Framework for Accurate Classification of Self-Reported Stress From Multisession Functional MRI Data of Veterans With Posttraumatic Stress

Abstract

Background: Posttraumatic stress disorder (PTSD) is a significant burden among combat Veterans returning from the wars in Iraq and Afghanistan. While empirically supported treatments have demonstrated reductions in PTSD symptomatology, there remains a need to improve treatment effectiveness. Functional magnetic resonance imaging (fMRI) neurofeedback has emerged as a possible treatment to ameliorate PTSD symptom severity. Virtual reality (VR) approaches have also shown promise in increasing treatment compliance and outcomes. To facilitate fMRI neurofeedback-associated therapies, it would be advantageous to accurately classify internal brain stress levels while Veterans are exposed to trauma-associated VR imagery. Methods: Across 2 sessions, we used fMRI to collect neural responses to trauma-associated VR-like stimuli among male combat Veterans with PTSD symptoms (N = 8). Veterans reported their self-perceived stress level on a scale from 1 to 8 every 15 s throughout the fMRI sessions. In our proposed framework, we precisely sample the fMRI data on cortical gray matter, blurring the data along the gray-matter manifold to reduce noise and dimensionality while preserving maximum neural information. Then, we independently applied 3 machine learning (ML) algorithms to this fMRI data collected across 2 sessions, separately for each Veteran, to build individualized ML models that predicted their internal brain states (self-reported stress responses). Results: We accurately classified the 8-class self-reported stress responses with a mean (± standard error) root mean square error of 0.6 (± 0.1) across all Veterans using the best ML approach. Conclusions: The findings demonstrate the predictive ability of ML algorithms applied to whole-brain cortical fMRI data collected during individual Veteran sessions. The framework we have developed to preprocess whole-brain cortical fMRI data and train ML models across sessions would provide a valuable tool to enable individualized real-time fMRI neurofeedback during VR-like exposure therapy for PTSD.

Keywords: algorithm design and analysis; behavioral sciences; functional magnetic resonance imaging; machine learning algorithms; mental disorders; posttraumatic stress disorder (PTSD); psychiatry; veterans.

Figure 1.

Individualized VR sequences included 2 fMRI sessions of five 8-min runs with a break of 1 to 2 min between each run. Veterans received an audio cue every 15 s to provide their subjective stress level on a 1 to 8 scale (top). VR-like sequences included alternating calming and stress-inducing stimuli such that Veteran stress levels varied across each run (middle). Calming, mildly stressful, and stressful stimuli were created based on clinical interviews with each Veteran (bottom). VR, virtual reality.

Figure 2.

Preprocessing and transformation of the collected whole-brain fMRI data into cortical fMRI data, which were then used in machine learning (ML) analysis.

Figure 3.

(A-C) Mean root mean square error (RMSE) values for 3 different ML models for each of the 3 different block size combinations, for each of the 8 Veterans. (D) Summary of RMSE across all Veterans for each of the 3 different ML models and for the 3 models combined for each of the 3 block sizes. Error bars represent standard deviation for A to C and standard error across Veterans for D. SVM, support vector machine; SVR, support vector regression; DL, deep learning; BS#, block size of # frames; ML, machine learning.

Figure 4.

Box plots illustrating the results of the permutation tests for 2 of the Veterans. The blue boxplots represent the RMSE values for the ML models run with the original labels, while the orange boxplots represent the RMSE values for the ML models run with permuted labels. ML, machine learning.