Liver fibrosis processing, multiclassification, and diagnosis based on hybrid machine learning approaches

Abstract

The cirrhosis and cirrhosis-related problems are connected to the degree of fibrosis in the liver. The purpose of this paper is to propose an automated method for identifying liver fibrosis using ultrasound shear wave elastography (700) images that is based on a hybrid machine learning approach using a convolutional neural network (CNN) with two types of classifier (SoftMax and support vector machine (SVM)). The dataset gathered from hospitals is used in the training and testing phases of the model. The objective is to develop a hybrid machine learning model that can classify images based on their stage of fibrosis. The suggested system comprises three stages. The first is the preprocessing step, which starts with countor detection and continues with the "contrast limited adaptive histogram equalization (CLAHE)" technique to show the properties of liver tissue. In the second step, the CNN algorithm was utilized, which was based on several images to extract deep features and identify shear wave elastography (SWE) samples. In the third step, SVM and SoftMax functions are used to classify liver fibrosis. A five-class model (normal, F1, F2, F3, and F4) was developed. The result illustrates how successfully the CNN-SoftMax and CNN-SVM classifiers classified liver fibrosis in the test dataset, with 97.18% and 98.59% accuracy, respectively.