Diagnosis and grading of steatotic liver disease via clinical and laboratory data using machine learning

Steatotic liver disease (SLD) is one of the most common chronic liver conditions worldwide and may lead to severe complications if not diagnosed and managed promptly. Its high spread, particularly in the MENA region, highlights the urgent need for early, accessible, and cost-effective diagnostic strategies. While conventional diagnostic methods such as imaging and biopsy offer high accuracy, they are invasive, costly, and impractical for large-scale screening efforts. In contrast, clinical and laboratory data represent a non-invasive, readily available, and inexpensive alternative for SLD detection. In this study, a machine learning-based approach was developed to diagnose and grade SLD using such routinely collected clinical and biochemical features. Various machine learning algorithms were applied, with SVM-XGBOOST achieving an accuracy of 90% using only eight features. Following expert consultation with gastroenterologists, four clinically relevant features, including Ferritin, Fasting Blood Glucose, Triglycerides, and Body Mass Index, were identified. Using these four features alone, the model retained a classification accuracy of 70%. These findings indicate that the integration of machine learning techniques with clinical and laboratory data can substantially improve diagnostic accuracy and efficiency. This approach offers a scalable and non-invasive solution for SLD screening, particularly in resource-limited or underserved healthcare settings, thereby facilitating early detection and timely intervention.