First steps in using machine learning on fMRI data to predict intrusive memories of traumatic film footage

The prevailing neurocircuitry models of anxiety disorders have been amygdalocentric in form. The bases for such models have progressed from theoretical considerations, extrapolated from research in animals, to in vivo human imaging data. For example, one current model of posttraumatic stress disorder (PTSD) has been highly influenced by knowledge from rodent fear conditioning research. Given the phenomenological parallels between fear conditioning and the pathogenesis of PTSD, we have proposed that PTSD is characterized by exaggerated amygdala responses (subserving exaggerated acquisition of fear associations and expression of fear responses) and deficient frontal cortical function (mediating deficits in extinction and the capacity to suppress attention/response to trauma-related stimuli), as well as deficient hippocampal function (mediating deficits in appreciation of safe contexts and explicit learning/memory). Neuroimaging studies have yielded convergent findings in support of this model. However, to date, neuroimaging investigations of PTSD have not principally employed conditioning and extinction paradigms per se. The recent development of such imaging probes now sets the stage for directly testing hypotheses regarding the neural substrates of fear conditioning and extinction abnormalities in PTSD.

A challenging goal in neuroscience is to be able to read out, or decode, mental content from brain activity. Recent functional magnetic resonance imaging (fMRI) studies have decoded orientation, position and object category from activity in visual cortex. However, these studies typically used relatively simple stimuli (for example, gratings) or images drawn from fixed categories (for example, faces, houses), and decoding was based on previous measurements of brain activity evoked by those same stimuli or categories. To overcome these limitations, here we develop a decoding method based on quantitative receptive-field models that characterize the relationship between visual stimuli and fMRI activity in early visual areas. These models describe the tuning of individual voxels for space, orientation and spatial frequency, and are estimated directly from responses evoked by natural images. We show that these receptive-field models make it possible to identify, from a large set of completely novel natural images, which specific image was seen by an observer. Identification is not a mere consequence of the retinotopic organization of visual areas; simpler receptive-field models that describe only spatial tuning yield much poorer identification performance. Our results suggest that it may soon be possible to reconstruct a picture of a person's visual experience from measurements of brain activity alone.

A review of 2,647 studies of posttraumatic stress disorder (PTSD) yielded 476 potential candidates for a meta-analysis of predictors of PTSD or of its symptoms. From these, 68 studies met criteria for inclusion in a meta-analysis of 7 predictors: (a) prior trauma, (b) prior psychological adjustment, (c) family history of psychopathology, (d) perceived life threat during the trauma, (e) posttrauma social support, (f) peritraumatic emotional responses, and (g) peritraumatic dissociation. All yielded significant effect sizes, with family history, prior trauma, and prior adjustment the smallest (weighted r = .17) and peritraumatic dissociation the largest (weighted r = .35). The results suggest that peritraumatic psychological processes, not prior characteristics, are the strongest predictors of PTSD.