Can machine learning models help doctors diagnose my metabolic dysfunction-associated steatotic liver disease?

Machine learning (ML) is a type of artificial intelligence that can analyze large amounts of data to find patterns. For metabolic dysfunction-associated steatotic liver disease (MASLD), ML models are being developed to help doctors diagnose the condition and predict serious outcomes like liver cancer. These models use information from blood tests, ultrasound images, and other clinical data. Research shows that ML tools can be quite accurate, sometimes outperforming traditional methods, but they are not yet a replacement for a doctor's full evaluation.

What the research says

Several studies have tested ML models for MASLD. A 2024 study developed an ML model to predict liver cancer (hepatocellular carcinoma) in people with MASLD. The model used standard lab tests and clinical information, and it found that liver fibrosis (scarring) was the strongest predictor of cancer risk ⁹. Another 2024 study created a hybrid ML model that analyzes ultrasound images to classify fatty liver as normal, mild, moderate, or severe. This model achieved 91.34% accuracy, which was better than traditional neural networks ¹¹. A 2025 study combined genetic testing with ML algorithms (like random forest and logistic regression) to screen for MASLD. They found that a specific mitochondrial DNA variant (mt12361A>G) increased the risk of MASLD, and their ML model showed good performance for screening ¹⁰. A systematic review and meta-analysis from 2025 looked at many ML and deep learning models for diagnosing MASH (the inflammatory form of MASLD) and liver fibrosis. The review found that ML models had a pooled accuracy (AUROC) of 0.833 for diagnosing MASH, and deep learning models had an AUROC of 0.841, indicating strong diagnostic performance ². These results suggest that ML tools can assist doctors in diagnosing MASLD and identifying patients at higher risk for complications.

What to ask your doctor

• Are there any machine learning tools or risk calculators available at your clinic to help assess my MASLD risk?
• How do ML-based predictions compare with standard non-invasive tests like the Fibrosis-4 score or ultrasound?
• Could an ML model help determine how often I should be screened for liver cancer?
• Are there any ongoing studies using AI for MASLD that I might be eligible for?
• What are the limitations of current ML models for MASLD, and how do you factor them into your clinical decisions?