Diagnostic Accuracy of Ultrasound, Contrast-enhanced CT, and Conventional MRI for Differentiating Leiomyoma From Leiomyosarcoma

Rationale and Objectives This study aimed to determine whether uterine leiomyoma can be distinguished from uterine leiomyosarcoma on ultrasound (US), computed tomography (CT), and/or magnetic resonance imaging (MRI) without diffusion-weighted imaging. Materials and Methods Institutional review board approval was obtained and informed consent was waived for this Health Insurance Portability and Accountability Act–compliant retrospective case-control diagnostic accuracy study. All subjects with resected uterine leiomyosarcoma diagnosed over a 17-year period (1998–2014) at a single institution for whom pre-resection US ( n  = 10), CT ( n  = 11), or MRI ( n  = 7) was available were matched by tumor size and imaging modality with 28 subjects with resected uterine leiomyoma. Six blinded radiologists (three attendings, three residents) assigned 5-point Likert scores for the following features: (1) margins, (2) necrosis, (3) hemorrhage, (4) vascularity, (5) calcifications, (6) heterogeneity, and (7) likelihood of malignancy (primary end point). Mean suspicion scores were calculated and receiver operating characteristic curves were generated. The ability of individual morphologic features to predict malignancy was assessed with logistic regression. Results Mean suspicion scores were 2.5 ± 1.2 (attendings) and 2.4 ± 1.3 (residents) for leiomyoma, and 2.7 ± 1.3 (attendings) and 2.7 ± 1.4 (residents) for leiomyosarcoma. The areas under the receiver operating characteristic curves (range: 0.330–0.685) were not significantly different from chance, either overall ( P  = .36–.88) or by any modality ( P  = .28–.96), for any reader. Reader experience had no effect on diagnostic accuracy. No morphologic parameter was significantly predictive of malignancy ( P  = .10–.97). Conclusions Uterine leiomyoma cannot be differentiated accurately from leiomyosarcoma on US, CT, or MRI without diffusion-weighted imaging.