Metabolic dysfunction-associated steatotic liver disease (MASLD), formerly known as non-alcoholic fatty liver disease (NAFLD), and its progressive form, metabolic dysfunction-associated steatohepatitis (MASH), are liver manifestations of the metabolic syndrome. Their global prevalence is rising alongside obesity, highlighting the need for new therapeutic targets. In this work, a deep learning-based image analysis model was developed to quantify liver steatosis in histological whole-slide images of preclinical MASLD models. The algorithm accurately identifies parenchymal tissue while excluding background, fixation artifacts, and vessels, then detects and quantifies regions with microvesicular and macrovesicular steatosis. This approach was applied to study the role of HSD17B12 in the development of steatotic liver disease using mice with a hepatocyte-specific knockout of HSD17B12 (LiB12cKO). On a chow diet, these mice exhibited hepatic lipid accumulation with reduced whole-body fat, characterized predominantly by microvesicular steatosis, indicating impaired lipid droplet expansion. These changes were associated with hepatocellular hypertrophy and hepatic injury. A high-fat diet intervention was implemented, and subsequent morphological and functional alterations were evaluated using histological image analysis, transmission electron microscopy, and micro-ultrasound imaging. Notably, LiB12cKO mice were resistant to high-fat diet–induced obesity and hepatic steatosis, as demonstrated by image analysis and complementary biomarkers. Transcriptomic profiling further supported these findings, revealing significant changes in LiB12cKO males compared to wild-type controls. Taken together, these results suggest that inactivation of HSD17B12 alters liver morphology and lipid metabolism, but offers a potential approach to reduce diet-induced obesity.