Multimodal AI and Deep Learning in the Early Detection of Metabolic Dysfunction-Associated Steatotic Liver Disease: A Robust Multi-Cohort Evaluation

Metabolic Dysfunction-Associated Steatotic Liver Disease (MASLD) is characterized by hepatic steatosis in the presence of key metabolic risk factors such as obesity, type2 diabetes mellitus (Type2DM), hypertension (HTN), or hypothyroidism and the absence of significant alcohol intake, affecting approximately 35% of the global adult population, with nearly 30% progressing to liver cirrhosis. Early-stage hepatic steatosis is typically asymptomatic, however as the disease progresses, patients may develop symptoms such as weight loss, itchy skin, fatigue and nausea. This study aims to perform a comprehensive analysis of early MASLD detection across all the risk factors using Multimodal Convolutional Neural Networks (CNN) and AI Models across multiple cohorts of ultrasound (US) and Electronic Medical Records (EMR) data. The multimodal GPT-5.2 AI model demonstrated an AUROC of 0.85 (95% CI 0.78–0.93) for the cross-platform test set and an AUROC of 0.82 (95% CI 0.68–0.93) for the Site-Independent Test Set. The GPT-5.2 model demonstrated strong performance across diverse datasets, indicating its potential utility for large-scale MASLD detection in clinical healthcare settings.