Anorexia nervosa (AN) is one of the most severe and persistent mental health disorders, characterized by profound disturbances in eating behavior and food-related cognition. Behavioral symptoms such as restrictive eating, meal avoidance, rigid eating patterns, prolonged meal duration, and excessive control over food intake are central features of the disorder and are closely linked to illness severity, treatment response, and relapse risk.
Despite the critical role of eating behavior in AN, current clinical assessment remains largely dependent on self-report measures, retrospective questionnaires, food diaries, and clinical interviews. Although these approaches provide valuable insights into patients’ experiences, they are limited by recall bias, social desirability effects, and the considerable cognitive and emotional burden they place on individuals with eating disorders. Importantly, they provide only intermittent snapshots of eating behavior and do not capture how patients actually eat in their daily lives.
Recent advances in artificial intelligence (AI) and wearable sensing technologies create new opportunities for objective, continuous, and ecologically valid assessment of eating behavior. Wrist-worn inertial measurement unit (IMU) sensors offer a promising approach to unobtrusively monitor hand-to-mouth movements associated with eating without the privacy concerns, stigma, or practical limitations associated with camera- or audio-based monitoring systems. However, existing wearable-based eating detection approaches have primarily been developed in healthy populations and controlled environments, with a strong focus on classification accuracy rather than clinical applicability, uncertainty estimation, and behavioral interpretability.
This PhD project aims to address this critical gap by developing an AI-enhanced wearable system for automated tracking and characterization of eating behavior in individuals with AN. The project will combine wearable sensing, advanced signal processing, machine learning, and longitudinal behavioral analysis to establish clinically meaningful digital biomarkers of eating behavior. These biomarkers will quantify fine-grained characteristics of eating patterns, including eating rate, temporal organization, behavioral rigidity, variability, and changes over time.
By integrating technology development with clinical expertise, this project seeks to enable objective monitoring of eating behavior in real-world settings and provide new tools for early identification of behavioral deterioration, treatment response, and recovery trajectories in AN.
As a PhD researcher, you will:
- Design, optimize, and validate data acquisition protocols for wearable-based eating behavior monitoring.
- Develop robust signal processing and machine learning pipelines for extracting eating-related behavioral patterns from IMU sensor data.
- Develop interpretable digital biomarkers that capture key micro-structural properties of eating behavior in AN, including eating speed, temporal regularity, rigidity, and behavioral variability.
- Investigate relationships between wearable-derived biomarkers and clinical outcomes, including illness severity, treatment progress, and recovery, with particular attention to within-person changes over time.
- Collaborate closely with clinicians, psychologists, and researchers in a multidisciplinary environment.
- Disseminate research findings through international peer-reviewed publications, scientific conferences, and outreach activities.
