Integrated CT-MRI deep learning plus clinical features improves preoperative HCC grading

An integrated CT-MRI deep learning model plus clinical features achieved AUCs of 0.89-0.90 for preoperative hepatocellular carcinoma (HCC) grading, significantly outperforming multimodality alone (0.72-0.82) and clinical model (0.59-0.73) in 250 patients across three institution…

• Multicenter retrospective study with 250 patients having paired CT and MRI from three institutions, including external validation.
• Multimodality CT-MRI deep learning alone outperformed the best single-modality CT model (AUC 0.58-0.76) and MRI model (AUC 0.62-0.72).
• Decision curve analysis confirmed the integrated model provided the greatest net clinical benefit.

Automated summary

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