EP01.41: Performance of an AI software in automatic detection of quality criteria and characteristics on fetal ultrasound images

Le résumé fourni par la source

Objectives: This study aimed to evaluate interoperator reproducibility for contrast-to-noise ratio (CNR) assessment at the 18-week scan.Methods: This retrospective cross-sectional study was conducted at a single-centre.The study sample comprised 187 consecutive 18-week scans, including 205 fetuses and 565 ultrasound images.Ultrasound examinations were performed by a single operator using a General Electric Medical Systems (Zipf, Austria) machine and an RM6C 2-7-MHz curved array transducer with consistent settings.Two independent and blinded sonographers scored the contrast and noise levels on a scale of 1 to 3 for two regions of interest, each corresponding to a small and a large anatomical structures, in the four-chamber, abdominal perimeter, and longitudinal spine views.The CNR was classified as poor (CNR < .75),fair (CNR: .75÷1.5), good (CNR: 1.5÷2.25)and excellent (CNR >2.25).Results: Interobserver agreements of CNR for the two operators were fair to good.Conclusions: We demonstrated moderate interobserver reproducibility of the CNR.Sonographer CNR assessment of the 4-chamber view may be the best approach for evaluating the quality of ultrasound images in daily routine practice.EP01.39:Table 1.Small structure Large structure ICC (95% CI) P ICC (95% CI) P Abdominal perimeter view .50(.43 -.62) <.001 .58(.45 -.68) <.001 4-chamber view .64(.49 -.74) <.001 .54(.42 -.63) <.001 Longitudinal spine view .52 (.40 -.63) <.001 .52 (.39 -.64) <.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.