A Review of Artificial Intelligence-Based Approaches for Non-Invasive Liver Disease Diagnosis

Liver disease is a significant clinical problem on a global scale and a leading cause of morbidity and mortality. Early and prompt diagnosis is essential to successful management, but the traditional diagnosis methods tend to require invasive procedures, or they are restricted by the limitations of varying observers. In recent years, the use of artificial intelligence (AI) and computer-aided diagnosis (CAD) systems has acquired significant popularity in the field of hepatology. Machine learning (ML) and deep learning (DL) technologies are actively used in medical imaging data, including ultrasound, CT, and MRI, to assist clinicians in identifying and classifying liver pathologies, i.e., fatty liver disease, cirrhosis, and hepatocellular carcinoma. The review is a synthesis of the recent developments in image processing, feature detection, and classification algorithms of liver disease diagnosis by AI. The most common ML algorithms include Support Vector Machines, random forests, decision trees, naive bayes, and K-nearest neighbors, in many cases using radiomic features derived using imaging data. Although deep learning models, especially convolutional neural networks and transfer learning implementations, are highly sensitive and highly perform in segmentation and classification tasks, traditional ML systems with radiomic features are frequently able to offer robust and efficient solutions to resource-bound environments. Even though these results are promising, there are still a number of challenges such as data heterogeneity, insufficient multi-center validation, and model interpretability. To reliably translate clinical findings into clinical practice and enhance patient outcomes, future research should focus on large-scale validation studies, multimodal data integration, and explainable AI frameworks.