Liver Cirrhosis Stage Estimation from MRI with Deep Learning

TL;DR

This paper addresses the critical need for early and accurate liver cirrhosis staging from MRI by proposing an end-to-end deep learning framework that analyzes multi-sequence (T1W and T2W) MRI. It introduces sequence-specific encoders with cross-sequence attention to fuse information, and evaluates six state-of-the-art architectures on CirrMRI600+ (628 scans from 339 patients), comparing them against radiomics baselines. The best results show 72.8% accuracy on T1W (VGG-19) and 63.8% on T2W (MambaVision-T), establishing a new benchmark for automated cirrhosis staging and revealing that deeper models do not always guarantee better performance. The work highlights the potential for clinical deployment, while noting limitations in moderate-stage discrimination and suggesting future directions including 3D architectures, clinical metadata integration, longitudinal data, and foundation models tailored for medical imaging.

Abstract

We present an end-to-end deep learning framework for automated liver cirrhosis stage estimation from multi-sequence MRI. Cirrhosis is the severe scarring (fibrosis) of the liver and a common endpoint of various chronic liver diseases. Early diagnosis is vital to prevent complications such as decompensation and cancer, which significantly decreases life expectancy. However, diagnosing cirrhosis in its early stages is challenging, and patients often present with life-threatening complications. Our approach integrates multi-scale feature learning with sequence-specific attention mechanisms to capture subtle tissue variations across cirrhosis progression stages. Using CirrMRI600+, a large-scale publicly available dataset of 628 high-resolution MRI scans from 339 patients, we demonstrate state-of-the-art performance in three-stage cirrhosis classification. Our best model achieves 72.8% accuracy on T1W and 63.8% on T2W sequences, significantly outperforming traditional radiomics-based approaches. Through extensive ablation studies, we show that our architecture effectively learns stage-specific imaging biomarkers. We establish new benchmarks for automated cirrhosis staging and provide insights for developing clinically applicable deep learning systems. The source code will be available at https://github.com/JunZengz/CirrhosisStage.

Figures (1)

• Figure 1: T1W and T2W MRI scans of cirrhotic patients are shown for mild, moderate, and severe cases.