Super-resolution hepatobiliary MRI reconstruction improves radiomic prediction of microvascular invasion in liver cancer

Super-resolution hepatobiliary-phase magnetic resonance imaging improved radiomics prediction of microvascular invasion in liver cancer: area under the curve 0.827 vs 0.614 for original images (p<0.001) in 544 patients across two centers.

• Retrospective two-center study included 544 patients with hepatocellular carcinoma: 450 training, 51 internal validation, and 43 external test.
• Radiomics models used lesion-aware 3D Restormer super-resolution of hepatobiliary-phase MRI, nnU-Net segmentation, and 50 LASSO-selected features.
• Original-image model AUCs were 0.710, 0.764, and 0.614; super-resolution model AUCs were 0.855, 0.798, and 0.827 in training, validation, and test sets.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.