A Hybrid Deep Learning Approach Combining CNN and BiLSTM for Improved Early Detection and Classification of Liver Diseases

Abstract

A wide range of conditions, such as fatty liver and cancer contributes to global illness and mortality rates. Although early detection is the solution, it is often difficult owing to the absence of noticeable early symptoms. The healthcare industry has seen significant progress via Artificial Intelligence (AI) approach. This research proposes a novel automated prediction of liver disease more effectively. The input liver data is initially preprocessed where it undergoes data cleaning to remove missing values and irrelevant information. Following, this feature selection is performed using select K-Best method to rank and choose the predictive features, by enhancing model's efficiency. The classification stage is performed using Convolutional Neural Network- Bidirectional Long Short-Term Memory (CNN-BiLSTM), where CNN automatically extract high-level features, while BiLSTM captures sequential dependencies, modeling both past and future information in data. This combined approach, leads to significant improvements in classification accuracy.