Application of a denoising doppler algorithm in adnexal masses ultrasound: impact on color score attribution and perceived image quality.

Abstract

Purpose: Vascularization assessment is crucial for ovarian masses diagnosis; however, its subjectivity often results in suboptimal interobserver agreement. This study evaluated the impact of the OvAi denoising algorithm on IOTA color score assignment and perceived image quality in adnexal mass ultrasound.

Materials and methods: In this multicenter prospective study (May 2018-August 2022), color/power Doppler videos of 100 adnexal masses from 100 women were acquired in two outpatient clinics. Six experts, five mid-experienced, and eight beginner examiners assessed IOTA color score and image quality before and after applying the denoising algorithm. Interobserver agreement on color score assignments was measured using Fleiss' Kappa; perceived image quality changes were tested via Chi-square.

Results: Interobserver agreement for color score assignment was fair across all observer groups (κ: 0.296-0.396) and unchanged post-denoising. In benign lesions, CS1 increased (+14.9%), while CS2-CS4 decreased. In malignant cases, CS1 and CS2 rose (+76.2% and +30.8%), with smaller reduction in CS3 (-3.7%) and CS4 (-21.4%). The CS4 decrease was less marked in malignant than in benign lesions (-21.4% vs -38.3%). A significant improvement in perceived image quality was observed in all observer groups (p < 0.05).

Conclusion: The OvAi algorithm improved perceived image quality without significantly impacting interobserver agreement. Continued development may enable more objective and automated color score assignments, potentially reducing interobserver variability and supporting diagnostic reliability.