Machine learning advances dietary and liver disease detection

Researchers are utilizing machine learning to improve the identification of dietary patterns and liver disease through existing medical data. One study focused on developing machine learning computable phenotypes to identify high-fat diets (HFD) using variables typically found in electronic health records (EHRs), such as demographics, comorbidities, and laboratory results. By training models like random forest on data from the National Health and Nutrition Examination Survey, researchers were able to successfully classify high-fat dietary patterns, which can help bridge the gap in diet-disease research where detailed dietary information is often missing from clinical records.

In a separate development, a machine learning approach using MRI radiomics has shown high accuracy in the noninvasive detection of nonalcoholic fatty liver disease (NAFLD). By extracting radiomics features from liver MRI images and employing a TabPFN classifier, researchers achieved an accuracy of 95.65% in classifying patients with the disease. This method utilizes clinical metadata such as age, BMI, and sex alongside image-based features to enhance noninvasive liver disease assessment.