Accurate liver fibrosis detection through hybrid mrmr bilstm cnn architecture with histogram equalization and optimization

Abstract

Liver fibrosis is the gradual scarring of liver tissue caused by prolonged liver injury and inflammation. Major contributors to liver fibrosis include chronic viral hepatitis, autoimmune hepatitis, and exposure to hepatotoxic substances. Early detection and intervention are vital to halt fibrosis progression and mitigate severe liver-related complications. Treatment approaches primarily target the underlying cause of liver injury, symptom management, and disease progression prevention through lifestyle adjustments, medication, and occasionally, liver transplantation. In Phase 1, DenseNet-169, a deep convolutional neural network (DCNN), is utilized to diagnose liver fibrosis by classifying liver images into three categories: normal, hepatitis, and cirrhosis. The study employs the UCI dataset, consisting of 583 liver disease records across various age groups (21 to over 61 years old), divided into training (70%) and testing (30%) sets. Prior to training, images undergo preprocessing, including histogram equalization for noise and anomaly removal. Feature extraction and data selection via student t-test enhance image quality. Enhanced images are then classified using DenseNet-169, achieving an impressive 98% accuracy when evaluated against diverse performance metrics, surpassing other state-of-the-art models. In Phase 2, the timely detection as well as diagnosing of liver fibrosis are paramount for efficient medical intervention, given its progressive nature and the potential risks and costs associated with invasive procedures like biopsies. newline