Neuroimaging and machine learning in eating disorders: a systematic review

Area of Science:

• Neuroscience
• Psychiatry
• Computer Science

Background:

• Eating disorders (EDs) are severe psychiatric conditions with significant mortality.
• Neuroimaging and machine learning (ML) offer potential for improved diagnosis and treatment prediction in EDs.

Purpose of the Study:

• To systematically review the application of ML techniques to neuroimaging data in anorexia nervosa (AN), bulimia nervosa (BN), and binge eating disorder (BED).

Main Methods:

• A systematic literature search was conducted on PubMed and APA PsycINFO (2014-2024) following PRISMA guidelines.
• Included studies used neuroimaging and ML on human subjects with AN, BN, or BED.
• Quality was assessed using GRADE and ROBINS-I tools.

Main Results:

• Five studies met inclusion criteria, primarily using support vector machines (SVMs) or other supervised ML models with structural MRI or diffusion tensor imaging.
• Observed cortical thickness and diffusion metrics differentiated ED subtypes.
• Studies were observational, heterogeneous, small in sample size, and had moderate to serious risk of bias, lacking external validation.

Conclusions:

• ML in neuroimaging holds potential for enhancing ED characterization and predicting outcomes.
• Methodological limitations currently hinder generalizability.
• Future research requires larger, multicenter, multimodal studies to improve clinical applicability.