FETOLY

FETOLY is software that uses machine learning to analyze fetal ultrasound images in real-time to automatically detect fetal heart and brain views, identify quality criteria within those views, and provide biometric measurements. It is intended as a concurrent reading aid during routine fetal heart and brain examinations in the second and third trimesters (17–40 weeks gestational age) for qualified ultrasound professionals.

FETOLY expands the predicate device (FETOLY-HEART) by adding fetal brain detection and measurement capabilities alongside heart analysis. It detects nine ultrasound views (five cardiac, four brain) versus five cardiac-only in the predicate, identifies 95 quality criteria (52 heart + 43 brain) versus 52 heart-only, and provides eight biometric measurements including cardiac angles and ratios plus brain cephalic index, versus cardiac measurements only.

IEC 62304 (software verification testing); acceptance criteria based on sensitivity ≥85% and specificity ≥85% for view detection, ≥90% sensitivity/specificity for heart quality criteria detection, and limits of agreement thresholds for biometric measurements validated against reference standards established using a 2+1 ground truth annotation procedure.

FETOLY is substantially equivalent because it shares the same intended use, user population, and clinical application (concurrent ultrasound reading aid) as predicate FETOLY-HEART. The expansion to brain views and measurements does not raise new safety or effectiveness questions because performance testing on 750 patient cases (completeness) and 441 patient cases (measurements) across diverse subgroups demonstrates sensitivity and specificity meeting or exceeding acceptance criteria. The brain measurement functionality is addressed by reference device FETAL HS (Voluson Expert), confirming biometric measurement equivalence. The frozen machine learning model with a predetermined change control plan ensures controlled future modifications.