Classification of suicidality by training supervised machine learning models with brain MRI findings: A systematic...

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Imaging

Background:

• Suicide is a critical global public health issue, necessitating improved methods for early identification of suicidal thoughts and behaviors.
• Magnetic Resonance Imaging (MRI) offers a non-invasive approach to investigate brain structures and functions potentially associated with suicidality.

Approach:

• A systematic literature review was conducted across major scientific databases (PubMed, Scopus, Web of Science).
• Studies applying supervised Machine Learning (ML) methods to MRI data for suicidality identification were included.
• The Prediction Model Risk of Bias Assessment Tool (PROBAST) was utilized for quality assessment of the included studies.

Key Points:

• Twenty-three studies met the inclusion criteria, with most developing prediction models lacking external validation.
• Machine learning model performance varied, but over half achieved high accuracy (≥0.8) and Area Under the Curve (AUC) (≥0.8).
• Deep learning models outperformed other ML approaches, identifying resting-state functional connectivity and grey matter volume in prefrontal-limbic regions as key discriminative features.

Conclusions:

• While many studies developed ML models for suicide identification, their predictive performance is inconsistent.
• Limitations include small sample sizes, lack of external validation, and heterogeneous study designs.
• Further rigorous research is essential to fully realize the potential of ML in identifying suicidality from MRI data.