Enhancing diagnostic accuracy for chronic liver disease through machine learning

The increasing incidence of chronic disorders highlights the serious necessity for immediate and accurate diagnostic techniques to remove the burden on healthcare systems, especially with respect to Chronic Liver Diseases (CLD) and Decompensated Chronic Liver Disease (DCLD). This study aims to create strong Machine Learning (ML) models to efficiently detect and classify CLD, DCLD, and non-diseased cases using a dataset of 51 patients with varied clinical, biochemical, and demographic profiles. A comparison of different ML algorithms, such as AdaBoost, K-Nearest Neighbours (KNN), Random Forest, Artificial Neural Networks (ANN), Support Vector Machines (SVM), and Extra Trees, was made to identify best method for this classification problem. AdaBoost achieved the highest observed classification accuracy of 91.11% among the evaluated models on available test data. However, this result should be interpreted within the scope of the study dataset and does not establish generalizability to independent patient populations. Data visualization and pre-processing were utilized which improving not only the evaluation of the model's performance but also pointing out areas to be improved. The study showcases the immense promise of ML in handling intricate data sets, ensuring that the data produced is precise, trustworthy, and scalable to be used for wider applications in medicine.