Applications of Machine Learning in Early Identification and Prediction of Non-Alcoholic Fatty Liver Disease: A Systematic Review and Meta-Analysis (Preprint)

BACKGROUND Non-alcoholic fatty liver disease (NAFLD) is one of the leading causes of chronic liver disease worldwide. Research based on artificial intelligence (AI) and machine learning (ML) algorithms has increasingly been applied to the identification and diagnosis of NAFLD. Early screening for NAFLD is beneficial in reducing the progression to inflammation, liver fibrosis, and cirrhosis. To date, no study has quantitatively evaluated the application of ML algorithms in the early identification and prediction of NAFLD. Therefore, this systematic review and meta-analysis is conducted to comprehensively assess the predictive accuracy of ML in early NAFLD detection. OBJECTIVE To systematically evaluate and quantify the performance of ML techniques in the early identification, screening, risk prediction, and diagnosis of patients with NAFLD. METHODS A literature search was performed in the PubMed, Web of Science, Cochrane Library, and EMBASE databases from January 2020 to March 2025. The quality of studies and risk of bias were assessed via the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) and the Prediction Model Risk of Bias Assessment Tool (PROBAST). Studies reporting AUC, sensitivity, and specificity were included. Data analysis was conducted using R (version 4.3.1) and Stata (version 17.0). This study adhered to the PRISMA guidelines. RESULTS A systematic review and bias assessment were performed on 23 articles, with 19 outcomes from 17 studies included in the meta-analysis. The pooled sensitivity and specificity were 0.81 [95% CI 0.76–0.85] and 0.80 [95% CI 0.76–0.84], respectively. The area under the summary receiver operating characteristic (SROC) curve (AUC) was 0.88 [95% CI 0.84–0.90]. Significant heterogeneity was observed among studies, with meta-regression indicating that reference standards and cross-validation were potential sources of heterogeneity. CONCLUSIONS ML demonstrated robust performance in the early identification and prediction of NAFLD, showing promise in assisting clinicians with screening and early diagnosis. However, further research is necessary due to the lack of external validation and limited exploration of certain algorithm types.