Multimodal MRI-Based Classification of Trauma Survivors with and without Post-Traumatic Stress Disorder

Systematic review of machine learning in PTSD studies for automated diagnosis evaluation

Post-traumatic stress disorder (PTSD) is frequently underdiagnosed due to its clinical and biological heterogeneity. Worldwide, many people face barriers to accessing accurate and timely diagnoses. Machine learning (ML) techniques have been utilized for early assessments and outcome prediction to address these challenges. This paper aims to conduct a systematic review to investigate if ML is a promising approach for PTSD diagnosis. In this review, statistical methods were employed to synthesize the outcomes of the included research and provide guidance on critical considerations for ML task implementation. These included (a) selection of the most appropriate ML model for the available dataset, (b) identification of optimal ML features based on the chosen diagnostic method, (c) determination of appropriate sample size based on the distribution of the data, and (d) implementation of suitable validation tools to assess the performance of the selected ML models. We screened 3186 studies and included 41 articles based on eligibility criteria in the final synthesis. Here we report that the analysis of the included studies highlights the potential of artificial intelligence (AI) in PTSD diagnosis. However, implementing AI-based diagnostic systems in real clinical settings requires addressing several limitations, including appropriate regulation, ethical considerations, and protection of patient privacy.

Imaging the cerebellum in post-traumatic stress and anxiety disorders: a mini-review

Post-traumatic stress disorder (PTSD) and anxiety disorders are among the most prevalent psychiatric conditions worldwide sharing many clinical manifestations and, most likely, neural mechanisms as suggested by neuroimaging research. While the so-called fear circuitry and traditional limbic structures of the brain, particularly the amygdala, have been extensively studied in sufferers of these disorders, the cerebellum has been relatively underexplored. The aim of this paper was to present a mini-review of functional (task-activity or resting-state connectivity) and structural (gray matter volume) results on the cerebellum as reported in magnetic resonance imaging studies of patients with PTSD or anxiety disorders (49 selected studies in 1,494 patients). While mixed results were noted overall, e.g., regarding the direction of effects and anatomical localization, cerebellar structures like the vermis seem to be highly involved. Still, the neurofunctional and structural alterations reported for the cerebellum in excessive anxiety and trauma are complex, and in need of further evaluation.

Neurobiological Clusters Are Associated With Trajectories of Overall Psychopathology in Youth

BACKGROUND

Integrating multiple neuroimaging modalities to identify clusters of individuals and then associating these clusters with psychopathology is a promising approach for understanding neurobiological mechanisms that underlie psychopathology and the extent to which these features are associated with clinical symptoms.

METHODS

We leveraged neuroimaging data from T1-weighted, diffusion-weighted, and resting-state functional magnetic resonance images from the Adolescent Brain Cognitive Development (ABCD) Study (N = 8035) and used similarity network fusion and spectral clustering to identify subgroups of participants. We examined neuroimaging measures as a function of clustering profiles using 1, 2, or 3 imaging modalities (i.e., data combinations), calculated the stability of the clustering assignment in each respective data combination, and compared the consistency of clusters across different data combinations. We then compared the extent to which clusters were associated with overall psychopathology at the baseline assessment and at 2 yearly follow-up visits.

RESULTS

Each data combination resulted in optimal clusters ranging from 2 to 4 subgroups for each data combination. Clusters were stable across subsampling of the ABCD Study cohort. Widespread structural measures (surface area, fractional anisotropy, and mean diffusivity) were important features contributing to clustering across different data combinations. Five of the seven data combinations were associated with overall psychopathology, both at baseline and over time (d = 0.08-0.41). Generally, lower global cortical volume and surface area, widespread reduced fractional anisotropy, and increased radial diffusivity were associated with increased overall psychopathology.

CONCLUSIONS

Profiles constructed from neuroimaging data combinations are associated with concurrent and future psychopathology trajectories.

Modern views of machine learning for precision psychiatry

In light of the National Institute of Mental Health (NIMH)'s Research Domain Criteria (RDoC), the advent of functional neuroimaging, novel technologies and methods provide new opportunities to develop precise and personalized prognosis and diagnosis of mental disorders. Machine learning (ML) and artificial intelligence (AI) technologies are playing an increasingly critical role in the new era of precision psychiatry. Combining ML/AI with neuromodulation technologies can potentially provide explainable solutions in clinical practice and effective therapeutic treatment. Advanced wearable and mobile technologies also call for the new role of ML/AI for digital phenotyping in mobile mental health. In this review, we provide a comprehensive review of ML methodologies and applications by combining neuroimaging, neuromodulation, and advanced mobile technologies in psychiatry practice. We further review the role of ML in molecular phenotyping and cross-species biomarker identification in precision psychiatry. We also discuss explainable AI (XAI) and neuromodulation in a closed human-in-the-loop manner and highlight the ML potential in multi-media information extraction and multi-modal data fusion. Finally, we discuss conceptual and practical challenges in precision psychiatry and highlight ML opportunities in future research.

Neurobiological Features of Posttraumatic Stress Disorder (PTSD) and Their Role in Understanding Adaptive Behavior and Stress Resilience

Posttraumatic stress disorder (PTSD) has been impacting the functioning of a large number of people in military activities and victims of violence for many generations. However, investments in research aiming to understand the neurobiological aspects of the disorder started relatively late, around the last third of the 20th century. The development of neuroimaging methods has greatly supported further understanding of the structural and functional changes in the re-organization processes of brains with PTSD. This helps to better explain the severity and evolution of behavioral symptoms, and opens the possibilities for identifying individual preexisting structural characteristics that could increase symptom severity and the risk of development. Here, we review the advances in neuroanatomical research on these adaptations in PTSD and discuss how those modifications in prefrontal and anterior cingulate circuitry impact the severity and development of the disorder, detaching the research from an amygdalocentric perspective. In addition, we investigate existing and contradictory evidence regarding the preexisting neurobiological features found mostly in twin studies and voxel-based morphometry (VBM) reports.

The cerebellum and psychological trauma: A systematic review of neuroimaging studies

Psychological trauma is highly prevalent among psychiatric disorders, however, the relationship between trauma, neurobiology and psychopathology is not yet fully understood. The cerebellum has been recognized as a crucial structure for cognition and emotion, however, it has been relatively ignored in the literature of psychological trauma, as it is not considered as part of the traditional fear neuro-circuitry. The aim of this review is to investigate how psychological trauma affects the cerebellum and to make conclusive remarks on whether the cerebellum forms part of the trauma-affected brain circuitry. A total of 267 unique records were screened and 39 studies were included in the review. Structural cerebellar alterations and aberrant cerebellar activity and connectivity in trauma-exposed individuals were consistently reported across studies. Early-onset of adverse experiences was associated with cerebellar alterations in trauma-exposed individuals. Several studies reported alterations in connectivity between the cerebellum and nodes of large-brain networks, which are implicated in several psychiatric disorders, including the default mode network, the salience network and the central executive network. Also, trauma-exposed individuals showed altered resting state and task based cerebellar connectivity with cortical and subcortical structures that are involved in emotion and fear regulation. Our preferred interpretation of the results is through the lens of the Universal Cerebellar Transform, the hypothesis that the cerebellum, given its homogeneous cytoarchitecture, performs a common computation for motor, cognitive and emotional functions. Therefore, trauma-induced alterations in this computation might set the ground for a variety of psychiatric symptoms.

Cortical thickness abnormalities in patients with post-traumatic stress disorder: A vertex-based meta-analysis

Neuroimaging studies report altered cortical thickness in patients with post-traumatic stress disorder (PTSD), but the results are inconsistent. Using anisotropic effect-size seed-based d mapping (AES-SDM) software with its recently-developed meta-analytic thickness mask, we conducted a meta-analysis of published studies which used whole-brain surface-based morphometry, in order to define consistent cortical thickness alterations in PTSD patients. Eleven studies with 438 patients and 396 controls were included. Compared with all controls, patients with PTSD showed increased cortical thickness in right superior temporal gyrus, and in left and right superior frontal gyrus; the former survived in subgroup analysis of adult patients, and in subgroup comparison with only non-PTSD trauma-exposed controls, the latter in subgroup comparison with only non-trauma-exposed healthy controls. Cortical thickness in right superior frontal gyrus was positively associated with percentage of female patients, and cortical thickness in left superior frontal gyrus was positively associated with symptom severity measured by the clinician-administered PTSD scale. These robust results may help to elucidate the pathophysiology of PTSD.

Combining Deep Learning and Graph-Theoretic Brain Features to Detect Posttraumatic Stress Disorder at the Individual Level

Previous studies using resting-state functional MRI (rs-fMRI) have revealed alterations in graphical metrics in groups of individuals with posttraumatic stress disorder (PTSD). To explore the ability of graph measures to diagnose PTSD and capture its essential features in individual patients, we used a deep learning (DL) model based on a graph-theoretic approach to discriminate PTSD from trauma-exposed non-PTSD at the individual level and to identify its most discriminant features. Our study was performed on rs-fMRI data from 91 individuals with PTSD and 126 trauma-exposed non-PTSD patients. To evaluate our DL method, we used the traditional support vector machine (SVM) classifier as a reference. Our results showed that the proposed DL model allowed single-subject discrimination of PTSD and trauma-exposed non-PTSD individuals with higher accuracy (average: 80%) than the traditional SVM (average: 57.7%). The top 10 DL features were identified within the default mode, central executive, and salience networks; the first two of these networks were also identified in the SVM classification. We also found that nodal efficiency in the left fusiform gyrus was negatively correlated with the Clinician Administered PTSD Scale score. These findings demonstrate that DL based on graphical features is a promising method for assisting in the diagnosis of PTSD.

Supervised machine learning for diagnostic classification from large-scale neuroimaging datasets

There are growing concerns about the generalizability of machine learning classifiers in neuroimaging. In order to evaluate this aspect across relatively large heterogeneous populations, we investigated four disorders: Autism spectrum disorder (N = 988), Attention deficit hyperactivity disorder (N = 930), Post-traumatic stress disorder (N = 87) and Alzheimer's disease (N = 132). We applied 18 different machine learning classifiers (based on diverse principles) wherein the training/validation and the hold-out test data belonged to samples with the same diagnosis but differing in either the age range or the acquisition site. Our results indicate that overfitting can be a huge problem in heterogeneous datasets, especially with fewer samples, leading to inflated measures of accuracy that fail to generalize well to the general clinical population. Further, different classifiers tended to perform well on different datasets. In order to address this, we propose a consensus-classifier by combining the predictive power of all 18 classifiers. The consensus-classifier was less sensitive to unmatched training/validation and holdout test data. Finally, we combined feature importance scores obtained from all classifiers to infer the discriminative ability of connectivity features. The functional connectivity patterns thus identified were robust to the classification algorithm used, age and acquisition site differences, and had diagnostic predictive ability in addition to univariate statistically significant group differences between the groups. A MATLAB toolbox called Machine Learning in NeuroImaging (MALINI), which implements all the 18 different classifiers along with the consensus classifier is available from Lanka et al. (2019) The toolbox can also be found at the following URL: https://github.com/pradlanka/malini .

Multi-voxel pattern analysis of amygdala functional connectivity at rest predicts variability in posttraumatic stress severity

INTRODUCTION

Resting state functional magnetic resonance imaging (rsfMRI) studies demonstrate that individuals with posttraumatic stress disorder (PTSD) exhibit atypical functional connectivity (FC) between the amygdala, involved in the generation of emotion, and regions responsible for emotional appraisal (e.g., insula, orbitofrontal cortex [OFC]) and regulation (prefrontal cortex [PFC], anterior cingulate cortex). Consequently, atypical amygdala FC within an emotional processing and regulation network may be a defining feature of PTSD, although altered FC does not seem constrained to one brain region. Instead, altered amygdala FC involves a large, distributed brain network in those with PTSD. The present study used a machine-learning data-driven approach, multi-voxel pattern analysis (MVPA), to predict PTSD severity based on whole-brain patterns of amygdala FC.

METHODS

Trauma-exposed adults (N = 90) completed the PTSD Checklist-Civilian Version to assess symptoms and a 5-min rsfMRI. Whole-brain FC values to bilateral amygdala were extracted and used in a relevance vector regression analysis with a leave-one-out approach for cross-validation with permutation testing (1,000) to obtain significance values.

RESULTS

Results demonstrated that amygdala FC predicted PCL-C scores with statistically significant accuracy (r = .46, p = .001; mean sum of squares = 130.46, p = .001; R²  = 0.21, p = .001). Prediction was based on whole-brain amygdala FC, although regions that informed prediction (top 10%) included the OFC, amygdala, and dorsolateral PFC.

CONCLUSION

Findings demonstrate the utility of MVPA based on amygdala FC to predict individual severity of PTSD symptoms and that amygdala FC within a fear acquisition and regulation network contributed to accurate prediction.

Toward Robust Anxiety Biomarkers: A Machine Learning Approach in a Large-Scale Sample

BACKGROUND

The field of psychiatry has long sought biomarkers that can objectively diagnose patients, predict treatment response, or identify individuals at risk of illness onset. However, reliable psychiatric biomarkers have yet to emerge. The recent application of machine learning techniques to develop neuroimaging-based biomarkers has yielded promising preliminary results. However, much of the work in this domain has not met best practice standards from the field of machine learning. This is especially true for studies of anxiety, creating uncertainty about the potential for anxiety biomarker development.

METHODS

We applied machine learning tools to predict trait anxiety from neuroimaging measurements in humans. Using publicly available data from the Brain Genomics Superstruct Project, we compared a suite of neuroimaging-based machine learning models predicting anxiety within a discovery sample (n = 531, 307 women) via k-fold cross-validation, and we tested the final model (a stacked model incorporating region-to-region functional connectivity, amygdala seed-to-voxel connectivity, and volumetric and cortical thickness data) in a held-out, unseen test sample (n = 348, 209 women).

RESULTS

Though the best model was able to predict anxiety within the discovery sample (cross-validated R2 of .06, permutation test p < .001), the generalization test within the holdout sample failed (R2 of -.04, permutation test p > .05).

CONCLUSIONS

In this study, we did not find evidence of a generalizable anxiety biomarker. However, we encourage other researchers to investigate this topic, utilizing large samples and proper methodology, to clarify the potential of neuroimaging-based anxiety biomarkers.

Diagnostic and Predictive Neuroimaging Biomarkers for Posttraumatic Stress Disorder

BACKGROUND

Comorbidity between posttraumatic stress disorder (PTSD) and major depressive disorder (MDD) has been commonly overlooked by studies examining resting-state functional connectivity patterns in PTSD. The current study used a data-driven approach to identify resting-state functional connectivity biomarkers to 1) differentiate individuals with PTSD (with or without MDD) from trauma-exposed healthy control subjects (TEHCs), 2) compare individuals with PTSD alone with those with comorbid PTSD+MDD, and 3) explore the clinical utility of the identified biomarkers by testing their associations with clinical symptoms and treatment response.

METHODS

Resting-state magnetic resonance images were obtained from 51 individuals with PTSD alone, 52 individuals with PTSD+MDD, and 76 TEHCs. Of the 103 individuals with PTSD, 55 were enrolled in prolonged exposure treatment. A support vector machine model was used to identify resting-state functional connectivity biomarkers differentiating individuals with PTSD (with or without MDD) from TEHCs and differentiating individuals with PTSD alone from those with PTSD+MDD. The associations between the identified features and symptomatology were tested with Pearson correlations.

RESULTS

The support vector machine model achieved 70.6% accuracy in discriminating between individuals with PTSD and TEHCs and achieved 76.7% accuracy in discriminating between individuals with PTSD alone and those with PTSD+MDD for out-of-sample prediction. Within-network connectivity in the executive control network, prefrontal network, and salience network discriminated individuals with PTSD from TEHCs. The basal ganglia network played an important role in differentiating individuals with PTSD alone from those with PTSD+MDD. PTSD scores were inversely correlated with within-executive control network connectivity (p < .001), and executive control network connectivity was positively correlated with treatment response (p < .001).

CONCLUSIONS

Results suggest that unique brain-based abnormalities differentiate individuals with PTSD from TEHCs, differentiate individuals with PTSD from those with PTSD+MDD, and demonstrate clinical utility in predicting levels of symptomatology and treatment response.

Multivariate classification of earthquake survivors with post-traumatic stress disorder based on large-scale brain networks

OBJECTIVE

The identification of post-traumatic stress disorder (PTSD) among natural disaster survivors is remarkably challenging, and there are no reliable objective signatures that can be used to assist clinical diagnosis and optimize treatment. The current study aimed to establish a neurobiological signature of PTSD from the connectivity of large-scale brain networks and clarify the brain network mechanisms of PTSD.

METHODS

We examined fifty-seven unmedicated survivors with chronic PTSD and 59 matched trauma-exposed healthy controls (TEHCs) using resting-state functional magnetic resonance imaging (rs-fMRI). We extracted the node-to-network connectivity and obtained a feature vector with a dimensionality of 864 (108 nodes × 8 networks) to represent each subject's functional connectivity (FC) profile. Multivariate pattern analysis with a relevance vector machine was then used to distinguish PTSD patients from TEHCs.

RESULTS

We achieved a promising diagnostic accuracy of 89.2% in distinguishing PTSD patients from TEHCs. The most heavily weighted connections for PTSD classification were among the default mode network (DMN), visual network (VIS), somatomotor network, limbic network, and dorsal attention network (DAN). The strength of the anticorrelation of FC between the ventral medial prefrontal cortex (vMPFC) in DMN and the VIS and DAN was associated with the severity of PTSD.

CONCLUSIONS

This study achieved relatively high accuracy in classifying PTSD patients vs. TEHCs at the individual level. This performance demonstrates that rs-fMRI-derived multivariate classification based on large-scale brain networks can provide potential signatures both to facilitate clinical diagnosis and to clarify the underlying brain network mechanisms of PTSD caused by natural disasters.

Exploring memory function in earthquake trauma survivors with resting-state fMRI and machine learning

BACKGROUND

Traumatized earthquake survivors may develop poor memory function. Resting-state functional magnetic resonance imaging (rs-fMRI) and machine learning techniques may one day aid the clinical assessment of individual psychiatric patients. This study aims to use machine learning with Rs-fMRI from the perspectives of neurophysiology and neuroimaging to explore the association between it and the individual memory function of trauma survivors.

METHODS

Rs-fMRI data was acquired for eighty-nine survivors (male (33%), average age (SD):45.18(6.31) years) of Wenchuan earthquakes in 2008 each of whom was screened by experienced psychiatrists based on the clinician-administered post-traumatic stress disorder (PTSD) scale (CAPS), and their memory function scores were determined by the Wechsler Memory Scale-IV (WMS-IV). We explored which memory function scores were significantly associated with CAPS scores. Using simple multiple kernel learning (MKL), Rs-fMRI was used to predict the memory function scores that were associated with CAPS scores. A support vector machine (SVM) was also used to make classifications in trauma survivors with or without PTSD.

RESULTS

Spatial addition (SA), which is defined by spatial working memory function, was negatively correlated with the total CAPS score (r = - 0.22, P = 0.04). The use of simple MKL allowed quantitative association of SA scores with statistically significant accuracy (correlation = 0.28, P = 0.03; mean squared error = 8.36; P = 0.04). The left middle frontal gyrus and the left precuneus contributed the largest proportion to the simple MKL association frame. The SVM could not make a quantitative classification of diagnosis with statistically significant accuracy.

LIMITATIONS

The use of the cross-sectional study design after exposure to an earthquake and the leave-one-out cross-validation (LOOCV) increases the risk of overfitting.

CONCLUSION

Spontaneous brain activity of the left middle frontal gyrus and the left precuneus acquired by rs-fMRI may be a brain mechanism of visual working memory that is related to PTSD symptoms. Machine learning may be a useful tool in the identification of brain mechanisms of memory impairment in trauma survivors.

The use of machine learning techniques in trauma-related disorders: a systematic review

Establishing the diagnosis of trauma-related disorders such as Acute Stress Disorder (ASD) and Posttraumatic Stress Disorder (PTSD) have always been a challenge in clinical practice and in academic research, due to clinical and biological heterogeneity. Machine learning (ML) techniques can be applied to improve classification of disorders, to predict outcomes or to determine person-specific treatment selection. We aim to review the existing literature on the use of machine learning techniques in the assessment of subjects with ASD or PTSD. We systematically searched PubMed, Embase and Web of Science for articles published in any language up to May 2019. We found 806 abstracts and included 49 studies in our review. Most of the included studies used multiple levels of biological data to predict risk factors or to identify early symptoms related to PTSD. Other studies used ML classification techniques to distinguish individuals with ASD or PTSD from other psychiatric disorder or from trauma-exposed and healthy controls. We also found studies that attempted to define outcome profiles using clustering techniques and studies that assessed the relationship among symptoms using network analysis. Finally, we proposed a quality assessment in this review, evaluating methodological and technical features on machine learning studies. We concluded that etiologic and clinical heterogeneity of ASD/PTSD patients is suitable to machine learning techniques and a major challenge for the future is to use it in clinical practice for the benefit of patients in an individual level.

Individual prediction of psychotherapy outcome in posttraumatic stress disorder using neuroimaging data

Trauma-focused psychotherapy is the first-line treatment for posttraumatic stress disorder (PTSD) but 30-50% of patients do not benefit sufficiently. We investigated whether structural and resting-state functional magnetic resonance imaging (MRI/rs-fMRI) data could distinguish between treatment responders and non-responders on the group and individual level. Forty-four male veterans with PTSD underwent baseline scanning followed by trauma-focused psychotherapy. Voxel-wise gray matter volumes were extracted from the structural MRI data and resting-state networks (RSNs) were calculated from rs-fMRI data using independent component analysis. Data were used to detect differences between responders and non-responders on the group level using permutation testing, and the single-subject level using Gaussian process classification with cross-validation. A RSN centered on the bilateral superior frontal gyrus differed between responders and non-responder groups (P_FWE < 0.05) while a RSN centered on the pre-supplementary motor area distinguished between responders and non-responders on an individual-level with 81.4% accuracy (P < 0.001, 84.8% sensitivity, 78% specificity and AUC of 0.93). No significant single-subject classification or group differences were observed for gray matter volume. This proof-of-concept study demonstrates the feasibility of using rs-fMRI to develop neuroimaging biomarkers for treatment response, which could enable personalized treatment of patients with PTSD.

Integrating machining learning and multimodal neuroimaging to detect schizophrenia at the level of the individual

Schizophrenia is a severe psychiatric disorder associated with both structural and functional brain abnormalities. In the past few years, there has been growing interest in the application of machine learning techniques to neuroimaging data for the diagnostic and prognostic assessment of this disorder. However, the vast majority of studies published so far have used either structural or functional neuroimaging data, without accounting for the multimodal nature of the disorder. Structural MRI and resting‐state functional MRI data were acquired from a total of 295 patients with schizophrenia and 452 healthy controls at five research centers. We extracted features from the data including gray matter volume, white matter volume, amplitude of low‐frequency fluctuation, regional homogeneity and two connectome‐wide based metrics: structural covariance matrices and functional connectivity matrices. A support vector machine classifier was trained on each dataset separately to distinguish the subjects at individual level using each of the single feature as well as their combination, and 10‐fold cross‐validation was used to assess the performance of the model. Functional data allow higher accuracy of classification than structural data (mean 82.75% vs. 75.84%). Within each modality, the combination of images and matrices improves performance, resulting in mean accuracies of 81.63% for structural data and 87.59% for functional data. The use of all combined structural and functional measures allows the highest accuracy of classification (90.83%). We conclude that combining multimodal measures within a single model is a promising direction for developing biologically informed diagnostic tools in schizophrenia.

Aberrant regional homogeneity in post-traumatic stress disorder after traffic accident: A resting-state functional MRI study

OBJECTIVES

The present study explored the changes in spontaneous regional activity in post-traumatic stress disorder (PTSD) patients, who experienced severe traffic accidents.

METHODS

20 drug-naive PTSD patients and 18 healthy control subjects were imaged using resting-state functional magnetic resonance imaging (rs-fMRI) and analyzed by the algorithm of regional homogeneity (ReHo).

RESULTS

Compared to the healthy control group, the PTSD group showed decreased ReHo values in the right angular gyrus. In addition, a negative correlation was found between the activity level of the angular gyrus and the CAPS score.

CONCLUSION

The dysfunctions were found in the memory- and emotion-related areas, suggested a possible mechanism of memory dysregulation that might be related to the intrusive memory symptoms of PTSD. These results provided imaging evidence that might provide an in-depth understanding of the intrinsic functional architecture of PTSD.

Machine Learning for Prediction of Posttraumatic Stress and Resilience Following Trauma: An Overview of Basic Concepts and Recent Advances

Posttraumatic stress responses are characterized by a heterogeneity in clinical appearance and etiology. This heterogeneity impacts the field's ability to characterize, predict, and remediate maladaptive responses to trauma. Machine learning (ML) approaches are increasingly utilized to overcome this foundational problem in characterization, prediction, and treatment selection across branches of medicine that have struggled with similar clinical realities of heterogeneity in etiology and outcome, such as oncology. In this article, we review and evaluate ML approaches and applications utilized in the areas of posttraumatic stress, stress pathology, and resilience research, and present didactic information and examples to aid researchers interested in the relevance of ML to their own research. The examined studies exemplify the high potential of ML approaches to build accurate predictive and diagnostic models of posttraumatic stress and stress pathology risk based on diverse sources of available information. The use of ML approaches to integrate high-dimensional data demonstrates substantial gains in risk prediction even when the sources of data are the same as those used in traditional predictive models. This area of research will greatly benefit from collaboration and data sharing among researchers of posttraumatic stress disorder, stress pathology, and resilience.

Multimodal MRI-based classification of migraine: using deep learning convolutional neural network

Background

Recently, deep learning technologies have rapidly expanded into medical image analysis, including both disease detection and classification. As far as we know, migraine is a disabling and common neurological disorder, typically characterized by unilateral, throbbing and pulsating headaches. Unfortunately, a large number of migraineurs do not receive the accurate diagnosis when using traditional diagnostic criteria based on the guidelines of the International Headache Society. As such, there is substantial interest in developing automated methods to assist in the diagnosis of migraine.

Methods

To the best of our knowledge, no studies have evaluated the potential of deep learning technologies in assisting with the classification of migraine patients. Here, we used deep learning methods in combination with three functional measures (the amplitude of low-frequency fluctuations, regional homogeneity and regional functional correlation strength) based on rs-fMRI data to distinguish not only between migraineurs and healthy controls, but also between the two subtypes of migraine. We employed 21 migraine patients without aura, 15 migraineurs with aura, and 28 healthy controls.

Results

Compared with the traditional support vector machine classifier, which has an accuracy of 83.67%, our Inception module-based convolutional neural network approach showed a significant improvement in classification output (over 86.18%). Our data also indicate that the Inception module-based CNN performs better than the AlexNet-based CNN (Inception module-based CNN reached an accuracy of 99.25%). Finally, we also found that regional functional correlation strength (RFCS) could be regarded as the optimum input out of the three indices (ALFF, ReHo, RFCS).

Conclusions

Overall, our study shows that combining the three functional measures of rs-fMRI with deep learning classification is a powerful method to distinguish between migraineurs and healthy individuals. Our data also highlight that deep learning-based frameworks could be used to develop more complicated models or systems to aid in clinical decision making in the future.

Electronic supplementary material

The online version of this article (10.1186/s12938-018-0587-0) contains supplementary material, which is available to authorized users.

Pre-treatment Resting-State Functional MR Imaging Predicts the Long-Term Clinical Outcome After Short-Term Paroxtine Treatment in Post-traumatic Stress Disorder

Background: The chronic phase of post-traumatic stress disorder (PTSD) and the limited effectiveness of existing treatments creates the need for the development of potential biomarkers to predict response to antidepressant medication at an early stage. However, findings at present focus on acute therapeutic effect without following-up the long-term clinical outcome of PTSD. So far, studies predicting the long-term clinical outcome of short-term treatment based on both pre-treatment and post-treatment functional MRI in PTSD remains limited. Methods: Twenty-two PTSD patients were scanned using resting-state functional MRI (rs-fMRI) before and after 12 weeks of treatment with paroxetine. Twenty patients were followed up using the same psychopathological assessments 2 years after they underwent the second MRI scan. Based on clinical outcome, the follow-up patients were divided into those with remitted PTSD or persistent PTSD. Amplitude of low-frequency fluctuations (ALFF) and degree centrality (DC) derived from pre-treatment and post-treatment rs-fMRI were used as classification features in a support vector machine (SVM) classifier. Results: Prediction of long-term clinical outcome by combined ALFF and DC features derived from pre-treatment rs-fMRI yielded an accuracy rate of 72.5% (p < 0.005). The most informative voxels for outcome prediction were mainly located in the precuneus, superior temporal area, insula, dorsal medial prefrontal cortex, frontal orbital cortex, supplementary motor area, lingual gyrus, and cerebellum. Long-term outcome could not be successfully classified by post-treatment imaging features with accuracy rates <50%. Conclusions: Combined information from ALFF and DC from rs-fMRI data before treatment could predict the long-term clinical outcome of PTSD, which is critical for defining potential biomarkers to customize PTSD treatment and improve the prognosis.

Combat Exposure Severity is Associated with Reduced Cortical Thickness in Combat Veterans: A Preliminary Report

Background

Chronic stress and related physiological responses are known to have deleterious effects on neural integrity. Combat exposure is a notoriously pathogenic stressor, and with over 2 million U.S. troops deployed to active combat zones since 2001, there is an urgent need to advance our understanding of its potential neural impact. Previous evidence suggests structural alterations in posttraumatic stress disorder (PTSD) and more recent studies have explored cortical thinning specifically. This preliminary study investigates the impact of combat exposure on cortical thickness, controlling for history of early life stress and age.

Methods

Twenty-one combat-exposed Veterans with PTSD and 20 non-PTSD combat-exposed controls (mean age 32.7) completed the Combat Exposure Scale, Childhood Trauma Questionnaire, and structural magnetic resonance imaging in a Siemens 3T TIM trio system. General linear model was used to examine the effect of combat exposure on cortical thickness, controlling for early life trauma exposure and age using cluster-wise correction (p < 0.05).

Results

This preliminary study found a negative correlation between combat exposure severity (CES) and cortical thickness in the left superior temporal and left rostral middle frontal regions, as well as an interaction between PTSD diagnosis status and CES, in the superior temporal/insular region showing a stronger negative correlation between CES and cortical thickness in the non-PTSD group.

Conclusions

Though caution should be taken with interpretation given the preliminary nature of the findings, the results indicate combat exposure may affect cortical structure beyond possible alterations due to early life stress exposure or PTSD psychopathology. Though replication in larger samples is required, these results provide useful information regarding possible neural biomarkers and treatment targets for combat-related psychopathology as well as highlighting the pathogenic effects of combat.