Radiomic Feature Analysis via MRI for Assessing Severity in Advanced Liver Disease

Overview

This study investigates the use of MRI-derived radiomic features from liver and spleen to predict disease severity in cirrhotic patients, approximated by MELD score and decompensation status. By focusing on a controlled single-center MRI dataset, the research aims to develop a noninvasive imaging biomarker to supplement existing clinical scoring systems.

Background

Cirrhosis, a leading cause of global mortality, results from chronic liver diseases of various etiologies and is characterized by architectural disruption of the liver. The Model for End-Stage Liver Disease (MELD) score is widely used to estimate mortality risk and prioritize liver transplantation. Radiomics, involving quantitative extraction of imaging features, has shown promise in oncology and fibrosis staging but faces challenges in MRI due to variability in acquisition and signal normalization. This study uniquely focuses on MRI radiomic features to assess cirrhosis severity and decompensation without relying on predefined imaging manifestations.

Data Highlights

The study retrospectively analyzed MRI scans from cirrhotic patients undergoing HCC screening at a single center, using uniform MRI acquisition protocols to ensure feature reproducibility. Radiomic features were extracted from T1-weighted images of liver and spleen. MELD scores, ranging from 6 to 40, served as the clinical reference for disease severity, with a score of 15 marking a critical threshold for transplant benefit. The cohort included patients with compensated and decompensated cirrhosis, defined by clinical complications such as variceal bleeding, ascites, or hepatic encephalopathy.

Key Findings

• Radiomic features from liver and spleen MRI can noninvasively predict cirrhosis severity as approximated by MELD score.
• These imaging biomarkers also show potential to distinguish between compensated and decompensated cirrhosis.
• Focusing on a single MRI scanner model and protocol minimized confounding variability, enhancing feature stability.
• The approach does not rely on subjective imaging findings but on quantitative texture and morphological heterogeneity metrics.
• This is the first study to propose an MRI-based radiomic signature specifically for liver cirrhosis severity assessment.

Clinical Implications

MRI-derived radiomic features could serve as adjunctive tools to improve risk stratification in cirrhotic patients, potentially guiding transplant prioritization beyond MELD scoring alone. Noninvasive imaging biomarkers may facilitate earlier detection of decompensation, enabling timely clinical interventions. Standardizing MRI acquisition protocols is critical to ensure reproducibility of radiomic analyses in clinical practice.

Conclusion

MRI-based radiomic analysis offers a promising noninvasive method to assess liver disease severity and decompensation in cirrhosis, complementing existing clinical scoring systems. Further validation could establish radiomic signatures as valuable biomarkers in managing advanced liver disease.

References

1. Yasaka et al. 2018 -- Deep learning with convolutional neural network for staging liver fibrosis using contrast-enhanced MR images
2. Choi et al. 2017 -- Liver fibrosis staging using contrast-enhanced CT images
3. Park et al. 2019 -- Liver fibrosis stage estimation using spleen-based intensity normalization
4. He et al. 2020 -- Radiomic features for predicting liver stiffness on T2-weighted MRI
5. UNOS 2016 -- Modification of MELD score to include serum sodium