491 ScanNav® audit: an AI-powered screening assistant for fetal anatomical ultrasound

To develop and evaluate a real-time Artificial Intelligence (AI) based system to automatically keep track of acquired images; check that the images conform imaging protocol standards, in essence, replacing human peer reviewer. We developed an AI (ScanNav) which (1) checks completeness record during fetal anomaly screening, ensuring all 19 required views are recorded; (2) assesses quality these (in this case according guidelines ISUOG). First, algorithms were trained on manually evaluated by pool experienced sonologists using state-of-the-art deep learning technology. The resulting algorithm was then assessed separate testing set; it deemed correct if 2 or more (from panel 5) independent agreed with its decision. Due lack expert agreement for “marginal” images, appropriate include as such 2:3 split. 479,322 anonymised from 48,161 routine mid-trimester scans. Agreement between ScanNav sonologist performed set 38,840 (4,284 scans). For scan mean (standard deviation) 93.5% (±5.6%); image 92.2% (±4.7%). processes 11 frames per second PC RTX4000 GPU. A extensively assess quality-check ultrasound against guidelines. Acting peer-reviewer performs well sonologist. Within clinical practice could ensure records scans complete conforming well-defined protocol. By improving confidence trainee less sonologists, has potential help expansion workforce, while can support workflow.