MVPA to enhance the study of rare cognitive events: An investigation of experimental PTSD

Abstract

Many cognitive processes are challenging to study due to their scarce occurrence. Here we demonstrate how pattern recognition and brain imaging can enhance the study of such processes by providing fast, sensitive, and non-intrusive detection of these events. This can enable efficient experimental and clinical intervention. We focus on the study of traumatic events producing flashbacks associated with post-traumatic stress disorder (PTSD), using an experimental analogue of trauma (a traumatic film). These events are rare and challenging to reliably elicit in experimental settings. We show that a classifier can be built to predict, based upon brain response, which stimuli are likely to induce these rare flashbacks at the point of exposure. An ability to predict these stimuli makes possible the trialing of context-specific preventative clinical interventions. We present results from two independent datasets, outlining key analytic challenges.

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

“…Going through the very few PTSD classification studies, we found the one used structural imaging feature (Gong et al, 2014 ) also identified a widely distributed brain network that comprised all brain lobes. In the other two studies that used task fMRI (Niehaus et al, 2014 ) and rs-fMRI (Liu et al, 2015 ) data respectively, limbic and prefrontal areas were all considered played key role in discriminating PTSD subjects with healthy controls that showed certain consistency with our discrimination maps. It is also worth noting that the network pattern depends on the threshold selection, so the regions shown on discrimination maps only indicate their relatively high contributions.…”

“…Recent studies have successfully applied multimodal analysis on Alzheimer's disease (Fan et al, 2008 ; Zhang et al, 2011 ; Dai et al, 2012 ; Liu et al, 2014b ), Parkinson's disease (Long et al, 2012 ) and sexual dimorphism (Wang et al, 2012 ). However, so far as we know, most of the very few studies that performed classification on PTSD only utilized single modal imaging data (Gong et al, 2014 ; Niehaus et al, 2014 ). A very recent study (Liu et al, 2015 ) has explored the power of multivariate approach in classifying PTSD in which features at three different levels derived from rs-fMRI data were combined, although this should be still considered as a single-modality study.…”

Neuromarkers for Mental Disorders: Harnessing Population Neuroscience

“…However, a more easily accessible method is cross-validation (CV). One of the most frequently used methods is leave-one-out CV [LOOCV; e.g., ( 67 – 69 )], or leave-k-out CV [e.g., ( 70 )]. A somewhat less computationally expensive method is k-fold CV (e.g., ( 71 )].…”

A Review on Machine Learning Techniques in the Diagnosis of Psychiatric Disorders

“…ML approaches are sensitive to ease inference in the single-subject degree, and may identify spatially dispersed patterns in the mind which may be undetectable using set comparisons. Recently, an increasing number of studies have applied ML approaches to neuroimaging data to forecast and describe psychiatric ailments 10,11 , in addition to Post Traumatic Stress Disorder (PTSD) With a Multivariate Voxel Pattern Investigation (MVPA) or supervised ML 12 ; an individual can classify psychiatric disorder in person neuroimaging data. In keeping with this belief, it's been indicated further the multivariate patterns of mind changes detected by system learning might be exceptionally sensitive to operational changes in the brain as a consequence of psychiatric disorder, and thus can facilitate the translation of neuroimaging in the chair to the bedside.…”