EP01.35: A national audit on the quality of ultrasound practices relating to the fetal heart assessment in a screening perspective in France

To perform a quantitative and qualitative audit of the images present in the reports of obstetrical ultrasounds in France using purpose-built Deep Learning tools. We asked a major cloud-based pregnancy ultrasound report editor (Monecho) to send us 5000 anonymised T2 files and 5000 anonymised T3 files. We instructed our trained algorithms to retain only those files containing at least 10 images (≥ 10). Then we applied a quantitative and qualitative approach in order to assess the completeness and the quality of these reports. We have categorised the results on a national scale, by regions (13) and by departments (96), and by operators. We received 9626 files, comprising 4910 / 5166 at T2 and 4228 / 4460 at T3, resulting in a total of 228,036 images analysed at T2 and 161,535 at T3, and involving 912 operators across France. 37.5% of the files are complete with significative difference between regions (p<.0001). When considering the average number of images per scan by region at T2 and T3, there is a significative difference between regions as well (p < 0.0001). When considering the qualitative assessment, the mean exam quality was 52,1% (24,48 / 47) with significative difference between regions (p < 0.0001). Our study demonstrate for the first time the quality of practice related to fetal heart assessment at a national level. The global quality related to the reports is medium and need to be improved.