How machine learning applied on fMRI could improve the prognosis of the post-traumatic stress disorder: A systematic review.

BACKGROUND: Post-traumatic stress disorder (PTSD) is a stressor-related disorder that affects a significant proportion of the population worldwide. Despite the neurological nature of this disorder, its functional biomarkers, due to mixed results remain far to be understood. Therefore, the present review study aims to explore functional brain differences between individuals with PTSD and healthy controls with (TEHC) and without (TUHC) trauma exposure, using a machine learning approach. METHOD: A systematic search was performed in PubMed, Embase, and Scopus to identify relevant studies published before 1st of November 2024. A total of 19 studies were included in our data extraction process. The paper has been registered on OSF (https://archive.org/details/osf-registrations-n6bc7-v1)/(https://osf.io/n6bc7/). RESULTS: Default mode and salience networks have been determinant in classifying PTSD participants from both TEHC and TUHC. However, middle frontal gyrus as well as sensory motor area were only involved in classifying PTSD participants from TEHC. Finally, the results have also shown that the association between left amygdala and hippocampus is determinant in identification of PTSD severity. LIMITATIONS: The analysis of the available literature was restricted due to the non-homogeneous characteristics of studies - both in terms of methodology and clinical aspects - which restricted our ability to draw comprehensive conclusions. In addition, some studies used overlapping samples therefore limiting the generalizability of the results. CONCLUSION: The identified networks play a crucial role in distinguishing PTSD participants from healthy participants. These findings could aid in developing more accurate diagnostic tools and may even help predict individuals at higher risk of developing PTSD following exposure to trauma events.