Facial expression analysis and machine learning for objective assessment of alcohol craving in alcohol use disorder:...

Alcohol craving is a key predictor of relapse in alcohol use disorder (AUD), yet current assessments rely mainly on self-reported scales, lacking objective evaluation methods. This single-center observational study aims to explore subtle differences in facial expressions among alcohol use disorder patients during craving states, long-term abstinent individuals, and healthy controls, with the goal of identifying objective biomarkers of alcohol craving. A secondary aim is to establish a rigorous craving evaluation system using machine learning.

We plan to recruit 200 participants per group: alcohol use disorder patients, long-term abstinent individuals, and healthy controls. Patients with alcohol use disorder will first undergo inpatient detoxification (approximately two weeks) and will be eligible once their Clinical Institute Withdrawal Assessment for Alcohol, Revised (CIWA-AR) score falls below 7. At enrollment, participants will complete psychological and clinical assessments, including sociodemographic and drinking history questionnaires, the Alcohol Use Disorders Identification Test (AUDIT), Penn Alcohol Craving Scale (PACS), Hamilton Anxiety Rating Scale (HAM-A), Hamilton Depression Rating Scale (HAM-D), Barratt Impulsiveness Scale, and Pittsburgh Sleep Quality Index (PSQI). Personalized drinking-environment preferences will be collected via semi-structured interviews. During the experiment, participants will provide craving ratings using a visual analog scale (VAS) before and after viewing a 120-second relaxation video and a 120-second alcohol cue-related video, while facial expressions are recorded simultaneously. The alcohol use disorder group will undergo the same assessments again after six weeks of inpatient treatment.

Developing an objective system for evaluating alcohol craving has the potential to enhance routine screening, differential diagnosis, and treatment monitoring in alcohol use disorder. Furthermore, integrating facial expression analysis with machine learning may enable the development of reliable craving assessment tools and treatment response prediction models. Such approaches could provide clinicians with evidence-based guidance for psychological and psychosocial interventions, ultimately reducing relapse rates among individuals with alcohol use disorder.