A framework for identification and classification of liver diseases based on machine learning algorithms.

Area of Science:

• Hepatology and Medical Informatics

Background:

• Hepatocellular carcinoma (HCC) is a leading cause of cancer death, with rising incidence rates globally.
• Accurate diagnosis is crucial for patient care but advanced imaging is often inaccessible, particularly in low-income regions.
• There's a need for accessible diagnostic frameworks for early liver disease detection using basic clinical data.

Purpose of the Study:

• To identify significant risk factors for liver diseases using machine learning algorithms.
• To develop a predictive model for early liver disease detection based on routine blood tests.
• To evaluate the performance of various machine learning classifiers in liver disease classification.

Main Methods:

• Utilized machine learning algorithms including regularized regression, logistic regression, random forest, decision tree, and extreme gradient boosting.
• Analyzed clinical data from 525 patients to extract significant risk factors.
• Compared the performance of five distinct machine learning classifiers.

Main Results:

• The Random Forest classifier achieved the highest performance with an accuracy of 0.762, recall of 0.843, F1-score of 0.775, and AUC of 0.999.
• Identified 14 significant risk factors, with Total bilirubin, Gamma-glutamyl transferase (GGT), and Direct bilirubin being the most critical.
• Established an order of importance for risk factors including hemoglobin, age, platelet count, and liver enzymes.

Conclusions:

• Machine learning classifiers can effectively aid in the early detection and classification of liver disease.
• The identified risk factors provide valuable insights for disease prevention and treatment strategies.
• This AI-driven approach offers a cost-effective solution for liver disease diagnosis in underserved areas.