Performance Evaluation of Attention Mechanism and Spiking Neural Networks on sMRI Data for Suicide Ideation Assessment

The coronavirus disease 2019 (COVID-19) pandemic has had a substantial detrimental impact on mental health, especially depression, and this has led to a high incidence of suicidal ideation (SI) around the globe, with the pandemic's post-peak period seeing the highest incidence in young adults. This study aims to propose an effective non-intrusive method for early detection of SI in young adults utilizing depression as a biomarker in structural magnetic resonance imaging. This paper introduces a hybrid machine learning approach utilizing attention mechanisms and spiking neural networks to differentiate between depression patients without SI and healthy controls. The hybrid method successfully completed the classification task after stratified 5-fold cross-validation, achieving test accuracy, sensitivity, specificity, and area under curve of 94%, 100%, 92%, and 0.96, respectively. The proposed algorithms offer an objective tool for identifying early SI risk in depressed patients without suicidal thoughts, alongside clinical assessment