Abstract
BACKGROUND: Comprehensive sonographic evaluation of pre- and postmenopausal women with abnormal uterine bleeding is essential for accurate diagnosis and the optimization of curative outcomes for endometrial malignancy. Microvascular flow imaging (MVFI), a state-of-the-art Doppler technique, enables high-resolution, noninvasive mapping of tumor-specific neovascularity that critically drives the initiation, growth, and progression of endometrial malignancy. This study aimed to compare the diagnostic performance of MVFI with that of conventional color Doppler imaging (CDI) for detecting endometrial malignancy using histopathology as the reference standard. METHODS: From June 2023 to October 2024, 283 females with abnormal uterine bleeding over the age of 40 years were enrolled in this prospective single-center study. Transvaginal grayscale ultrasound imaging was performed with a HERA W10 system (Samsung Medison Co., Ltd.), with standardized documentation of endometrial features. Endometrial vascularity was subsequently evaluated through tandem CDI and MVFI assessments. Two senior radiologists independently assessed endometrial vascularity using International Endometrial Tumor Analysis (IETA) consensus criteria with a 4-point scale. Intra- and interobserver agreement were evaluated through sequential and reversed-order interpretations. All imaging results were validated against histopathological outcomes as the reference standard. Diagnostic performance was evaluated for (I) MVFI and CDI individually and for (II) integrated models combining vascular detection techniques with grayscale features for endometrial malignancy detection. RESULTS: With histological outcomes used as the reference standards, there were 32 malignant and 251 nonmalignant endometrial specimens. MVFI yielded significantly higher vascular scores than did CDI (median score: 2 vs. 1; P<0.001). MVFI demonstrated significantly superior diagnostic performance as compared to CDI in terms of area under the curve (AUC) (0.770 vs. 0.681; P=0.011) and sensitivity (71.9% vs. 59.4%; P=0.031) and had comparable specificity (72.9%). Furthermore, the MVFI-grayscale combination model, as compared to the CDI-grayscale combination model, yielded a higher AUC (0.855 vs. 0.809; P=0.046) and sensitivity (87.5% vs. 68.8%; P=0.031) and comparably specificity (72.9%). Both CDI and MVFI showed excellent intraobserver agreements (κ>0.8) and good interobserver agreement (κ>0.7). CONCLUSIONS: MVFI demonstrated significantly higher diagnostic performance than did CDI for the detection of endometrial malignancy; MVFI is thus a promising adjunctive technique for precisely diagnosing endometrial malignancy in females with abnormal uterine bleeding.