Review of machine learning solutions for eating disorders

Area of Science:

• Psychiatry
• Computer Science
• Medical Informatics

Background:

• Eating Disorders (EDs) present complex challenges in diagnosis, treatment, and recovery.
• Current methods face limitations in early detection, leading to significant health and psychosocial impairments.

Purpose of the Study:

• To review current Machine Learning (ML) and Artificial Intelligence (AI) applications in Eating Disorder (ED) clinical management.
• To bridge the knowledge gap between ED researchers and AI practitioners regarding state-of-the-art AI applications and their limitations.

Main Methods:

• Narrative review of AI/ML applications in various ED use-cases.
• Comparative analysis of AI techniques based on complexity, flexibility, functionality, explainability, and adaptability.

Main Results:

• AI/ML has been applied to ED risk factor identification, incidence prediction (including social media analysis), diagnosis, patient monitoring, and treatment response prediction.
• Analysis considered various AI techniques and their suitability for healthcare settings.

Conclusions:

• Existing ML and Causality methods face restrictions, including insufficient high-quality data and the need for flexible, explainable, and trustworthy AI models.
• Future AI development in ED clinical management requires careful AI model selection and collaborative efforts to build robust datasets and secure AI frameworks.