An ultrasound algorithm for identification of endometrial cancer.

摘要: Objective To propose a scoring system to predict endometrial cancer using different ultrasound image characteristics at gray-scale, with and without enhancement by gel infusion, Doppler transvaginal sonography (TVS) evaluate intra- interobserver variability in assessment of these characteristics. Method Unenhanced TVS, examinations infusion (GIS) were performed prospectively 174 consecutive postmenopausal women thickness ≥ 5 mm. The reference standard all was hysteroscopy or hysterectomy pathological evaluation the malignancy. presence various pattern indicative malignancy their evaluated. Multivariate logistic regression used correlate clinical parameters cancer. Results A simple flow score (which considered only vascularity not single/double dominant vessel, multiple vessels, large color splash densely packed vessels) had an area under receiver–operating curve (AUC) 0.83 prediction cancer. Models including thickness, interrupted endomyometrial junction on unenhanced TVS predicted AUC 0.95 (95% CI, 0.92–0.99) and, addition irregular surface GIS, 0.97 0.94–0.99). A risk (REC) based body mass index, score, GIS very well identifying cancer; REC-score ≥ 4 sensitivity for detection 91% specificity 94%. Observers agreed 82.3% cases (kappa, 0.63 (0.48–0.78)) when subjective analyzed stored videoclips. Conclusion Our observer-dependent proposed seems perform should be tested future studies. Copyright © 2013 ISUOG. Published John Wiley & Sons Ltd.