Preoperative differentiation between benign and malignant uterine mesenchymal tumours remains a critical challenge in gynaecological imaging. While benign leiomyomas are common and frequently asymptomatic, malignant tumours such as leiomyosarcomas pose a serious risk due to their aggressive behaviour and poor prognosis. Existing diagnostic algorithms based on MRI have limited sensitivity, raising concerns about misdiagnosis and inappropriate surgical interventions. A recent retrospective cohort study addressed this issue by validating a revised MRI-based algorithm that integrates five key imaging and clinical features to improve the accuracy of tumour classification. This modified approach enables more effective risk stratification and clinical decision-making, reducing the risks associated with both underdiagnosis and overtreatment.
Limitations of Current Diagnostic Approaches
Traditional MRI algorithms for uterine tumour evaluation rely on features such as T2-weighted signal intensity, diffusion-weighted imaging (DWI) and apparent diffusion coefficient (ADC) values. Despite being widely adopted, these criteria often fail to distinguish malignant lesions from benign atypical leiomyomas, leading to false negatives and suboptimal treatment. The original model, which used pelvic lymph node status, T2W signal intensity, DWI compared to the endometrium and an ADC cut-off of 0.9 × 10⁻³ mm²/sec, showed an overall accuracy of 92.5% but demonstrated limited sensitivity at 61.1%. Importantly, it misclassified seven malignant or borderline tumours, including leiomyosarcomas and smooth muscle tumours of uncertain malignant potential (STUMPs).
Must Read: Improving HIFU Prognosis of Uterine Fibroids with Radiomics
The shortcomings of this initial model were particularly evident in premenopausal women and in cases with ambiguous imaging features. Inconsistent signal intensities and overlapping ADC values between benign and malignant lesions contributed to diagnostic uncertainty. These limitations highlighted the need for a more robust, sensitive framework capable of minimising both false positives and false negatives, especially in cases where surgical choices have high-stakes consequences, such as the risk of tumour dissemination from morcellation.
Enhancing Accuracy with a Refined Algorithm
The modified algorithm introduced in this study builds upon existing criteria by incorporating two additional predictors—irregular tumour margins and menopausal status—and refining the interpretation of diffusion signals relative to the bladder rather than the endometrium. The ADC threshold was also raised to 1.23 × 10⁻³ mm²/sec based on multivariate analysis, which improved discriminatory capacity. These adjustments significantly enhanced the model’s performance: sensitivity increased to 83.3%, specificity rose to 98.6%, and overall accuracy reached 98%.
Among the 455 patients included in the cohort, the revised model correctly classified 446 cases. Only three malignant or borderline tumours were misclassified, a substantial improvement compared to the original algorithm. Irregular tumour margins, seen in 83.3% of malignant or borderline cases, emerged as a highly predictive feature with an odds ratio of over 100. Menopausal status was also a powerful discriminator, with postmenopausal women having a notably higher prevalence of malignancy.
These refinements led to the development of a five-tier scoring system ranging from clearly benign to high-risk malignant tumours. Tumours assigned to score categories 1 or 2 were uniformly benign, while those in score category 5 had a malignancy prevalence exceeding 90%. The intermediate categories facilitated nuanced decision-making, such as recommending biopsy or short-term follow-up for uncertain lesions. This structured stratification tool provides a practical framework for radiologists and surgeons to tailor patient management while reducing unnecessary radical surgeries.
Clinical Impact and Future Directions
The implications of this improved scoring model are substantial. With more reliable identification of malignant and borderline lesions, clinicians can make better-informed decisions regarding surgical planning, avoiding the use of morcellation in high-risk cases and opting for conservative or minimally invasive procedures where appropriate. This is particularly important in premenopausal women and those with fertility concerns, where overtreatment may lead to significant quality-of-life impacts.
Furthermore, the model’s strong negative predictive value—bolstered by the absence of malignancy in cases with low T2W and DW signal or high ADC values—offers reassurance in excluding cancer in many patients, allowing for de-escalated surveillance strategies. The ability to confidently identify benign disease can also support broader adoption of interventional radiology approaches or medical management.
Nonetheless, the study’s retrospective nature and the low prevalence of malignant tumours within the cohort underscore the need for external validation in larger, prospective settings. Differentiating between STUMPs and frankly invasive malignancies also remains a challenge, which may benefit from the integration of molecular diagnostics or biopsy-based techniques. As imaging technologies and radiomic tools continue to evolve, future iterations of diagnostic models may incorporate advanced quantitative features, further refining their utility in clinical practice.
The refinement of MRI-based diagnostic criteria for uterine mesenchymal tumours offers a meaningful advance in gynaecologic oncology. By integrating five key features—T2W signal, DW signal, ADC values, tumour margins and menopausal status—the new model achieves greater accuracy, sensitivity and specificity than existing approaches. The proposed scoring system enhances clinical decision-making, minimises unnecessary surgical risk and supports personalised care pathways. Although further validation is needed, this novel stratification tool represents a promising step forward in the safe and effective management of uterine tumours.
Source: European Journal of Radiology
Image Credit: iStock
