Abstract
PURPOSE: To develop and validate a radiomics model that uses multiple magnetic resonance imaging (MRI) sequences to accurately distinguish hepatocellular carcinoma (HCC) from focal nodular hyperplasia (FNH), thereby improving diagnostic precision and decision-making. METHODS: We conducted a retrospective analysis including 196 patients (97 in HCC and 99 in FNH) diagnosed at the Zhangzhou Affiliated Hospital of Fujian Medical University (August 2011-December 2021). Radiomics features were extracted from the MRI images. LASSO logistic regression models were constructed for feature selection and to differentiate HCC from FNH. The model was further validated using a temporally independent cohort of 91 patients (49 HCC, 42 FNH) from the same institution (January 2022-December 2023). The area under the curve (AUC), accuracy, sensitivity, and specificity were used to evaluate the model's performance. RESULTS: We obtained 34 features for T2-weighted imaging (T2WI), diffusion-weighted imaging (DWI), and contrast enhanced imaging (CEI). The radiomics model demonstrated high diagnostic performance, with AUCs of 0.992 and 0.958 in the training and internal validation, respectively. In the independent external validation set, the model maintained strong performance with an AUC of 0.903, sensitivity of 88.9%, and specificity of 87.2%. In the training and internal validation, the model also showed high accuracy (0.956 and 0.867, respectively) and sensitivity (0.957 and 0.900, respectively). The integrated T2WI + DWI + CEI (TDC)-clinical data model demonstrated higher diagnostic accuracy than the TDC-only model. CONCLUSION: The developed multimodal MRI radiomics model effectively differentiated HCC from FNH and offers a non-invasive diagnostic tool that surpasses traditional imaging techniques. Further research is warranted to confirm these findings and explore the model's applications in broader clinical settings.