LAUSR

Abstract Objective This study aimed to assess the predictive value of conventional magnetic resonance imaging (MRI) combined with radiomics in determining the nonperfused volume ratio (NPVR) following high-intensity focused ultrasound (HIFU) ablation for uterine fibroids. Methods and materials A total of 216 symptomatic uterine fibroids in 216 women were subjected to HIFU ablation from October 2015 to March 2020. Baseline clinical and MR parameters acquired before and after HIFU ablation were analyzed, and the NPVR was calculated accordingly. Radiomics features were extracted using A.K. software on T2-weighted imaging (T2WI). The minimum redundancy and maximum relevancy (mRMR) method were used to refine the selected radiomics features. Then, multiple linear regression models, the Wilcoxon signed-rank test, and Spearman’s rank correlation and Bland–Altman analyses were conducted. Results Conventional MRI combined with radiomics revealed the signal intensity on T2WI (X9), enhancement degree on T1-weighted imaging (T1WI) (X11), uterine fibroid location (X4), wavelet_glszm_SizeZoneNonUniformity first order (X12) and wavelet_HHH_firstorder_Skewness (X13) negatively affected the NPVR. The resulting regression equation was NPVR = 104.030 − 11.886 × X9 − 5.459 × X11 − 2.776 × X4 − 0.20 × X12 − 16.913 × X13. The adjusted R 2 values of the conventional MRI model and combined model were 0.385 and 0.408, respectively, and the two fitted models were statistically significant (p < 0.05). No significant differences were observed between the predicted NPVR value [81 (71, 91) %] of the combined model and the actual NPVR value [89 (77, 97) %] (p > 0.05). In addition, the predicted NPVR was correlated with the actual NPVR (r = 0.655, p < 0.001). Conclusions The efficiency of the combined model was better than that of the conventional MRI model in predicting the NPVR following HIFU ablation for uterine fibroids. Radiomics is an important supplemental modality to conventional MRI.