Automated Liver Disease Classification Via Modified ResNet-50 Architecture on CT Images

Abstract

The classification of liver diseases is fundamental for the prompt identification and management of hepatic illnesses. In recent years, many computer-aided diagnostic systems for liver lesions have been developed based on deep learning techniques. This paper describes the design and assessment of the Automated classification of liver diseases with the use of Computed tomography (CT) imaging using a Modified residual convolutional neural network (ResNet-50) model. The dataset comprised many classifications of liver tissues, including cirrhosis, benign tumors, malignant tumors, and normal liver cells.  The model performed training, validation, and testing, attaining excellent results with a training accuracy of 99.1%, validation accuracy of 96.4%, and test accuracy of 99.3%.  This precision surpasses that of several leading approaches.  The findings achieved with the proposed framework illustrate the effective execution of the experiment for practical application in liver tumor screening. By augmenting the accuracy of diagnostics, this research addresses Goal 3: Good Health and Well Being. It also encourages new developments in AI technologies and medical imaging, which shifts the focus to the integration of AI powered systems within health care for achieving Goal 9: Industry, Innovation and Infrastructure and subsequently, the development of medical technology and healthcare globally.