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

Practitioners practicing fetal ultrasound (US) exams can nowadays be supported by artificial intelligence (AI) software (under development, not approved) evaluating images captured during routine clinical practice. The objective was to assess the performance of an AI software for fetal US exams, by automatic detection of anatomical structures and characteristics. 643 fetal US exams, performed in all trimesters (T1, T2, T3), were collected from 36 North-American and European centres. The sensitivity and specificity of the AI software (see table 1) were assessed by comparing the algorithm's blinded assessment to experts' assessment. The concordances between experts' and the algorithm's assessments for the presence (resp. absence) of an anatomical structure or characteristic was defined as true positive (resp. true negative) on the corresponding structure or characteristic. Reported in table 1. With proven high performance, such an AI software appears relevant in supporting fetal US practitioners by providing live automatic objective evaluation of captured images supporting quality assessment procedures. Specificity - Mean (95% CI) Transthalamic Transventricular Transcerebellar 4 chambers LVOT RVOT 3VV or 3VT Abdominal Circumference Axial view of the kidneys Crown–rump Length Nuchal Translucency Profile Placenta insertion Placenta location