OBJECTIVE. The purpose of this article was to evaluate MRI features of uterine leiomyomas that predict volumetric response after uterine artery embolization (UAE). MATERIALS AND METHODS. This retrospective study included 75 patients with 212 uterine leiomyomas who were successfully treated between August 2013 and December 2018. To predict uterine volumetric response, age, number of lesions, and baseline uterine volume were assessed. To predict leiomyoma volumetric response, a multivariate regression analysis was performed to evaluate six predictive factors: location, baseline leiomyoma volume, signal intensity on T1-weighted and T2-weighted MRI, heterogeneity of signal intensity on T2-weighted MRI, and vascularity on subtraction imaging (SI). A five-variable predictive ROC model was developed to evaluate the diagnostic accuracy of the signal intensity ratio on T2-weighted MRI, enhancement ratio, heterogeneity ratio on T2-weighted MRI, location, and baseline leiomyoma volume in predicting at least 40% leiomyoma volumetric response. RESULTS. Age, number of leiomyomas, and baseline uterine volume were not predictive of uterine volumetric response. A submucosal location was the best predictive factor of leiomyoma volumetric response, and it showed 32.2% more leiomyoma volumetric response compared with a nonsubmucosal location (p < .001). Hyperintensity on T2-weighted MRI was the second best predictive factor of leiomyoma volumetric response, and it showed 16.9% more volumetric response compared with hypointense leiomyomas (p = .013). A small baseline leiomyoma volume (< 58 cm3) was associated with 10.2% more leiomyoma volumetric response compared with larger leiomyomas (p = .01). Leiomyomas that were hyperintense on SI showed 7.9% more leiomyoma volumetric response compared with those that were hypointense (p = .014). The five-variable ROC model showed high diagnostic accuracy with an AUC of 0.85, sensitivity of 82%, and specificity of 71%. CONCLUSION. A submucosal location, hyperintensity on T2-weighted MRI, small baseline leiomyoma volume (< 58 cm3), and hyperintense leiomyoma on subtraction imaging are the main independent favorable predictors of leiomyoma volumetric response after UAE. An accurate predictive ROC model was developed that may help in selecting patients suitable for UAE. Quantitative assessment of heterogeneity on T2-weighted MRI showed promising results as a predictor of volumetric response, and further research in this area using texture analysis and radiomics is suggested.
Keywords: MRI; embolization; predictive modeling; uterine artery embolization; uterine leiomyoma.