An effective approach for early liver disease prediction using deep learning method with immunity-based Boosted Ebola optimization search algorithm

• Published in: Expert systems with applications Vol. 285; p. 127711
• Main Authors: Madhavi, A. Venu, Prasad, Srinivas
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.08.2025
• Subjects: Adaptive Level-Set Algorithm; Immunity-based Boosted Ebola Optimization Search Algorithm; Liver Disease Prediction; Vision Transformer based Dilated Residual DenseNet; Immunity-based Boosted Ebola Optimization Search Algorithm; Liver Disease Prediction; Adaptive Level-Set Algorithm; Vision Transformer based Dilated Residual DenseNet
• ISSN: 0957-4174
• DOI: 10.1016/j.eswa.2025.127711

Liver problems considerably affect liver function and are among the top causes of mortality in India. The primary factors contributing to liver diseases include alcohol consumption, inhalation of toxic gases and the intake of contaminated food, and medications. Accurate and precise diagnosis of liver conditions is crucial, as even minor errors in detecting liver disorders can lead to severe consequences. Early diagnosis of diseases of the liver presents significant hurdles for doctors. Research into the use of machine learning for the prediction and classification of liver disease is ongoing, aiming to enhance diagnostic accuracy. Despite the higher cost and longer development time, machine learning models offer promising prospects. Deep learning algorithms, in particular, represent a promising approach to aid physicians in the rapid and accurate diagnosis of liver diseases. Here, developed an effective deep learning-assisted liver disease prediction model to overcome the challenges in the conventional models. Initially, the raw image is collected using the standard source of data. Further are inputted into the liver disease segmentation process using the proposed Adaptive Level-Set Algorithm (A-LSA). Here, the suggested Immunity-based Boosted Ebola Optimisation Search Algorithm (IBEOSA) adjusts the parameters to improve accuracy and lower cost. The Vision Transformer-based Dilated Residual DenseNet (ViT-DRDNet) then completes the classification step. Lastly, the classified result was provided by the ViT-DRDNet model. The introduced model’s efficiency is validated with different conventional models and optimization algorithms to showcase its effectiveness.