Application of Machine Learning Techniques for Clinical Predictive Modeling: A Cross-Sectional Study on Nonalcoholic Fatty Liver Disease in China

• Published in: BioMed research international Vol. 2018; no. 2018; pp. 1 - 9
• Main Authors: Yu, Chaohui, Shen, Zhe, Li, Youming, Ma, Han, Li, You-ming
• Format: Journal Article
• Language: English
• Published: Cairo, Egypt Hindawi Publishing Corporation 01.01.2018; Hindawi; John Wiley & Sons, Inc
• Subjects: Adult; Alanine; Alanine transaminase; Alanine Transaminase - blood; Algorithms; Artificial intelligence; Bayesian analysis; Big Data; Biomarkers; Biomedical research; Body mass; Body Mass Index; China; Classification; Colleges & universities; Computer science; Cross-Sectional Studies; Data collection; Data mining; Diabetes; Diagnostic imaging; Diagnostic systems; Fatty liver; Female; gamma-Glutamyltransferase - blood; Hepatitis; Humans; Information systems; Laboratory tests; Learning algorithms; Liver; Liver diseases; Machine Learning; Male; Medical research; Medicine, Experimental; Methods; Middle Aged; Models, Biological; Non-alcoholic Fatty Liver Disease - blood; Non-alcoholic Fatty Liver Disease - pathology; Non-alcoholic Fatty Liver Disease - physiopathology; Pattern recognition; Physical examinations; Prediction models; Public software; Regression models; Screening; Software; Statistical analysis; Statistical methods; Studies; Triglycerides; Triglycerides - blood; Type 2 diabetes; Ultrasound; Uric acid; γ-Glutamyltransferase; China; Japan
• ISSN: 2314-6133; 2314-6141; 2314-6141
• DOI: 10.1155/2018/4304376

Background. Nonalcoholic fatty liver disease (NAFLD) is one of the most common chronic liver diseases. Machine learning techniques were introduced to evaluate the optimal predictive clinical model of NAFLD. Methods. A cross-sectional study was performed with subjects who attended a health examination at the First Affiliated Hospital, Zhejiang University. Questionnaires, laboratory tests, physical examinations, and liver ultrasonography were employed. Machine learning techniques were then implemented using the open source software Weka. The tasks included feature selection and classification. Feature selection techniques built a screening model by removing the redundant features. Classification was used to build a prediction model, which was evaluated by the F-measure. 11 state-of-the-art machine learning techniques were investigated. Results. Among the 10,508 enrolled subjects, 2,522 (24%) met the diagnostic criteria of NAFLD. By leveraging a set of statistical testing techniques, BMI, triglycerides, gamma-glutamyl transpeptidase (γGT), the serum alanine aminotransferase (ALT), and uric acid were the top 5 features contributing to NAFLD. A 10-fold cross-validation was used in the classification. According to the results, the Bayesian network model demonstrated the best performance from among the 11 different techniques. It achieved accuracy, specificity, sensitivity, and F-measure scores of up to 83%, 0.878, 0.675, and 0.655, respectively. Compared with logistic regression, the Bayesian network model improves the F-measure score by 9.17%. Conclusion. Novel machine learning techniques may have screening and predictive value for NAFLD.