Application of different MR sequence combinations in radiomics assessment of liver localized lesions

Abstract Objective To explore the possibility of using radiomics to differentiate hepatocellular carcinoma (HCC) from focal nodular hyperplasia(FNH). Materials and methods The MRI images of 196 patients diagnosed in Zhangzhou Municipal Hospital Affiliated to Fujian Medical University from 2011 to 2021 were retrospectively analyzed. A variety of imaging sequences were used in the study, including T2WI, DWI and C sequence. Using these images, a variety of radiomics features were extracted and analyzed. Using these images, a variety of radiomics features were extracted and combined into T2WI/DWI combination, T2WI/C combination and DWI/C combination. The support vector machine (SVM) classifier was used to construct a radiomics model for differentiating HCC from FNH. Results Radiomics models based on different MRI sequence combinations showed high accuracy in the diagnosis of HCC. Radiomics models based on different MRI sequence combinations showed high accuracy in the diagnosis of HCC and FNH. The model based on the combination of T2WI and magnetic resonance enhanced sequence (TC) had the best AUC value and the highest AUC value. The model based on the combination of T2WI and magnetic resonance enhanced sequence (TC) had the best AUC value and sensitivity. The model constructed by the combination of DWI and magnetic resonance enhanced sequence (DC) performed the best in terms of specificity and accuracy. Conclusion The MRI sequence-based radiomics model provides a new auxiliary tool for the differential diagnosis of HCC and FNH, which is helpful to improve the accuracy of clinical imaging diagnosis and promote the individualized precise treatment of the disease.