Deep Learning-Enhanced Quantitative Ultrasound and Magnetic Resonance Elastography Analysis for the Precise Non-Invasive Staging of Liver Fibrosis and Inflammation

• Ada John

  ladoke Akintola university of technology

  Author

Chronic liver diseases (CLDs) affect approximately 25–30% of the global population, with liver fibrosis and inflammation serving as key pathological drivers of disease progression toward cirrhosis and hepatocellular carcinoma . Current gold-standard liver biopsy is invasive, suffers from sampling variability, and poses procedural risks, while existing non-invasive methods lack sufficient diagnostic accuracy for precise staging. This study addresses the critical research gap in simultaneous, accurate, non-invasive grading of fibrosis and inflammation by developing a deep learning-enhanced multimodal imaging framework integrating quantitative ultrasound (QUS) and magnetic resonance elastography (MRE) with clinical biomarkers. Using a retrospective dataset of 142 patients with biopsy-confirmed CLD, we employed a hybrid convolutional neural network (CNN) architecture with physics-informed regularization to fuse radiomic features from B-mode ultrasound, shear wave elastography, and MRE. The proposed framework achieved area under the curve (AUC) values of 0.91 for fibrosis stage ≥F4 and 0.93 for inflammation grade ≥A3, significantly outperforming single-modality approaches and conventional clinical indicators . The model maintained robust performance across intermediate fibrosis stages (F2: AUC 0.89, F3: AUC 0.91). These findings demonstrate that deep learning-enhanced multimodal elastography analysis provides a replicable, non-invasive framework for precise CLD staging, with direct implications for clinical decision-making, patient monitoring, and reduced biopsy dependence.

• pdf

Copyright (c) 2026 Ada John (Author)

This work is licensed under a Creative Commons Attribution 4.0 International License.