CNN with machine learning approaches using ExtraTreesClassifier and MRMR feature selection techniques to detect liver diseases on cloud

Abstract

Liver disease is a significant global burden on health, with about a few hundred million people suffering from chronic liver disease (CLD), with approximately 2 million deaths each year. Liver diseases are tough to identify and usually ignored in the early stages as it does not show any symptoms. The liver disease diagnosis in the early stage will help to take precautions to prevent future illness. Generally, recognition of people with liver illness is accomplished via liver biopsy and visual assessment of MRI by experienced professionals, which is a laborious and time-consuming practice. As a result, there is a need for the development of an automated detection method that can offer results with minimal and greater precision. The primary motivation of this work is to implement a machine learning (ML) based real-time liver diseases classification framework onto the cloud to reduce clinicians' burden. The Indian Liver Patient Dataset (ILPD) was applied to classify liver diseases. The dataset has eleven attributes or features employed to train the models. The Convolutional Neural Network (CNN) was implemented and then the flatten layer output was given to the Logistic regression (LR), Random Forest (RF), and Support Vector Machine (SVM) classifier and achieved a precision of 100 percent for all models. The ExtraTreesClassifier (ETC) and Maximum Relevance Minimum Redundancy (MRMR) techniques were applied to select the features extracted by CNN and achieved remarkable 100 percent precision. The stratified K-fold method was used to evaluate the model performance. The comparative results confirm that the CNN-RF outperforms the literature-reported models. After the evaluation, the model was deployed successfully to the Platform-as-a-Service (PaaS) cloud.