T2 hyperintense myometrial tumors: can MRI features differentiate leiomyomas from leiomyosarcomas?

PURPOSE

To establish MRI features that help differentiate atypical leiomyomas and leiomyomas with degeneration that show hyperintensity on T2WI from leiomyosarcomas.

METHODS AND MATERIALS

This retrospective study evaluated 41 women who performed MRI before undergoing hysterectomy and had histologically proven atypical leiomyomas, leiomyomas with degeneration or leiomyosarcomas (21 leiomyomas; 20 leiomyosarcomas); only patients with T2 hyperintense myometrial tumors were included. The association between MRI features (contours; free pelvic fluid; intra-tumoral hemorrhagic areas, T2 heterogeneity; T2 dark areas; flow voids; restriction on diffusion-weighted images; signal intensity and heterogeneity after contrast administration; unenhanced areas, localization of unenhanced areas; necrosis; cystic areas) and the histology (leiomyoma vs. leiomyosarcoma) were calculated using Fisher's exact test. For those features that showed a significant association, a univariate linear regression was performed.

RESULTS

Five MRI features demonstrated a significant correlation with malignant histology: irregular borders (p = 0.03); "T2 dark" areas (p = 0.02); presence of central necrosis (p = 0.01); presence of high signal on b1000 DWI (p < 0.001); ADC value lower than 0.82 × 10 mm/s; hyperenhancement of the tumor relative to the myometrium on post-contrast images (p = 0.02); and type 3 enhancing curve on DCE. Two of these features demonstrated a significant result predicting a malignant histology: lobulated contours and central necrosis [F(3;34) = 8,95; p < 0.001; R = 0.506].

CONCLUSION

The presence of lobulated borders, T2 dark areas, necrosis, hyperintensity relative to the myometrium after contrast administration, central necrosis, presence of high signal on b1000 DWI, ADC value lower than 0.82 × 10 mm/s and type 3 enhancing curve on DCE can help distinguish between leiomyoma and leiomyosarcoma. The association of lobulated borders and central necrosis can help predict a malignant histology.