INTERPRETABLE AI-BASED MODEL FOR LIVER DISEASE DETECTION AND CLASSIFICATION USING BIOCHEMICAL MARKERS OF HOSPITAL-BASED DATA FROM LAHORE, PAKISTAN

Abstract

Purpose: Liver disease accounts for two million deaths annually. Liver Function tests have been widely used for the non-invasive diagnosis and interpretation of liver disease. This study aims to make predictive model of Artificial intelligence to find characteristic patterns in Liver function tests for detecting conditions like hepatitis, cirrhosis and Non-alcoholic fatty liver disease.

Methods: A total of 250 samples data were collected from different hospitals of Lahore and analysed on the cobas c 311 analyzer. SPSS 30 was used to calculate the statistical descriptives.  Diseases were classified into groups hepatitis, cirrhosis, and non-alcoholic fatty liver disease. These datasets were then integrated with AI model which was developed using Classification and Prediction Approach and Multiple machine learning algorithms, including decision trees, support vector machines, and ensemble learning models.

Results: Out of 250 samples being tested 215 (86%) showed abnormalities and 35 (14%) were normal. These results were interpreted and cross validated by patient records. After training of AI model, it showed strong performance in classifying liver diseases using LFT parameters. They accurately differentiate between hepatitis, cirrhosis, and non-alcoholic fatty liver disease

Conclusion: This study shows that artificial intelligence-based models can effectively use routine biochemical markers to classify liver disease.  The proposed framework supports early intervention, and provides a scalable solution for data-driven clinical decision support.