AUTHOR=Wei Yichen , Gong Jie , He Xin , Liu Bo , Liu Tiejun , Yang Shuohui , Zhou Zhipeng , Liang Lingyan , Zhan Songhua , Xia Ziqiang , Duan Gaoxiong , Lin Bin , Han Qiuli , Li Shasha , Qin Wei , Pickhardt Perry J. , Deng Demao TITLE=An MRI-Based Radiomic Model for Individualized Prediction of Hepatocellular Carcinoma in Patients With Hepatitis B Virus-Related Cirrhosis JOURNAL=Frontiers in Oncology VOLUME=12 YEAR=2022 URL=https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2022.800787 DOI=10.3389/fonc.2022.800787 ISSN=2234-943X ABSTRACT=Objective

To develop and validate a radiomic nomogram for individualized prediction of hepatocellular carcinoma (HCC) in HBV cirrhosis patients based on baseline magnetic resonance imaging examinations and clinical data.

Methods

364 patients with HBV cirrhosis from five hospitals were assigned to the training, internal validation, external validation-1 or external validation-2 cohort. All patients underwent baseline magnetic resonance image (MRI) scans and clinical follow-up within three-year time. Clinical risk factors and MRI-based features were extracted and analyzed. The radiomic signatures were built using the radiomics-score (Rad-score) that calculated for each patient as a linear weighted combination of selected MRI-based features. Prognostic performances of the clinical and radiomic nomograms were evaluated with Cox modeling in the training and validation cohorts.

Results

Eighteen features were selected for inclusion in the Rad-score prognostic model. The radiomic signature from multi-sequence MRI yielded a concordance index (C-index) of 0.710, 0.681, 0.632 and 0.658 in the training, internal validation, external validation-1, external validation-2 cohorts, respectively. Sex and Child-Turcotte-Pugh (CTP) class were the most prognostic clinical risk factors in univariate Cox proportional hazards analyses. The radiomic combined nomogram that integrated the radiomic signature with the clinical factors yielded a C-index of 0.746, 0.710, and 0.641 in the training, internal validation, and external validation-1 cohorts, respectively, which was an improvement over either the clinical nomogram or radiomic signature alone.

Conclusion

We developed an MRI-based radiomic combined nomogram with good discrimination ability for the individualized prediction of HCC in HBV cirrhosis patients within three-year time.