Deep learning model of fMRI connectivity predicts PTSD symptom trajectories in recent trauma survivors

Abstract

Highlights We propose a novel end-to-end neural network that employs resting-state and task-based functional MRI (fMRI) datasets, obtained one month after trauma exposure, to predict PTSD one, six and 14-months after the exposure. The method utilizes connectivity maps extracted from pairs of brain regions which are subsequently updated by applying the algorithmic technique of pairwise attention. The proposed deep learning method predicts PTSD status, PTSD symptom clusters and survival analysis within the prospective design. We demonstrate a significant improvement in performance on all the datasets and experiments in comparison to other relevant analytical techniques. Pairwise association analysis reveals several significant functional connectivity patterns, in line with previous PTSD neuroimaging literature.

Brain correlates of depression, post-traumatic distress, and inflammatory biomarkers in COVID-19 survivors: A multimodal magnetic resonance imaging study

“…Abnormal neural responses are paralleled by reduced regional grey matter (GM) volumes at MRI voxel-based morphometry (VBM), affecting in particular the anterior cingulate cortex (ACC) ( Matsuo et al, 2019 ; Vai et al, 2020 ; Wise et al, 2017 ; Bora et al, 2012 ), and by abnormal resting state functional connectivity suggesting circuit dysfunction mainly involving the default mode network (DMN), dorsolateral prefrontal cortex, orbitofrontal cortex, ACC, insula, amygdala, hippocampus, basal ganglia, thalamus, and cerebellum ( Spellman and Liston, 2020 ; Scalabrini et al, 2020 ; Vai et al, 2019 ). Consistent findings in patients with post-traumatic stress disorder (PTSD) also affirm aberrant GM reduction ( Li et al, 2014 ; Meng et al, 2016 ) and resting-state connectivity ( Koch et al, 2016 ; Sheynin et al, 2021 ) in corticolimbic structures, largely overlapping with those implicated in mood disorders. Moreover, GM volume reduction and altered connectivity associate with neuropsychiatric symptoms in inflammatory medical conditions ( Schrepf et al, 2018 ; Schweinhardt et al, 2008 ), suggesting that regional GM microstructure and function could mediate the relationship between a medical illness and its psychopathological sequelae.…”

Modern views of machine learning for precision psychiatry

“… 87 supervised classification, deep DNN PTSD rs-fMRI/task fMRI Sheynin et al. 88 supervised classification SVM schizophrenia sMRI Nieuwenhuis et al. 89 supervised classification, deep SVM, random forest, DNN schizophrenia task fMRI Smucny et al.…”

Machine learning for post‐traumatic stress disorder identification utilizing resting‐state functional magnetic resonance imaging

“…ML approaches are progressively utilized to classify, select, predict, and characterize individuals suffering from a range of psychiatric disorders (Rehman et al, 2021; Schultebraucks et al, 2019). Recently, ML approaches have been used for the classification, emotion regulation, treatment selection, prediction, or characterization the PTSD individuals by Zandvakili et al (2020), Nixon et al (2021), Siegel et al (2021), Christ et al (2021), Shahzad et al (2021), and Sheynin et al (2021). The most frequently used ML approaches are logistic regression (LR), K‐nearest neighbor (KNN), and support vector machine linear (SVM) (Jabeen et al, 2018; Saba et al, 2018; Yousaf et al, 2019).…”