Fetal EchoScan (v1.2)

Fetal EchoScan (v1.2) is a cloud-based machine learning software device that analyzes fetal heart ultrasound video clips to detect suspicious cardiac findings as an adjunct to physician interpretation during second-trimester anatomic ultrasound exams in pregnant women aged 18 or older. It identifies eight types of morphological abnormalities suggestive of congenital heart defects and is intended to assist—not replace—physician diagnosis.

Both the subject device (v1.2) and predicate device (v1.1) use identical machine learning algorithms trained on the same eight suspicious radiographic findings, accept identical ultrasound video inputs (4-chamber, left ventricular outflow tract, right ventricular outflow tract views), and output present/absent/inconclusive assessments per frame with exam summary tables. The v1.2 differs only in output display: it uses annotated DICOMs in a PACS viewer plus an optional third-party user interface, whereas v1.1 used a device web interface.

IEC 62304:2016 (Medical device software—Software life cycle processes); FDA Guidance for Content of Premarket Submissions for Software Contained in Medical Devices; FDA Guidance for Management of Cybersecurity in Medical Devices.

The subject and predicate devices share identical fundamental technology (same machine learning algorithm, same eight detectable findings, same input/output logic) and differ only in the user interface implementation. Bench testing demonstrated the v1.2 achieves high sensitivity (0.984–0.990) and specificity (0.958–0.970) for detecting any suspicious finding across 877 exams, with consistent performance across demographic and equipment subgroups. Since the algorithm and clinical performance are equivalent, the minor UI differences do not raise different safety or effectiveness questions, supporting substantial equivalence.