Abstract
Early detection of liver fibrosis in chronic hepatitis B (CHB) patients is crucial for improving their prognosis. This study aims to develop a machine learning (ML)-based diagnostic model for evaluating the degree of liver fibrosis in patients with CHB. The present study comprised 268 patients with CHB who underwent liver biopsy from January 2022 to May 2023. Liver fibrosis was staged according to the Scheuer scoring system. The dataset is divided into a training set and a validation set in a ratio of 7.5:2.5. The feature selection process was executed through the utilization of least absolute shrinkage and selection operator regression, and 8 ML algorithms (including random forest [RF], extreme gradient enhancement, support vector machine, etc) were employed for the prediction of the performance of the significant liver fibrosis assessment model. The RF model performs best among ML models, with an area under the curve value of 0.810 for the training set and 0.793 for the validation set. Decision curve analysis indicates that the RF model exhibits the highest net benefit under most threshold probabilities. When the probability threshold is 30%, the sensitivity and net benefit are the highest, significantly outperforming other traditional scores. RF ML models can effectively assess liver fibrosis in patients with CHB. Compared with traditional indicators, they can become a safe, effective and more personalized screening method, enabling early and dynamic risk decision-making.