Machine Learning-Based Liver Cirrhosis Prediction: A Review of

Feature Selection, Optimization Algorithms, Clinical

Applications, and Future Challenges

P. K. Dutta1,*

1 School of Engineering and Technology, Amity University Kolkata, India

Received: January 14, 2025 Revised: March 03, 2025 Accepted: May 06, 2025 ⋆ Corresponding author

ABSTRACT

Machine learning-based prediction has become an important research direction in the analysis of liver cirrhosis, where

clinical, laboratory, and demographic data can be used to support early risk assessment, disease-stage identification,

and outcome prediction. Liver cirrhosis represents a progressive and life-threatening condition associated with

chronic hepatic injury, fibrosis, portal hypertension, liver-function deterioration, and several severe complications

that may reduce patient survival and quality of life. Conventional diagnostic and prognostic approaches remain

clinically valuable; however, they may be limited when dealing with complex, nonlinear, heterogeneous, and highdimensional

patient data. This review examines the role of machine learning in liver cirrhosis prediction by analyzing

how supervised learning, ensemble models, deep learning, feature-selection strategies, and optimization-based

approaches contribute to improving classification accuracy, reducing irrelevant clinical variables, and enhancing

model reliability. The review highlights major predictive tasks, including cirrhosis diagnosis, disease severity

classification, survival prediction, complication-risk estimation, and treatment-outcome assessment. It also discusses

how machine learning can assist clinical decision-making by identifying hidden relationships among biochemical

markers, patient characteristics, and disease progression patterns that may not be easily captured through traditional

statistical methods. Despite these advantages, the implementation of machine learning models for liver cirrhosis

prediction remains associated with several challenges, including limited dataset size, class imbalance, missing

clinical values, insufficient external validation, model interpretability concerns, data heterogeneity, and the need for

clinically transparent decision-support systems. Overall, this review shows that machine learning-based prediction

can provide a valuable computational framework for improving liver cirrhosis assessment when it is developed

using reliable datasets, robust validation procedures, explainable modeling strategies, and clinically meaningful

performance evaluation.

Keywords: Machine learning Liver cirrhosis Disease prediction Feature selection Clinical decision support

1. INTRODUCTION

Liver cirrhosis is a progressive and clinically serious condition

that develops as a consequence of sustained hepatic

injury, long-term inflammation, and irreversible fibrotic remodeling

of liver tissue. As the disease advances, normal

hepatic architecture is gradually replaced by scar tissue and regenerative

nodules, leading to impaired liver function, portal

hypertension, metabolic dysregulation, and increased susceptibility

to life-threatening complications. The clinical