Purpose: To retrospectively evaluate the ability of magnetic resonance imaging (MRI) to differentiate malignant from benign myometrial tumours.
Methods: Fifty-one women underwent MRI before surgery for evaluation of a solitary myometrial tumour. At histopathology, there were 25 uncertain or malignant mesenchymal tumours and 26 benign leiomyomas. Conventional morphological MRI criteria were recorded in addition to b 1,000 signal intensity and apparent diffusion coefficient (ADC). Odds ratios (OR) were calculated for each criterion. A multivariate analysis was performed to construct an interpretation model.
Results: The significant criteria for prediction of malignancy were high b 1,000 signal intensity (OR = +∞), intermediate T2-weighted signal intensity (OR = +∞), mean ADC (OR = 25.1), patient age (OR = 20.1), intra-tumoral haemorrhage (OR = 21.35), endometrial thickening (OR = 11), T2-weighted signal heterogeneity (OR = 10.2), menopausal status (OR = 9.7), heterogeneous enhancement (OR = 8) and non-myometrial origin on MRI (OR = 4.9). In the recursive partitioning model, using b 1,000 signal intensity, T2 signal intensity, mean ADC, and patient age, the model correctly classified benign and malignant tumours in 47 of the 51 cases (92.4 %).
Conclusion: We have developed an interpretation model usable in routine practice for myometrial tumours discovered at MRI including T2 signal, b 1,000 signal and ADC measurement.
Key points: • MRI is widely used to differentiate benign from malignant myometrial tumours. • By combining T2-weighted, b 1,000 and ADC features, MRI is 92.4 % accurate. • DWI may limit misdiagnoses of uterine sarcoma as benign leiomyoma. • Patient age is important when considering a solitary myometrial tumour.