Preoperative Risk Assessment in Endometrial Cancer: MRI-Based Histogram Analysis

A new study suggests that a more advanced form of MRI analysis could significantly improve how clinicians assess risk in patients with endometrial cancer before surgery. Researchers from the University of Gothenburg and Sahlgrenska University Hospital found that histogram analysis of diffusion kurtosis imaging (DKI) may outperform conventional MRI in identifying tumors with aggressive histological features.

Endometrial cancer, the most common gynecologic malignancy, generally has a favorable prognosis. However, for the subset of tumors with high-grade histology or non-endometrioid subtypes, the risk of metastasis—and the need for more extensive surgical intervention—rises sharply. Current preoperative assessments often rely on biopsy and standard imaging to gauge this risk, but discrepancies between pre- and postoperative diagnoses remain common. Roughly 20% of cases are misclassified prior to surgery, potentially leading to overtreatment or undertreatment.

In the prospective study, 94 women with biopsy-confirmed endometrial cancer underwent a comprehensive MRI protocol that extended beyond standard clinical imaging. In addition to conventional T1- and T2-weighted sequences and diffusion-weighted imaging (DWI), the protocol incorporated synthetic MRI for quantitative relaxation mapping, dynamic contrast enhancement (DCE), and DKI—an advanced diffusion technique that captures non-Gaussian water motion in tissues, offering deeper insight into microstructural complexity.

Tumors were segmented across three adjacent MRI slices, and histogram features such as mean, median, skewness, and kurtosis were extracted from various image types. Notably, the DKI-derived diffusion coefficient maps (Dapp) demonstrated statistically significant differences between low- and high-risk tumors across all four histogram parameters. In contrast, no significant differences emerged from synthetic MRI relaxation maps or most DCE perfusion metrics.

The skewness of the Dapp histogram achieved the highest area under the receiver operating characteristic curve (AUC = 0.744), slightly surpassing the visual performance of an experienced radiologist in stratifying tumors by histologic risk (sensitivity = 77.8%, specificity = 58.2%). The findings also indicated that Dapp metrics could better differentiate tumors across the three FIGO grades, offering more granular insights than conventional apparent diffusion coefficient (ADC) values derived from standard DWI.

“Our results suggest that DKI, particularly the histogram features of Dapp maps, could add valuable diagnostic precision in preoperative assessments,” the authors concluded. They noted that the differences likely reflect greater heterogeneity and cellular density in more aggressive tumors, phenomena previously observed in gliomas, breast, and prostate cancers using DKI techniques.

While the study benefits from a prospective design and consistent imaging protocols, the authors acknowledge some limitations. The number of high-risk tumors was relatively small, and tumor segmentations were limited to three slices rather than entire volumes. Differences in image resolution between sequences also meant that histogram data varied in pixel count, potentially influencing robustness. Still, the time-efficient slice-based approach may offer a practical advantage in clinical settings, and future studies could explore automated whole-tumor segmentation to standardize analysis.

Further validation across institutions and scanner types will be critical before DKI can be fully integrated into clinical pathways. But if confirmed, the technique could help reduce surgical overtreatment by offering a more reliable, image-based method of confirming or questioning preoperative biopsy results. It might also spare patients with low-risk tumors from unnecessary lymphadenectomy and its associated complications, such as lymphedema.