This systematic review and meta-analysis of 12 MRI studies, conducted to support the revision of the JRS Diagnostic Imaging Guidelines, confirmed pooled sensitivity and specificity of 84.2% and 89.1%, respectively, with an HSROC AUC of 0.916. These findings support the use of MRI as a robust diagnostic tool for evaluating myometrial masses.

MRI serves as a key modality for differential diagnosis of uterine myometrial masses. Nevertheless, its recommendation for the comprehensive diagnosis of sarcoma is based on evidence from both qualitative features and quantitative ADC values, as most studies involved small sample sizes and institution-specific ADC cutoff values that are not readily generalizable, reflecting the challenges inherent in conducting large-scale studies of this rare tumor. Recently, several systematic reviews have evaluated the diagnostic performance of MRI for uterine sarcomas [10,11,12]. Hindman et al. [10] reported an overall diagnostic accuracy of 88%–95%, with sensitivity ranging from 83%–100% and specificity from 88%–100%. Raffone et al. [12] demonstrated a pooled sensitivity of 90% (95% CI, 84%–94%), specificity of 96% (95% CI, 96%–97%), positive likelihood ratio of 13.55 (95% CI, 6.20–29.61), negative likelihood ratio of 0.08 (95% CI, 0.02–0.32), diagnostic odds ratio of 175.13 (95% CI, 46.53–659.09), and an AUC of 0.976. These systematic reviews reported substantial heterogeneity in diagnostic accuracy, indicating that performance varies with differences in study populations, reference standards, and MRI feature assessment, similar to what we observed in our review. However, the consistently high specificity underscores the clinical utility of MRI by enabling confident identification of malignancy, which is essential for guiding en bloc resection and avoiding hazardous morcellation.

The present meta-analysis demonstrates that MRI provides high diagnostic accuracy for differentiating uterine sarcomas from leiomyomas, with performance comparable to previous systematic reviews [10,11,12]. However, a significant challenge in interpreting these results lies in the pathological diversity of the malignant cohort. Most studies primarily targeted LMS, but several also included STUMP and ESS, each differing in biological behavior and MRI appearance. STUMP often shows intermediate features between leiomyoma and LMS, contributing to diagnostic ambiguity [26]. ESS may demonstrate characteristic imaging findings such as low-signal-intensity bands, cystic or necrotic areas, and worm-like nodular extension along the myometrium [27, 28]. Carcinosarcoma and adenosarcoma are no longer classified as true sarcomas but as endometrial epithelial tumors or mixed epithelial and mesenchymal tumors [15, 16]. These lesions typically present as heterogeneous large masses filling the uterine cavity, whereas adenosarcoma more commonly protrudes into the cervical canal with frequent cystic changes [29, 30]. Such pathological diversity likely contributed to variability in diagnostic performance, particularly specificity. Similarly, the control groups often comprised various leiomyoma subtypes, including cellular, bizarre, and degenerative variants. These benign subtypes frequently show heterogeneous T2 signal, irregular margins, hemorrhage, or restricted diffusion, mimicking sarcomatous features and leading to false-positive findings.

Among individual MRI parameters, T2WI provides excellent soft-tissue contrast and plays a key role in the initial assessment of myometrial lesions. The studies included in our meta-analysis evaluated T2WI using various criteria, including margin irregularity or infiltrative patterns [14,15,16,17,18, 20,21,22,23,24,25], signal intensity grading relative to reference tissues [14, 16,17,18,19,20, 22,23,24,25], and specific findings such as internal heterogeneity [17, 18, 20, 23, 24], T2-hypointense bands [20], or endometrial cavity interruption [15]. Generally, homogeneous low signal intensity on T2WI, comparable to that of skeletal muscle, is strongly associated with leiomyomas. In contrast, intermediate to high signal intensity on T2WI relative to the background myometrium has been correlated with malignant lesions such as LMS. However, this finding also overlaps with certain benign variants, including cellular and atypical leiomyomas, thereby limiting its specificity [10]. While an irregular margin was frequently identified as a malignancy indicator in the reviewed studies [14,15,16,17,18, 20,21,22,23,24,25], its diagnostic weight is often regarded as weak to moderate, partly because the definition of “irregular” has varied among reports. To address this, Hindman et al. [10] adopted the terminology from the American College of Radiology Thyroid Imaging Reporting and Data System, which defines “irregular,” “lobulated,” and “smooth” margins.

T1WI is useful for detecting intratumoral hemorrhage, and the studies evaluated this meta-analysis using various criteria, primarily focusing on high signal intensity relative to reference tissues such as skeletal muscle, myometrium, or bone marrow [14,15,16,17,18,19,20,21,22,23,24]. Specific attention was given to signal distribution, such as distinguishing central patchy foci from a peripheral hemorrhagic rim [23, 24], and this feature was adopted as a classification algorithm in some studies [18, 23]. Despite being traditionally regarded as a potential indicator of malignancy [20, 31, 32], the evidence linking hemorrhage to sarcoma remains limited and inconsistent because T1 high signal intensity is also observed in benign degenerative or infarcted leiomyomas [10], which can lead to diagnostic overlap.

The presence of necrosis is a characteristic feature of LMS [5, 26]. Historically, early and heterogeneous enhancement reflecting tumor vascularity has also been reported to contribute to the diagnosis [33]. Among the studies included in this meta-analysis that evaluated contrast enhancement, diagnostic criteria were inconsistent, ranging from the assessment of specific enhancement patterns [23, 24] to the simple detection of non-enhancing necrotic areas without pattern analysis [15, 16]. Consequently, its association with malignancy has not been firmly established by either qualitative or quantitative approaches [10]. Therefore, contrast administration is currently regarded most useful for distinguishing viable solid components from necrosis rather than for detailed perfusion analysis.

DWI provides objective quantitative data, where increased cellularity corresponds to restricted diffusion and low ADC values. The studies included in this meta-analysis evaluated these metrics [14, 16, 17, 19,20,21,22,23,24,25], generally confirming that sarcomas demonstrate marked diffusion restriction. Consistent with these findings, a large meta-analysis, by Woo et al. [11] reported that incorporating ADC values markedly improved diagnostic performance (sensitivity 94%, specificity 95%, AUC 0.94), and they identified an ADC threshold of approximately 0.904 × 10⁻³ mm²/s. This aligns with the consensus statement by Hindman et al. [10], which emphasizes that tumors with restricted diffusion (ADC value ≤ 0.9 × 10⁻³ mm²/s) are suggestive of malignancy, a threshold consistent with the cutoff proposed by Abdel Wahab et al. [16] in our reviewed studies. Recent external validations, such as by Horowitz et al. [34], have further confirmed the reproducibility of this consensus threshold. In a study comparing a radiomics model with human readers, Xie et al. [15] highlighted the clinical necessity of quantitative metrics, by reporting that standard qualitative assessment by radiologists based on conventional MRI sequences without DWI (T1WI, T2WI, and DCE), yielded low sensitivity (58.6%), because it relied heavily on typical malignant features such as frank necrosis or hemorrhage, so sarcomas lacking these specific signs were frequently misclassified as benign. In contrast, the ADC-based radiomic model achieved an AUC of 0.83 with sensitivity of 76.0%, specificity of 73.2% demonstrating the value of quantitative ADC analysis.

While individual sequences provide specific clues, the relative contribution of each sequence to the final diagnosis warrants careful consideration. Reflecting this, diagnostic strategies employing these parameters varied widely, encompassing scoring systems [18, 24], algorithm-based models [16, 23, 25], deep learning [21, 35], and radiomics [15]. Notably, given that contrast-enhanced imaging findings beyond necrosis often lack consistency, recent guidelines favor a hierarchical approach. For instance, the consensus statement by Hindman et al. [10] proposes a stepwise evaluation: initially assessing extra-uterine disease and T2 signal intensity, while relying on DWI and specific ADC cutoffs as the critical “branching points” to distinguish sarcomas from benign entities. Similarly, Rosa et al. [23] demonstrated high diagnostic accuracy using a multicenter MRI-based diagnostic algorithm integrating multiple imaging parameters, including T1WI, T2WI, DWI, and CE-T1WI. Furthermore, the integration of clinical and laboratory data is emerging as a vital strategy to enhance individualized malignancy prediction. Yamanishi et al. [25] reported promising results using a diagnostic algorithm that combines MRI findings, specifically T2 signal intensity, T2 margin characteristics, and DWI, with serum lactate dehydrogenase levels. In addition, recent studies published outside our review period have proposed updated algorithms that incorporate menopausal status, T2 signal, DWI signal, tumor margins, and an ADC cutoff value of 1.23 × 10⁻³ mm²/s to achieve high accuracy [36]. Thus, while MRI remains the cornerstone of diagnosis, integrated MRI assessment should prioritize the inclusion of DWI alongside morphological sequences. Current evidence suggests that a comprehensive approach integrating several MRI findings, especially ADC metrics, with clinical and laboratory data offers the most robust strategy for distinguishing sarcomas from benign simulators.

Although histopathology remains the definitive standard for diagnosing sarcoma, distinguishing it from leiomyoma is sometimes difficult even for experienced gynecologic pathologists. Coagulative tumor necrosis, a key criterion for malignancy, may resemble hyaline or ischemic degeneration, and interobserver agreement on its presence is only moderate (κ = 0.44) [37]. Cellular leiomyoma and STUMP further complicate interpretation, as their morphological features overlap extensively with those of sarcoma. Given that even histopathologic distinction is challenging, some degree of diagnostic difficulty on imaging is inevitable. These limitations in the reference standard highlight the need for careful imaging assessment. Clinically, misclassification has important consequences: a false-positive diagnosis may lead to unnecessary hysterectomy, whereas a false-negative result may delay appropriate surgery for a true malignancy. This diagnostic uncertainty is particularly critical in younger patients who aim to preserve fertility. Therefore, MRI evaluation should be integrated with clinical findings, laboratory data, follow-up imaging, and, when indicated, biopsy, ideally within a multidisciplinary framework to ensure individualized and oncologically appropriate management.

This review has several limitations. First, inclusion was restricted to English-language studies published from 2019 onward, which may introduce selection bias. Second, subgroup analyses were not performed, and potential sources of heterogeneity could not be formally evaluated. Third, the meta-analysis relied on study-level 2 × 2 data, preventing adjustment for covariates or harmonization of diagnostic thresholds. Fourth, most included studies were retrospective and used variable inclusion criteria, reference standards, and MRI protocols. Malignant cohorts sometimes included non-sarcomatous tumors, and benign lesions were occasionally verified by follow-up alone. Finally, disease prevalence varied widely across studies, limiting the comparability of predictive values and allowing only qualitative assessment of publication bias.

In conclusion, although uterine sarcomas are ultimately diagnosed histopathologically after surgical resection, preoperative MRI plays a crucial role in treatment planning. MRI provides substantial diagnostic value for distinguishing sarcomas from leiomyomas, despite unavoidable overlap in some cases. Standardized MRI protocols and diagnostic criteria are essential to improve diagnostic consistency and reproducibility. Future research should focus on establishing integrated diagnostic frameworks to support personalized management that balances oncologic safety with fertility preservation.