this study demonstrates that there are consistent patterns of structural brain abnormalities in patients with histories of TRD, including those with mild symptom burden, or in full clinical remission, at time of scanning when compared with healthy controls. These patterns were identified using feature selection and key regions included the caudate, the insula and periventricular grey matter—each of which have previously been reported to be reduced in previous group-level meta-analyses focusing on MDD [14–17]. The insula appears to represent an important brain region in TRD as it was found to correlate with MGH-S scores and was identified in the TRD vs. healthy control prediction. The insula has previously been implicated in the pathophysiology of depression as it has been associated with, among other things, emotional processing [39, 40] and human awareness and subjective experience [41]. As with our study, the volume of the insula was reduced in several other studies of patients with depression [20, 40, 42, 43]. Furthermore, these insula changes may be associated with clinical course and prognosis as Soriano-Mas and colleagues reported that insula volume was correlated with the number of relapses during a follow-up period and, along with the hippocampus and lateral parietal cortex, reduced volume of the left insula predicted a slower recovery [43]. The insula has also been reported to have decreased connectivity in TRD compared with treatment-sensitive depression [44].

The failure to accurately predict MGH-S scores in individual subjects may be due to the difficulties and methodological limitations inherent in quantifying the degree of treatment resistance. However, an alternative explanation is that there is no consistent association between brain structure and the number of trials of antidepressant medication without clinical response. The latter explanation is supported by the fact that so few brain regions correlated with MGH-S scores in the group-level regression analysis.

Hippocampal volume reductions have been reported in several TRD (and MDD) studies previously [9–12, 45]. Although no significant differences were identified in this region between TRD participants and healthy controls, hippocampal volume was found to negatively correlate with both the HDRS₁₇ and MADRS scores (see S1 File). Nonetheless, given the chronicity, severity and degree of treatment-refractoriness of the participants’ illnesses, the absence of hippocampal volumetric differences between the groups was somewhat surprising. The extensive treatment history all TRD patients, or inclusion of TRD participants with mild illness severity (or remission) at the time of scanning in the present study may account for the failure to replicate these between-group hippocampal volume reductions as a number of studies suggest a correlation between hippocampal volume and antidepressant treatment/decreasing depressive symptoms [46–50].

Caudate volume reductions have been reported in previous studies of depression [45, 51–54] and negative correlations between caudate volume and HDRS₁₇ scores have been reported previously [51, 55]. In addition, decreased connectivity between the precuneus/posterior cingulate cortex and the bilateral caudate has been reported in early depression [56]. Although some studies did not find significant differences in caudate volume between groups, there are various factors that could explain this discrepancy including, for example, younger study populations, lower illness severity and the exclusion of subjects with cerebrovascular risk factors [55, 57, 58]. It is particularly interesting that the dorsal striatum (including the caudate) and not the ventral striatum was found to have reduced volume in the TRD group since the dorsal striatum is generally implicated in learning stimulus-response associations whereas the ventral striatum is involved with reward and motivation, particularly the prediction of reward [59].

The decrease in periventricular grey matter is linked with ventricular enlargement. Visual inspection of the raw MR images showed that several participants had remarkably large ventricles, including a few controls. Ventricular enlargement is more commonly associated with aging but, as there were no significant differences in age between the groups, it is most likely due to MDD diagnosis. Elkis, Friedman [60] reported, through a meta-analysis, that ventricular volumes were larger in patients with mood disorders than healthy controls, but smaller than in patients with schizophrenia. However, a later review by Soares and Mann [61] could not find evidence for ventricular enlargement in mood disorders. In an early structural study by Andreasen, Olsen [62], patients with schizophrenia and ventricular enlargement typically had more “negative” symptoms, the symptoms more commonly associated with depression such as anhedonia, and, similarly, patients with schizophrenia and small ventricles had predominantly “positive” symptoms. It is, therefore, possible that ventricular enlargement may be more strongly associated with motor symptoms and loss of hedonic responsiveness. However this association requires further investigation as neither the “psychomotor retardation” question within the HDRS₁₇ score nor the “loss of pleasure” question within the BDI score correlated with CSF volume in this study.

Reduced habenula volume has also been reported, albeit inconsistently, in previous studies of patients with mood disorders [63, 64]. This reduction may be specific to mood disorder as reductions in this area have not been identified in studies of schizophrenia or post-traumatic stress disorder [63, 65]. Identification of a reduced habenular volume is interesting as it contains strong connections with both the limbic forebrain, consistently implicated in TRD studies, and the midbrain/brainstem and it is functionally associated with depression-relevant emotions (anxiety) and physiological responses such as those to stress and reward processing [63]. In addition, deep brain stimulation has been attempted previously in the lateral habenula in patients with treatment resistant depression [66].

Predictions of symptom severity in individual subjects have been performed previously in participants with current symptoms of depression [21] and other disorders such as OCD, Alzheimer’s disease and mild cognitive impairment [67–69]. Mwangi et al. [21] reported the ability to predict scores on the BDI in MDD participants using whole brain grey matter images. Using a similar approach, in this study we were able to predict HDRS₁₇ and MADRS scores. This may be because the participants in the present study were more treatment-resistant and had a wider range of symptom severity scores. In any event, both the HDRS₁₇ and MADRS scores, which are based upon clinician ratings of descriptions of symptoms over the preceding week, combined with some quantification of mental state findings, could be significantly predicted based upon a single grey matter MR image. When predicting symptom severity, adding feature selection to the process conferred little additional benefit. As some participants, despite having an extensive treatment history, were in remission at the time of scanning, it is a remarkable and novel finding that it was possible to accurately predict state measures of depression in individual subjects using structural measures.

Study Limitations

There are several study limitations to note. First, although the number of subjects in this study is comparable with similar studies, replication in a larger study population would be desirable. Second, the TRD participants were taking a range of antidepressant medications at the time of scanning and it remains unclear to what extent this may influence results. Sapolsky [70] has argued that it is unlikely that the often observed grey matter reductions in MDD are a consequence of antidepressant medication as there is evidence for the contrary, antidepressant-induced neurogenesis, with no clear rationale for reductions. We observed some small regions of increased grey matter in the posterior cerebellum and middle frontal gyrus and it is possible these represent medication effects that merit further investigation. Current medication status is also, theoretically, a potential confound when predicting symptom severity scores. However, Table 2 shows that there is no obvious link between current medication and symptom severity. Despite the TRD patients, by definition, having an extensive medication history, there were very few specific brain regions that correlated with the MGH-S scores. The insula was the only region of interest that was linked with medication exposure, implying that previous medication exposure is not strongly associated with widespread regional grey matter loss. This is despite the participants in this study having experienced far greater lifetime exposure to antidepressant medications (see MGH-S scores) than those in other studies. Finally, as the TRD group was recruited with a past or present diagnosis of chronic MDD, the differences between the two groups may not have been as distinct as other studies. However, this enhanced range of symptom severity may actually have been an advantage in the prediction of symptom severity.

Download:

Table 2. Treatment Resistance, State Illness Severity and Current Medication.

No patients had psychotic symptoms and quetiapine was prescribed as an augmentation agent for antidepressants [71], similar to the long established use of lithium, L-tryptophan and tri-iodothyronine in treatment resistant depression. No obvious relationships between current medication and treatment resistance/state illness severity were present. ‘mg’ indicates total dose per day, ‘mcg’ total micrograms per day.

https://doi.org/10.1371/journal.pone.0132958.t002

Conclusions

To summarise, it was possible to use grey matter MR images to predict diagnostic status when comparing TRD subjects and healthy controls. Although it was not possible to predict the level of treatment-resistance in individual TRD participants, the insula was identified as a region of interest. These results provide encouragement that machine learning methods can increase the understanding of the neurobiology of TRD.

Supporting Information

S1 File. Supporting Information and Severity Score Prediction Information.

https://doi.org/10.1371/journal.pone.0132958.s001

(DOC)

S1 Fig. The best fit lines for whole brain severity score predictions (left: HDRS₁₇ predictions, right: MADRS predictions).

https://doi.org/10.1371/journal.pone.0132958.s002

(TIF)

S2 Fig. The brain regions which were identified as the most predictive during the whole brain severity score predictions (left: HDRS₁₇ predictions, right: MADRS predictions).

https://doi.org/10.1371/journal.pone.0132958.s003

(TIF)

S3 Fig. The best fit line for the prediction of the BDI score using thresholded multiple linear regression feature selection and grey matter images.

https://doi.org/10.1371/journal.pone.0132958.s004

(TIF)

S4 Fig. Group-level positive correlations between grey matter and HAM_D (top) and MADRS (bottom).

https://doi.org/10.1371/journal.pone.0132958.s005

(TIF)

S5 Fig. Group-level negative correlations between grey matter and HAM_D.

https://doi.org/10.1371/journal.pone.0132958.s006

(TIF)

S6 Fig. Group-level negative correlations between grey matter and MADRS.

https://doi.org/10.1371/journal.pone.0132958.s007

(TIF)

S7 Fig. Group-level grey matter decreases in patients with TRD with increased BDI scores.

https://doi.org/10.1371/journal.pone.0132958.s008

(TIF)

S8 Fig. Group-level grey matter decreases in patients with TRD with increased MGH-S scores.

https://doi.org/10.1371/journal.pone.0132958.s009

(TIF)

S9 Fig. The best fit line for the correlation between the MGH-S scores and the volume of grey matter contained within a 5mm sphere centred within the mid-insulae (left (top), right (bottom)).

https://doi.org/10.1371/journal.pone.0132958.s010

(TIF)

S1 Data. Supplemental data.

https://doi.org/10.1371/journal.pone.0132958.s011

(ZIP)

Acknowledgments

We are very grateful to all the volunteers who took part in the study and to our research staff. Particular thanks go to Jennifer Macfarlane, Victoria Gradin, Mairi Stirling, Karen Walker, Christine Matthews and Craig Adam for assistance with recruitment, assessment, scanning and study administration.

Author Contributions

Conceived and designed the experiments: JDS KM. Performed the experiments: JDS KM DC. Analyzed the data: BAJ JDS ST. Wrote the paper: BAJ JDS ST DC KM.

References

1. Whiteford HA, Degenhardt L, Rehm J, Baxter AJ, Ferrari AJ, Erskine HE, et al. Global burden of disease attributable to mental and substance use disorders: findings from the Global Burden of Disease Study 2010. The Lancet. 2013;382(9904):1575–86.
- View Article
- Google Scholar
2. American Psychiatric Association. Diagnostic and statistical manual of mental disorders: DSM-IV-TR. Arlington, Virginia, U.S.A: American Psychiatric Publishing, Inc.; 2000.
3. American Psychiatric Association. Diagnostic and statistical manual of mental disorders: DSM-5. Arlington, Virginia, U.S.A: American Psychiatric Publishing, Inc.; 2013.
4. Rapaport MH, Schneider LS, Dunner DL, Davies JT, Pitts CD. Efficacy of controlled-release paroxetine in the treatment of late-life depression. The Journal of clinical psychiatry. 2003;64(9):1065–74. pmid:14628982
- View Article
- PubMed/NCBI
- Google Scholar
5. Trivedi MH, Rush AJ, Wisniewski SR, Nierenberg AA, Warden D, Ritz L, et al. Evaluation of outcomes with citalopram for depression using measurement-based care in STAR* D: implications for clinical practice. American journal of Psychiatry. 2006;163(1):28–40. pmid:16390886
- View Article
- PubMed/NCBI
- Google Scholar
6. Bennabi D, Aouizerate B, El-Hage W, Doumy O, Moliere F, Courtet P, et al. Risk factors for treatment resistance in unipolar depression: A systematic review. Journal of affective disorders. 2015;171:137–41. pmid:25305428
- View Article
- PubMed/NCBI
- Google Scholar
7. Fava M. Diagnosis and definition of treatment-resistant depression. Biological psychiatry. 2003;53(8):649–59. pmid:12706951
- View Article
- PubMed/NCBI
- Google Scholar
8. Sackheim H. The definition and meaning of treatment-resistant depression. Journal of Clinical Psychiatry. 2001;62(suppl 16):10–7. pmid:11480879
- View Article
- PubMed/NCBI
- Google Scholar
9. Shah P, Glabus M, Goodwin G, Ebmeier K. Chronic, treatment-resistant depression and right fronto-striatal atrophy. The British Journal of Psychiatry. 2002;180(5):434–40.
- View Article
- Google Scholar
10. Shah PJ, Ebmeier KP, Glabus MF, Goodwin GM. Cortical grey matter reductions associated with treatment-resistant chronic unipolar depression. Controlled magnetic resonance imaging study. The British journal of psychiatry. 1998;172(6):527–32.
- View Article
- Google Scholar
11. Maller JJ, Daskalakis ZJ, Thomson RH, Daigle M, Barr MS, Fitzgerald PB. Hippocampal volumetrics in treatment‐resistant depression and schizophrenia: The devil's in De‐Tail. Hippocampus. 2012;22(1):9–16. pmid:20882552
- View Article
- PubMed/NCBI
- Google Scholar
12. Furtado CP, Maller JJ, Fitzgerald PB. A magnetic resonance imaging study of the entorhinal cortex in treatment-resistant depression. Psychiatry Research: Neuroimaging. 2008;163(2):133–42. pmid:18511243
- View Article
- PubMed/NCBI
- Google Scholar
13. Haubold A, Peterson BS, Bansal R. Annual Research Review: Progress in using brain morphometry as a clinical tool for diagnosing psychiatric disorders. Journal of Child Psychology and Psychiatry. 2012;53(5):519–35. pmid:22394424
- View Article
- PubMed/NCBI
- Google Scholar
14. Koolschijn P, van Haren NEM, Lensvelt-Mulders GJLM, Pol H, Hilleke E, Kahn RS. Brain volume abnormalities in major depressive disorder: A meta-analysis of magnetic resonance imaging studies. Human brain mapping. 2009;30(11):3719–35. pmid:19441021
- View Article
- PubMed/NCBI
- Google Scholar
15. Fu CHY, Russell T, Murray R, Weinberger DR. Neuroimaging in psychiatry: Martin Dunitz; 2003.
- View Article
- Google Scholar
16. Bora E, Fornito A, Pantelis C, Yücel M. Gray matter abnormalities in major depressive disorder: a meta-analysis of voxel based morphometry studies. Journal of affective disorders. 2012;138(1):9–18.
- View Article
- Google Scholar
17. Kempton MJ, Salvador Z, Munafò MR, Geddes JR, Simmons A, Frangou S, et al. Structural neuroimaging studies in major depressive disorder: meta-analysis and comparison with bipolar disorder. Archives of general psychiatry. 2011;68(7):675. pmid:21727252
- View Article
- PubMed/NCBI
- Google Scholar
18. Kipli K, Kouzani AZ, Joordens M, editors. Evaluation of Feature Selection Algorithms for Detection of Depression from Brain sMRI Scans. Proceedings of 2013 ICME International Conference on Complex Medical Engineering; 2013; Beijing.
19. Costafreda SG, Chu C, Ashburner J, Fu CHY. Prognostic and Diagnostic Potential of the Structural Neuroanatomy of Depression. PLoS ONE. 2009;4(7):e6353. pmid:19633718
- View Article
- PubMed/NCBI
- Google Scholar
20. Mwangi B, Ebmeier KP, Matthews K, Douglas Steele J. Multi-centre diagnostic classification of individual structural neuroimaging scans from patients with major depressive disorder. Brain. 2012;135(5):1508–21. pmid:22544901.
- View Article
- PubMed/NCBI
- Google Scholar
21. Mwangi B, Matthews K, Steele JD. Prediction of illness severity in patients with major depression using structural MR brain scans. Journal Of Magnetic Resonance Imaging: JMRI. 2012;35(1):64–71. pmid:21959677.
- View Article
- PubMed/NCBI
- Google Scholar
22. Sheehan DV, Lecrubier Y. Mini-International Psychiatric Interview. MINI PLUS English Version. 1992;5(0).
- View Article
- Google Scholar
23. Nelson HE. National Adult Reading Test (NART): Test manual. Windsor, Berks: Nelson Publishing Company; 1982.
24. Oldfield RC. The assessment and analysis of handedness: The Edinburgh inventory. Neuropsychologia. 1971;9(1):97–113. pmid:5146491
- View Article
- PubMed/NCBI
- Google Scholar
25. McRobbie DW, Moore EA, Graves MJ, Prince MR. MRI from Picture to Proton. 2nd edition ed. New York: Cambridge University Press; 2010.
26. Ashburner J, Barnes G, Chen C-C, Daunizeau J, Flandin G, Friston K, et al. SPM8 Manual. 2012.
- View Article
- Google Scholar
27. Schwaighofer A. SVM toolbox. 2.51 ed. Retrieved from http://vision.ucsd.edu/~pdollar/toolbox/piotr_toolbox_V2.60.zip2001.
28. Bishop CM. Pattern recognition and machine learning. Jordan M, Kleinberg J, Scholkopf B, editors: Springer; 2006.
29. Johnston BA, Mwangi B, Matthews K, Coghill D, Steele JD. Predictive classification of individual magnetic resonance imaging scans from children and adolescents. European Child & Adolescent Psychiatry. 2013;22(12):733–44.
- View Article
- Google Scholar
30. Slotnick SD, Moo LR, Segal JB, Hart J Jr. Distinct prefrontal cortex activity associated with item memory and source memory for visual shapes. Brain Research Cognitive Brain Research. 2003;17(1):75–82. pmid:12763194.
- View Article
- PubMed/NCBI
- Google Scholar
31. Schrouff J, Rosa MJ, Rondina JM, Marquand AF, Chu C, Ashburner J, et al. PRoNTo: Pattern Recognition for Neuroimaging Toolbox. Neuroinformatics. 2013. pmid:23417655.
- View Article
- PubMed/NCBI
- Google Scholar
32. Tipping ME. Sparse bayesian learning and the relevance vector machine. J Mach Learn Res. 2001;1:211–44.
- View Article
- Google Scholar
33. Crozier R. Goodness of Fit (gfit2) http://www.mathworks.com/matlabcentral/fileexchange/22020-goodness-of-fit-modified/content/gfit2.m2009.
34. Hazari H, Christmas D, Matthews K. The clinical utility of different quantitative methods for measuring treatment resistance in major depression. Journal of affective disorders. 2013;150(2):231–6. pmid:23668902
- View Article
- PubMed/NCBI
- Google Scholar
35. Petersen T, Papakostas GI, Posternak MA, Kant A, Guyker WM, Iosifescu DV, et al. Empirical testing of two models for staging antidepressant treatment resistance. Journal of clinical psychopharmacology. 2005;25(4):336–41. pmid:16012276
- View Article
- PubMed/NCBI
- Google Scholar
36. Rush AJ, Trivedi MH, Ibrahim HM, Carmody TJ, Arnow B, Klein DN, et al. The 16-Item quick inventory of depressive symptomatology (QIDS), clinician rating (QIDS-C), and self-report (QIDS-SR): a psychometric evaluation in patients with chronic major depression. Biological Psychiatry. 2003;54(5):573–83. http://dx.doi.org/10.1016/S0006-3223(02)01866-8. pmid:12946886
- View Article
- PubMed/NCBI
- Google Scholar
37. Zimmerman M, Martinez JH, Young D, Chelminski I, Dalrymple K. Severity classification on the Hamilton depression rating scale. Journal of affective disorders. 2013;150(2):384–8. pmid:23759278
- View Article
- PubMed/NCBI
- Google Scholar
38. Lawson RP, Drevets WC, Roiser JP. Defining the habenula in human neuroimaging studies. Neuroimage. 2013;64:722–7. pmid:22986224
- View Article
- PubMed/NCBI
- Google Scholar
39. Sprengelmeyer R, Steele JD, Mwangi B, Kumar P, Christmas D, Milders M, et al. The insular cortex and the neuroanatomy of major depression. Journal of affective disorders. 2011;133(1):120–7.
- View Article
- Google Scholar
40. Takahashi T, Yücel M, Lorenzetti V, Tanino R, Whittle S, Suzuki M, et al. Volumetric MRI study of the insular cortex in individuals with current and past major depression. Journal of affective disorders. 2010;121(3):231–8. pmid:19540599
- View Article
- PubMed/NCBI
- Google Scholar
41. Craig A. How do you feel—now? the anterior insula and human awareness. 2009.
- View Article
- Google Scholar
42. Hwang J-P, Lee T-W, Tsai S-J, Chen T-J, Yang C-H, Lirng J-F, et al. Cortical and subcortical abnormalities in late-onset depression with history of suicide attempts investigated with MRI and voxel-based morphometry. Journal of geriatric psychiatry and neurology. 2010;23(3):171–84. pmid:20430976
- View Article
- PubMed/NCBI
- Google Scholar
43. Soriano-Mas C, Hernández-Ribas R, Pujol J, Urretavizcaya M, Deus J, Harrison BJ, et al. Cross-sectional and longitudinal assessment of structural brain alterations in melancholic depression. Biological psychiatry. 2011;69(4):318–25. pmid:20875637
- View Article
- PubMed/NCBI
- Google Scholar
44. Lui S, Wu Q, Qiu L, Yang X, Kuang W, Chan RC, et al. Resting-state functional connectivity in treatment-resistant depression. American Journal of Psychiatry. 2011;168(6):642–8. pmid:21362744
- View Article
- PubMed/NCBI
- Google Scholar
45. Redlich R, Almeida JR, Grotegerd D, Opel N, Kugel H, Heindel W, et al. Brain Morphometric Biomarkers Distinguishing Unipolar and Bipolar Depression: A Voxel-Based Morphometry—Pattern Classification Approach. JAMA psychiatry. 2014;71(11):1222–30. pmid:25188810
- View Article
- PubMed/NCBI
- Google Scholar
46. Schermuly I, Wolf D, Lieb K, Stoeter P, Fellgiebel A. State dependent posterior hippocampal volume increases in patients with major depressive disorder. Journal of affective disorders. 2011;135(1):405–9.
- View Article
- Google Scholar
47. Moore GJ, Bebchuk JM, Wilds IB, Chen G, Menji HK. Lithium-induced increase in human brain grey matter. The Lancet. 2000;356(9237):1241–2.
- View Article
- Google Scholar
48. Santarelli L, Saxe M, Gross C, Surget A, Battaglia F, Dulawa S, et al. Requirement of hippocampal neurogenesis for the behavioral effects of antidepressants. Science. 2003;301(5634):805–9. pmid:12907793
- View Article
- PubMed/NCBI
- Google Scholar
49. Sahay A, Hen R. Adult hippocampal neurogenesis in depression. Nature neuroscience. 2007;10(9):1110–5. pmid:17726477
- View Article
- PubMed/NCBI
- Google Scholar
50. Frodl T, Jäger M, Smajstrlova I, Born C, Bottlender R, Palladino T, et al. Effect of hippocampal and amygdala volumes on clinical outcomes in major depression: a 3-year prospective magnetic resonance imaging study. Journal of psychiatry & neuroscience: JPN. 2008;33(5):423.
- View Article
- Google Scholar
51. Butters MA, Aizenstein HJ, Hayashi KM, Meltzer CC, Seaman J, Reynolds Iii CF, et al. Three-Dimensional Surface Mapping of the Caudate Nucleus in Late-Life Depression. The American Journal of Geriatric Psychiatry. 2009;17(1):4–12. http://dx.doi.org/10.1097/JGP.0b013e31816ff72b. pmid:18790876
- View Article
- PubMed/NCBI
- Google Scholar
52. Kim MJ, Hamilton JP, Gotlib IH. Reduced caudate gray matter volume in women with major depressive disorder. Psychiatry Research: Neuroimaging. 2008;164(2):114–22. http://dx.doi.org/10.1016/j.pscychresns.2007.12.020. pmid:18930633
- View Article
- PubMed/NCBI
- Google Scholar
53. Krishnan K, McDonald WM, Escalona PR, Doraiswamy PM, Na C, Husain MM, et al. Magnetic resonance imaging of the caudate nuclei in depression: preliminary observations. Archives of general psychiatry. 1992;49(7):553. pmid:1627046
- View Article
- PubMed/NCBI
- Google Scholar
54. Parashos IA, Tupler LA, Blitchington T, Krishnan KRR. Magnetic-resonance morphometry in patients with major depression. Psychiatry Research: Neuroimaging. 1998;84(1):7–15. http://dx.doi.org/10.1016/S0925-4927(98)00042-0. pmid:9870413
- View Article
- PubMed/NCBI
- Google Scholar
55. Pillay SS, Renshaw PF, Bonello CM, Lafer B, Fava M, Yurgelun-Todd D. A quantitative magnetic resonance imaging study of caudate and lenticular nucleus gray matter volume in primary unipolar major depression: relationship to treatment response and clinical severity. Psychiatry Research: Neuroimaging. 1998;84(2–3):61–74. http://dx.doi.org/10.1016/S0925-4927(98)00048-1. pmid:10710164
- View Article
- PubMed/NCBI
- Google Scholar
56. Bluhm R, Williamson P, Lanius R, Théberge J, Densmore M, Bartha R, et al. Resting state default-mode network connectivity in early depression using a seed region-of-interest analysis: Decreased connectivity with caudate nucleus. Psychiatry and Clinical Neurosciences. 2009;63(6):754–61. pmid:20021629
- View Article
- PubMed/NCBI
- Google Scholar
57. Lacerda ALT, Nicoletti MA, Brambilla P, Sassi RB, Mallinger AG, Frank E, et al. Anatomical MRI study of basal ganglia in major depressive disorder. Psychiatry Research: Neuroimaging. 2003;124(3):129–40. http://dx.doi.org/10.1016/S0925-4927(03)00123-9. pmid:14623065
- View Article
- PubMed/NCBI
- Google Scholar
58. Lenze EJ, Sheline YI. Absence of striatal volume differences between depressed subjects with no comorbid medical illness and matched comparison subjects. American Journal of Psychiatry. 1999;156(12):1989–91. pmid:10588417
- View Article
- PubMed/NCBI
- Google Scholar
59. O'Doherty J, Dayan P, Schultz J, Deichmann R, Friston K, Dolan RJ. Dissociable roles of ventral and dorsal striatum in instrumental conditioning. Science. 2004;304(5669):452–4. pmid:15087550
- View Article
- PubMed/NCBI
- Google Scholar
60. Elkis H, Friedman L, Wise A, Meltzer HY. Meta-analyses of studies of ventricular enlargement and cortical sulcal prominence in mood disorders: comparisons with controls or patients with schizophrenia. Archives of General Psychiatry. 1995;52(9):735. pmid:7654125
- View Article
- PubMed/NCBI
- Google Scholar
61. Soares JC, Mann JJ. The anatomy of mood disorders—review of structural neuroimaging studies. Biological psychiatry. 1997;41(1):86–106. pmid:8988799
- View Article
- PubMed/NCBI
- Google Scholar
62. Andreasen NC, Olsen SA, Dennert JW, Smith MR. Ventricular enlargement in schizophrenia: Relationship to positive and negative symptoms. Am J Psychiatry. 1982;139(3):297–302. pmid:7058942
- View Article
- PubMed/NCBI
- Google Scholar
63. Ranft K, Dobrowolny H, Krell D, Bielau H, Bogerts B, Bernstein H-G. Evidence for structural abnormalities of the human habenular complex in affective disorders but not in schizophrenia. Psychological Medicine. 2010;40(04):557–67.
- View Article
- Google Scholar
64. Savitz JB, Nugent AC, Bogers W, Roiser JP, Bain EE, Neumeister A, et al. Habenula Volume in Bipolar Disorder and Major Depressive Disorder: A High-Resolution Magnetic Resonance Imaging Study. Biological Psychiatry. 2011;69(4):336–43. http://dx.doi.org/10.1016/j.biopsych.2010.09.027. pmid:21094939
- View Article
- PubMed/NCBI
- Google Scholar
65. Savitz JB, Bonne O, Nugent AC, Vythilingam M, Bogers W, Charney DS, et al. Habenula volume in post-traumatic stress disorder measured with high-resolution MRI. Biol Mood Anxiety Disord. 2011;1:7. pmid:22738208
- View Article
- PubMed/NCBI
- Google Scholar
66. Sartorius A, Henn FA. Deep brain stimulation of the lateral habenula in treatment resistant major depression. Medical hypotheses. 2007;69(6):1305–8. pmid:17498883
- View Article
- PubMed/NCBI
- Google Scholar
67. Hoexter MQ, Miguel EC, Diniz JB, Shavitt RG, Busatto GF, Sato JR. Predicting obsessive—compulsive disorder severity combining neuroimaging and machine learning methods. Journal of Affective Disorders. 2013;150(3):1213–6. http://dx.doi.org/10.1016/j.jad.2013.05.041. pmid:23769292
- View Article
- PubMed/NCBI
- Google Scholar
68. Wang Y, Fan Y, Bhatt P, Davatzikos C. High-dimensional pattern regression using machine learning: from medical images to continuous clinical variables. Neuroimage. 2010;50(4):1519–35. pmid:20056158.
- View Article
- PubMed/NCBI
- Google Scholar
69. Stonnington CM, Chu C, Klöppel S, Jack CR Jr., Ashburner J, Frackowiak RSJ. Predicting clinical scores from magnetic resonance scans in Alzheimer's disease. Neuroimage. 2010;51(4):1405–13. pmid:20347044.
- View Article
- PubMed/NCBI
- Google Scholar
70. Sapolsky RM. Depression, antidepressants, and the shrinking hippocampus. Proceedings of the National Academy of Sciences. 2001;98(22):12320–2.
- View Article
- Google Scholar
71. Dorée J-P, Rosiers JD, Lew V, Gendron A, Elie R, Stip E, et al. Quetiapine augmentation of treatment-resistant depression: a comparison with lithium. Current Medical Research and Opinion. 2007;23(2):333–41. pmid:17288688
- View Article
- PubMed/NCBI
- Google Scholar

[ref1] 1. Whiteford HA, Degenhardt L, Rehm J, Baxter AJ, Ferrari AJ, Erskine HE, et al. Global burden of disease attributable to mental and substance use disorders: findings from the Global Burden of Disease Study 2010. The Lancet. 2013;382(9904):1575–86.
View Article
Google Scholar

[2] View Article

[3] Google Scholar

[ref2] 2. American Psychiatric Association. Diagnostic and statistical manual of mental disorders: DSM-IV-TR. Arlington, Virginia, U.S.A: American Psychiatric Publishing, Inc.; 2000.

[ref3] 3. American Psychiatric Association. Diagnostic and statistical manual of mental disorders: DSM-5. Arlington, Virginia, U.S.A: American Psychiatric Publishing, Inc.; 2013.

[ref4] 4. Rapaport MH, Schneider LS, Dunner DL, Davies JT, Pitts CD. Efficacy of controlled-release paroxetine in the treatment of late-life depression. The Journal of clinical psychiatry. 2003;64(9):1065–74. pmid:14628982
View Article
PubMed/NCBI
Google Scholar

[7] View Article

[8] PubMed/NCBI

[9] Google Scholar

[ref5] 5. Trivedi MH, Rush AJ, Wisniewski SR, Nierenberg AA, Warden D, Ritz L, et al. Evaluation of outcomes with citalopram for depression using measurement-based care in STAR* D: implications for clinical practice. American journal of Psychiatry. 2006;163(1):28–40. pmid:16390886
View Article
PubMed/NCBI
Google Scholar

[11] View Article

[12] PubMed/NCBI

[13] Google Scholar

[ref6] 6. Bennabi D, Aouizerate B, El-Hage W, Doumy O, Moliere F, Courtet P, et al. Risk factors for treatment resistance in unipolar depression: A systematic review. Journal of affective disorders. 2015;171:137–41. pmid:25305428
View Article
PubMed/NCBI
Google Scholar

[15] View Article

[16] PubMed/NCBI

[17] Google Scholar

[ref7] 7. Fava M. Diagnosis and definition of treatment-resistant depression. Biological psychiatry. 2003;53(8):649–59. pmid:12706951
View Article
PubMed/NCBI
Google Scholar

[19] View Article

[20] PubMed/NCBI

[21] Google Scholar

[ref8] 8. Sackheim H. The definition and meaning of treatment-resistant depression. Journal of Clinical Psychiatry. 2001;62(suppl 16):10–7. pmid:11480879
View Article
PubMed/NCBI
Google Scholar

[23] View Article

[24] PubMed/NCBI

[25] Google Scholar

[ref9] 9. Shah P, Glabus M, Goodwin G, Ebmeier K. Chronic, treatment-resistant depression and right fronto-striatal atrophy. The British Journal of Psychiatry. 2002;180(5):434–40.
View Article
Google Scholar

[27] View Article

[28] Google Scholar

[ref10] 10. Shah PJ, Ebmeier KP, Glabus MF, Goodwin GM. Cortical grey matter reductions associated with treatment-resistant chronic unipolar depression. Controlled magnetic resonance imaging study. The British journal of psychiatry. 1998;172(6):527–32.
View Article
Google Scholar

[30] View Article

[31] Google Scholar

[ref11] 11. Maller JJ, Daskalakis ZJ, Thomson RH, Daigle M, Barr MS, Fitzgerald PB. Hippocampal volumetrics in treatment‐resistant depression and schizophrenia: The devil's in De‐Tail. Hippocampus. 2012;22(1):9–16. pmid:20882552
View Article
PubMed/NCBI
Google Scholar

[33] View Article

[34] PubMed/NCBI

[35] Google Scholar

[ref12] 12. Furtado CP, Maller JJ, Fitzgerald PB. A magnetic resonance imaging study of the entorhinal cortex in treatment-resistant depression. Psychiatry Research: Neuroimaging. 2008;163(2):133–42. pmid:18511243
View Article
PubMed/NCBI
Google Scholar

[37] View Article

[38] PubMed/NCBI

[39] Google Scholar

[ref13] 13. Haubold A, Peterson BS, Bansal R. Annual Research Review: Progress in using brain morphometry as a clinical tool for diagnosing psychiatric disorders. Journal of Child Psychology and Psychiatry. 2012;53(5):519–35. pmid:22394424
View Article
PubMed/NCBI
Google Scholar

[41] View Article

[42] PubMed/NCBI

[43] Google Scholar

[ref14] 14. Koolschijn P, van Haren NEM, Lensvelt-Mulders GJLM, Pol H, Hilleke E, Kahn RS. Brain volume abnormalities in major depressive disorder: A meta-analysis of magnetic resonance imaging studies. Human brain mapping. 2009;30(11):3719–35. pmid:19441021
View Article
PubMed/NCBI
Google Scholar

[45] View Article

[46] PubMed/NCBI

[47] Google Scholar

[ref15] 15. Fu CHY, Russell T, Murray R, Weinberger DR. Neuroimaging in psychiatry: Martin Dunitz; 2003.
View Article
Google Scholar

[49] View Article

[50] Google Scholar

[ref16] 16. Bora E, Fornito A, Pantelis C, Yücel M. Gray matter abnormalities in major depressive disorder: a meta-analysis of voxel based morphometry studies. Journal of affective disorders. 2012;138(1):9–18.
View Article
Google Scholar

[52] View Article

[53] Google Scholar

[ref17] 17. Kempton MJ, Salvador Z, Munafò MR, Geddes JR, Simmons A, Frangou S, et al. Structural neuroimaging studies in major depressive disorder: meta-analysis and comparison with bipolar disorder. Archives of general psychiatry. 2011;68(7):675. pmid:21727252
View Article
PubMed/NCBI
Google Scholar

[55] View Article

[56] PubMed/NCBI

[57] Google Scholar

[ref18] 18. Kipli K, Kouzani AZ, Joordens M, editors. Evaluation of Feature Selection Algorithms for Detection of Depression from Brain sMRI Scans. Proceedings of 2013 ICME International Conference on Complex Medical Engineering; 2013; Beijing.

[ref19] 19. Costafreda SG, Chu C, Ashburner J, Fu CHY. Prognostic and Diagnostic Potential of the Structural Neuroanatomy of Depression. PLoS ONE. 2009;4(7):e6353. pmid:19633718
View Article
PubMed/NCBI
Google Scholar

[60] View Article

[61] PubMed/NCBI

[62] Google Scholar

[ref20] 20. Mwangi B, Ebmeier KP, Matthews K, Douglas Steele J. Multi-centre diagnostic classification of individual structural neuroimaging scans from patients with major depressive disorder. Brain. 2012;135(5):1508–21. pmid:22544901.
View Article
PubMed/NCBI
Google Scholar

[64] View Article

[65] PubMed/NCBI

[66] Google Scholar

[ref21] 21. Mwangi B, Matthews K, Steele JD. Prediction of illness severity in patients with major depression using structural MR brain scans. Journal Of Magnetic Resonance Imaging: JMRI. 2012;35(1):64–71. pmid:21959677.
View Article
PubMed/NCBI
Google Scholar

[68] View Article

[69] PubMed/NCBI

[70] Google Scholar

[ref22] 22. Sheehan DV, Lecrubier Y. Mini-International Psychiatric Interview. MINI PLUS English Version. 1992;5(0).
View Article
Google Scholar

[72] View Article

[73] Google Scholar

[ref23] 23. Nelson HE. National Adult Reading Test (NART): Test manual. Windsor, Berks: Nelson Publishing Company; 1982.

[ref24] 24. Oldfield RC. The assessment and analysis of handedness: The Edinburgh inventory. Neuropsychologia. 1971;9(1):97–113. pmid:5146491
View Article
PubMed/NCBI
Google Scholar

[76] View Article

[77] PubMed/NCBI

[78] Google Scholar

[ref25] 25. McRobbie DW, Moore EA, Graves MJ, Prince MR. MRI from Picture to Proton. 2nd edition ed. New York: Cambridge University Press; 2010.

[ref26] 26. Ashburner J, Barnes G, Chen C-C, Daunizeau J, Flandin G, Friston K, et al. SPM8 Manual. 2012.
View Article
Google Scholar

[81] View Article

[82] Google Scholar

[ref27] 27. Schwaighofer A. SVM toolbox. 2.51 ed. Retrieved from http://vision.ucsd.edu/~pdollar/toolbox/piotr_toolbox_V2.60.zip2001.

[ref28] 28. Bishop CM. Pattern recognition and machine learning. Jordan M, Kleinberg J, Scholkopf B, editors: Springer; 2006.

[ref29] 29. Johnston BA, Mwangi B, Matthews K, Coghill D, Steele JD. Predictive classification of individual magnetic resonance imaging scans from children and adolescents. European Child & Adolescent Psychiatry. 2013;22(12):733–44.
View Article
Google Scholar

[86] View Article

[87] Google Scholar

[ref30] 30. Slotnick SD, Moo LR, Segal JB, Hart J Jr. Distinct prefrontal cortex activity associated with item memory and source memory for visual shapes. Brain Research Cognitive Brain Research. 2003;17(1):75–82. pmid:12763194.
View Article
PubMed/NCBI
Google Scholar

[89] View Article

[90] PubMed/NCBI

[91] Google Scholar

[ref31] 31. Schrouff J, Rosa MJ, Rondina JM, Marquand AF, Chu C, Ashburner J, et al. PRoNTo: Pattern Recognition for Neuroimaging Toolbox. Neuroinformatics. 2013. pmid:23417655.
View Article
PubMed/NCBI
Google Scholar

[93] View Article

[94] PubMed/NCBI

[95] Google Scholar

[ref32] 32. Tipping ME. Sparse bayesian learning and the relevance vector machine. J Mach Learn Res. 2001;1:211–44.
View Article
Google Scholar

[97] View Article

[98] Google Scholar

[ref33] 33. Crozier R. Goodness of Fit (gfit2) http://www.mathworks.com/matlabcentral/fileexchange/22020-goodness-of-fit-modified/content/gfit2.m2009.

[ref34] 34. Hazari H, Christmas D, Matthews K. The clinical utility of different quantitative methods for measuring treatment resistance in major depression. Journal of affective disorders. 2013;150(2):231–6. pmid:23668902
View Article
PubMed/NCBI
Google Scholar

[101] View Article

[102] PubMed/NCBI

[103] Google Scholar

[ref35] 35. Petersen T, Papakostas GI, Posternak MA, Kant A, Guyker WM, Iosifescu DV, et al. Empirical testing of two models for staging antidepressant treatment resistance. Journal of clinical psychopharmacology. 2005;25(4):336–41. pmid:16012276
View Article
PubMed/NCBI
Google Scholar

[105] View Article

[106] PubMed/NCBI

[107] Google Scholar

[ref36] 36. Rush AJ, Trivedi MH, Ibrahim HM, Carmody TJ, Arnow B, Klein DN, et al. The 16-Item quick inventory of depressive symptomatology (QIDS), clinician rating (QIDS-C), and self-report (QIDS-SR): a psychometric evaluation in patients with chronic major depression. Biological Psychiatry. 2003;54(5):573–83. http://dx.doi.org/10.1016/S0006-3223(02)01866-8. pmid:12946886
View Article
PubMed/NCBI
Google Scholar

[109] View Article

[110] PubMed/NCBI

[111] Google Scholar

[ref37] 37. Zimmerman M, Martinez JH, Young D, Chelminski I, Dalrymple K. Severity classification on the Hamilton depression rating scale. Journal of affective disorders. 2013;150(2):384–8. pmid:23759278
View Article
PubMed/NCBI
Google Scholar

[113] View Article

[114] PubMed/NCBI

[115] Google Scholar

[ref38] 38. Lawson RP, Drevets WC, Roiser JP. Defining the habenula in human neuroimaging studies. Neuroimage. 2013;64:722–7. pmid:22986224
View Article
PubMed/NCBI
Google Scholar

[117] View Article

[118] PubMed/NCBI

[119] Google Scholar

[ref39] 39. Sprengelmeyer R, Steele JD, Mwangi B, Kumar P, Christmas D, Milders M, et al. The insular cortex and the neuroanatomy of major depression. Journal of affective disorders. 2011;133(1):120–7.
View Article
Google Scholar

[121] View Article

[122] Google Scholar

[ref40] 40. Takahashi T, Yücel M, Lorenzetti V, Tanino R, Whittle S, Suzuki M, et al. Volumetric MRI study of the insular cortex in individuals with current and past major depression. Journal of affective disorders. 2010;121(3):231–8. pmid:19540599
View Article
PubMed/NCBI
Google Scholar

[124] View Article

[125] PubMed/NCBI

[126] Google Scholar

[ref41] 41. Craig A. How do you feel—now? the anterior insula and human awareness. 2009.
View Article
Google Scholar

[128] View Article

[129] Google Scholar

[ref42] 42. Hwang J-P, Lee T-W, Tsai S-J, Chen T-J, Yang C-H, Lirng J-F, et al. Cortical and subcortical abnormalities in late-onset depression with history of suicide attempts investigated with MRI and voxel-based morphometry. Journal of geriatric psychiatry and neurology. 2010;23(3):171–84. pmid:20430976
View Article
PubMed/NCBI
Google Scholar

[131] View Article

[132] PubMed/NCBI

[133] Google Scholar

[ref43] 43. Soriano-Mas C, Hernández-Ribas R, Pujol J, Urretavizcaya M, Deus J, Harrison BJ, et al. Cross-sectional and longitudinal assessment of structural brain alterations in melancholic depression. Biological psychiatry. 2011;69(4):318–25. pmid:20875637
View Article
PubMed/NCBI
Google Scholar

[135] View Article

[136] PubMed/NCBI

[137] Google Scholar

[ref44] 44. Lui S, Wu Q, Qiu L, Yang X, Kuang W, Chan RC, et al. Resting-state functional connectivity in treatment-resistant depression. American Journal of Psychiatry. 2011;168(6):642–8. pmid:21362744
View Article
PubMed/NCBI
Google Scholar

[139] View Article

[140] PubMed/NCBI

[141] Google Scholar

[ref45] 45. Redlich R, Almeida JR, Grotegerd D, Opel N, Kugel H, Heindel W, et al. Brain Morphometric Biomarkers Distinguishing Unipolar and Bipolar Depression: A Voxel-Based Morphometry—Pattern Classification Approach. JAMA psychiatry. 2014;71(11):1222–30. pmid:25188810
View Article
PubMed/NCBI
Google Scholar

[143] View Article

[144] PubMed/NCBI

[145] Google Scholar

[ref46] 46. Schermuly I, Wolf D, Lieb K, Stoeter P, Fellgiebel A. State dependent posterior hippocampal volume increases in patients with major depressive disorder. Journal of affective disorders. 2011;135(1):405–9.
View Article
Google Scholar

[147] View Article

[148] Google Scholar

[ref47] 47. Moore GJ, Bebchuk JM, Wilds IB, Chen G, Menji HK. Lithium-induced increase in human brain grey matter. The Lancet. 2000;356(9237):1241–2.
View Article
Google Scholar

[150] View Article

[151] Google Scholar

[ref48] 48. Santarelli L, Saxe M, Gross C, Surget A, Battaglia F, Dulawa S, et al. Requirement of hippocampal neurogenesis for the behavioral effects of antidepressants. Science. 2003;301(5634):805–9. pmid:12907793
View Article
PubMed/NCBI
Google Scholar

[153] View Article

[154] PubMed/NCBI

[155] Google Scholar

[ref49] 49. Sahay A, Hen R. Adult hippocampal neurogenesis in depression. Nature neuroscience. 2007;10(9):1110–5. pmid:17726477
View Article
PubMed/NCBI
Google Scholar

[157] View Article

[158] PubMed/NCBI

[159] Google Scholar

[ref50] 50. Frodl T, Jäger M, Smajstrlova I, Born C, Bottlender R, Palladino T, et al. Effect of hippocampal and amygdala volumes on clinical outcomes in major depression: a 3-year prospective magnetic resonance imaging study. Journal of psychiatry & neuroscience: JPN. 2008;33(5):423.
View Article
Google Scholar

[161] View Article

[162] Google Scholar

[ref51] 51. Butters MA, Aizenstein HJ, Hayashi KM, Meltzer CC, Seaman J, Reynolds Iii CF, et al. Three-Dimensional Surface Mapping of the Caudate Nucleus in Late-Life Depression. The American Journal of Geriatric Psychiatry. 2009;17(1):4–12. http://dx.doi.org/10.1097/JGP.0b013e31816ff72b. pmid:18790876
View Article
PubMed/NCBI
Google Scholar

[164] View Article

[165] PubMed/NCBI

[166] Google Scholar

[ref52] 52. Kim MJ, Hamilton JP, Gotlib IH. Reduced caudate gray matter volume in women with major depressive disorder. Psychiatry Research: Neuroimaging. 2008;164(2):114–22. http://dx.doi.org/10.1016/j.pscychresns.2007.12.020. pmid:18930633
View Article
PubMed/NCBI
Google Scholar

[168] View Article

[169] PubMed/NCBI

[170] Google Scholar

[ref53] 53. Krishnan K, McDonald WM, Escalona PR, Doraiswamy PM, Na C, Husain MM, et al. Magnetic resonance imaging of the caudate nuclei in depression: preliminary observations. Archives of general psychiatry. 1992;49(7):553. pmid:1627046
View Article
PubMed/NCBI
Google Scholar

[172] View Article

[173] PubMed/NCBI

[174] Google Scholar

[ref54] 54. Parashos IA, Tupler LA, Blitchington T, Krishnan KRR. Magnetic-resonance morphometry in patients with major depression. Psychiatry Research: Neuroimaging. 1998;84(1):7–15. http://dx.doi.org/10.1016/S0925-4927(98)00042-0. pmid:9870413
View Article
PubMed/NCBI
Google Scholar

[176] View Article

[177] PubMed/NCBI

[178] Google Scholar

[ref55] 55. Pillay SS, Renshaw PF, Bonello CM, Lafer B, Fava M, Yurgelun-Todd D. A quantitative magnetic resonance imaging study of caudate and lenticular nucleus gray matter volume in primary unipolar major depression: relationship to treatment response and clinical severity. Psychiatry Research: Neuroimaging. 1998;84(2–3):61–74. http://dx.doi.org/10.1016/S0925-4927(98)00048-1. pmid:10710164
View Article
PubMed/NCBI
Google Scholar

[180] View Article

[181] PubMed/NCBI

[182] Google Scholar

[ref56] 56. Bluhm R, Williamson P, Lanius R, Théberge J, Densmore M, Bartha R, et al. Resting state default-mode network connectivity in early depression using a seed region-of-interest analysis: Decreased connectivity with caudate nucleus. Psychiatry and Clinical Neurosciences. 2009;63(6):754–61. pmid:20021629
View Article
PubMed/NCBI
Google Scholar

[184] View Article

[185] PubMed/NCBI

[186] Google Scholar

[ref57] 57. Lacerda ALT, Nicoletti MA, Brambilla P, Sassi RB, Mallinger AG, Frank E, et al. Anatomical MRI study of basal ganglia in major depressive disorder. Psychiatry Research: Neuroimaging. 2003;124(3):129–40. http://dx.doi.org/10.1016/S0925-4927(03)00123-9. pmid:14623065
View Article
PubMed/NCBI
Google Scholar

[188] View Article

[189] PubMed/NCBI

[190] Google Scholar

[ref58] 58. Lenze EJ, Sheline YI. Absence of striatal volume differences between depressed subjects with no comorbid medical illness and matched comparison subjects. American Journal of Psychiatry. 1999;156(12):1989–91. pmid:10588417
View Article
PubMed/NCBI
Google Scholar

[192] View Article

[193] PubMed/NCBI

[194] Google Scholar

[ref59] 59. O'Doherty J, Dayan P, Schultz J, Deichmann R, Friston K, Dolan RJ. Dissociable roles of ventral and dorsal striatum in instrumental conditioning. Science. 2004;304(5669):452–4. pmid:15087550
View Article
PubMed/NCBI
Google Scholar

[196] View Article

[197] PubMed/NCBI

[198] Google Scholar

[ref60] 60. Elkis H, Friedman L, Wise A, Meltzer HY. Meta-analyses of studies of ventricular enlargement and cortical sulcal prominence in mood disorders: comparisons with controls or patients with schizophrenia. Archives of General Psychiatry. 1995;52(9):735. pmid:7654125
View Article
PubMed/NCBI
Google Scholar

[200] View Article

[201] PubMed/NCBI

[202] Google Scholar

[ref61] 61. Soares JC, Mann JJ. The anatomy of mood disorders—review of structural neuroimaging studies. Biological psychiatry. 1997;41(1):86–106. pmid:8988799
View Article
PubMed/NCBI
Google Scholar

[204] View Article

[205] PubMed/NCBI

[206] Google Scholar

[ref62] 62. Andreasen NC, Olsen SA, Dennert JW, Smith MR. Ventricular enlargement in schizophrenia: Relationship to positive and negative symptoms. Am J Psychiatry. 1982;139(3):297–302. pmid:7058942
View Article
PubMed/NCBI
Google Scholar

[208] View Article

[209] PubMed/NCBI

[210] Google Scholar

[ref63] 63. Ranft K, Dobrowolny H, Krell D, Bielau H, Bogerts B, Bernstein H-G. Evidence for structural abnormalities of the human habenular complex in affective disorders but not in schizophrenia. Psychological Medicine. 2010;40(04):557–67.
View Article
Google Scholar

[212] View Article

[213] Google Scholar

[ref64] 64. Savitz JB, Nugent AC, Bogers W, Roiser JP, Bain EE, Neumeister A, et al. Habenula Volume in Bipolar Disorder and Major Depressive Disorder: A High-Resolution Magnetic Resonance Imaging Study. Biological Psychiatry. 2011;69(4):336–43. http://dx.doi.org/10.1016/j.biopsych.2010.09.027. pmid:21094939
View Article
PubMed/NCBI
Google Scholar

[215] View Article

[216] PubMed/NCBI

[217] Google Scholar

[ref65] 65. Savitz JB, Bonne O, Nugent AC, Vythilingam M, Bogers W, Charney DS, et al. Habenula volume in post-traumatic stress disorder measured with high-resolution MRI. Biol Mood Anxiety Disord. 2011;1:7. pmid:22738208
View Article
PubMed/NCBI
Google Scholar

[219] View Article

[220] PubMed/NCBI

[221] Google Scholar

[ref66] 66. Sartorius A, Henn FA. Deep brain stimulation of the lateral habenula in treatment resistant major depression. Medical hypotheses. 2007;69(6):1305–8. pmid:17498883
View Article
PubMed/NCBI
Google Scholar

[223] View Article

[224] PubMed/NCBI

[225] Google Scholar

[ref67] 67. Hoexter MQ, Miguel EC, Diniz JB, Shavitt RG, Busatto GF, Sato JR. Predicting obsessive—compulsive disorder severity combining neuroimaging and machine learning methods. Journal of Affective Disorders. 2013;150(3):1213–6. http://dx.doi.org/10.1016/j.jad.2013.05.041. pmid:23769292
View Article
PubMed/NCBI
Google Scholar

[227] View Article

[228] PubMed/NCBI

[229] Google Scholar

[ref68] 68. Wang Y, Fan Y, Bhatt P, Davatzikos C. High-dimensional pattern regression using machine learning: from medical images to continuous clinical variables. Neuroimage. 2010;50(4):1519–35. pmid:20056158.
View Article
PubMed/NCBI
Google Scholar

[231] View Article

[232] PubMed/NCBI

[233] Google Scholar

[ref69] 69. Stonnington CM, Chu C, Klöppel S, Jack CR Jr., Ashburner J, Frackowiak RSJ. Predicting clinical scores from magnetic resonance scans in Alzheimer's disease. Neuroimage. 2010;51(4):1405–13. pmid:20347044.
View Article
PubMed/NCBI
Google Scholar

[235] View Article

[236] PubMed/NCBI

[237] Google Scholar

[ref70] 70. Sapolsky RM. Depression, antidepressants, and the shrinking hippocampus. Proceedings of the National Academy of Sciences. 2001;98(22):12320–2.
View Article
Google Scholar

[239] View Article

[240] Google Scholar

[ref71] 71. Dorée J-P, Rosiers JD, Lew V, Gendron A, Elie R, Stip E, et al. Quetiapine augmentation of treatment-resistant depression: a comparison with lithium. Current Medical Research and Opinion. 2007;23(2):333–41. pmid:17288688
View Article
PubMed/NCBI
Google Scholar

[242] View Article

[243] PubMed/NCBI

[244] Google Scholar

Figures

Abstract

Introduction

Materials and Methods

Participants

Image Acquisition

Image Pre-processing

Individual Subject Classification

Group-level Differences

Individual Subject MGH-S and Severity Score Predictions

Group-level Correlations

Results

Participant Characteristics

Individual Subject Classification

Group-level Differences

Individual Subject MGH-S and Severity Score Predictions

Group-level Correlations

Discussion

Study Limitations

Conclusions

Supporting Information

S1 File. Supporting Information and Severity Score Prediction Information.

S1 Fig. The best fit lines for whole brain severity score predictions (left: HDRS17 predictions, right: MADRS predictions).

S2 Fig. The brain regions which were identified as the most predictive during the whole brain severity score predictions (left: HDRS17 predictions, right: MADRS predictions).

S3 Fig. The best fit line for the prediction of the BDI score using thresholded multiple linear regression feature selection and grey matter images.

S4 Fig. Group-level positive correlations between grey matter and HAM_D (top) and MADRS (bottom).

S5 Fig. Group-level negative correlations between grey matter and HAM_D.

S6 Fig. Group-level negative correlations between grey matter and MADRS.

S7 Fig. Group-level grey matter decreases in patients with TRD with increased BDI scores.

S8 Fig. Group-level grey matter decreases in patients with TRD with increased MGH-S scores.

S9 Fig. The best fit line for the correlation between the MGH-S scores and the volume of grey matter contained within a 5mm sphere centred within the mid-insulae (left (top), right (bottom)).

S1 Data. Supplemental data.

Acknowledgments

Author Contributions

References

S1 Fig. The best fit lines for whole brain severity score predictions (left: HDRS₁₇ predictions, right: MADRS predictions).

S2 Fig. The brain regions which were identified as the most predictive during the whole brain severity score predictions (left: HDRS₁₇ predictions, right: MADRS predictions).