Constructing a Fatty Liver Prediction Model Based on Physical Examination Data and Machine Learning Algorithm

Read full text Summarise with VIRAT
Authors: Tao Meiyu, Huang Jiehong, Qin Qingrong, Yao Shunhan, Li Zi, Yao Jianmin, Qi Yue, Zeng Lanqing, Wu Xiangjun
Cited by: 0
References: 10

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Fatty liver disease is one of the most common chronic liver diseases and is closely associated with metabolic disorders. Early identification of high-risk populations is essential for prevention and intervention. Based on large-scale physical examination data, this study aimed to identify predictors of fatty liver disease and construct a prediction model using machine learning algorithms. A total of 4,700 individuals who underwent physical examinations at the Health Management Center of Guangxi Medical University Kaiyuan Langdong Hospital from 2020 to 2023 were retrospectively included and divided into a fatty liver group (1,372 cases) and a normal group (3,328 cases) according to abdominal ultrasound results. Demographic characteristics and biochemical indicators, including age, sex, body mass index (BMI), alanine aminotransferase (ALT), uric acid (UA), fasting plasma glucose (FPG), total cholesterol (TC), triglycerides (TG), high-density lipoprotein cholesterol (HDL-C), and low-density lipoprotein cholesterol (LDL-C), were collected. Univariate analysis and multivariate logistic regression were used to identify independent risk factors for fatty liver disease, and five machine learning models, including logistic regression, support vector machine (SVM), random forest, k-nearest neighbors (KNN), and neural network, were constructed to predict fatty liver risk. The predictive performance of each model was evaluated using accuracy, sensitivity, specificity, F1 score, and area under the receiver operating characteristic curve (AUC). The results showed that age, sex, BMI, ALT, UA, FPG, and TG were independent risk factors for fatty liver disease. Among the five models, logistic regression achieved the best performance with an accuracy of 0.7675 and an AUC of 0.7975. A nomogram based on the logistic regression model was further established and showed good calibration and predictive consistency. These findings suggest that a prediction model based on routine physical examination indicators can effectively assess the risk of fatty liver disease and may provide a convenient tool for early screening and risk assessment in health examination populations.

Fields of Study & Subjects

Liver Disease Diagnosis and Treatment Artificial Intelligence in Healthcare Alcohol Consumption and Health Effects

Publication details

Journal / source
MEDS Public Health and Preventive Medicine
Volume
6
Issue
1
Publication date
1 January 2026
Language
EN