Classify the fNIRS signals of first-episode drug-naive MDD patients with or without suicidal ideation using machine learning

Comprehensive Summary

This study explored whether functional near infrared spectroscopy can objectively identify suicidal ideation in first episode drug naive patients with major depressive disorder. To do this, the researchers recruited 91 first episode MDD patients along with 39 healthy controls and measured prefrontal cortex activation during a verbal fluency task. Patients were categorized based on suicidal ideation scores. A one dimensional convolutional neural network was applied to classify differences in brain activation patterns. Patients with suicidal ideation demonstrated significantly reduced activation across the prefrontal cortex, particularly in the dorsolateral prefrontal cortex, frontopolar cortex, and orbitofrontal cortex, compared to both non suicidal patients and healthy controls. Lower activation in these regions was significantly associated with greater suicidal ideation severity. The classification model achieved up to 69.8 percent accuracy, with area under the curve values reaching as high as 0.88 to 0.92 in specific regions. Overall, the findings suggest that dysfunction in the DLPFC, FPC, and OFC may serve as potential neuroimaging biomarkers of suicidal ideation in early untreated MDD.

Outcomes and Implications

Suicide risk assessment in depression still relies primarily on subjective clinical scales, which can miss patients who underreport suicidal thoughts. This study highlights the potential role of fNIRS as an objective and non invasive adjunct tool for identifying elevated suicide risk, particularly early in the course of illness. Because fNIRS is portable and relatively accessible, it could realistically be integrated into psychiatric evaluations in the future. Although additional validation and external replication are required before routine clinical implementation, this work provides an important step and foundation toward incorporating machine learning driven neuroimaging biomarkers into suicide prevention strategies.