FDA clears Sonio Suspect AI for fetal anomaly detection

FDA clears Sonio Suspect AI for fetal anomaly detection | Image Credit: © Tada Images - © Tada Images - stock.adobe.com.

The Sonio Suspect artificial intelligence (AI) model has received FDA clearance for the detection of fetal anomalies, according to Sonio.¹

The technology provides a 22-point improvement in reader performance. Additionally, it allows for earlier and more accurate detection and characterization of abnormalities. This addresses a critical gap in fetal anomaly detection, with data indicating up to 51% of anomalies are not detected during standard prenatal ultrasound screenings.

“By combining real-time AI quality control with AI-driven anomaly detection, Sonio supports ultrasound providers at every step of the patient pathway, from exhaustive documentation to accurate diagnosis,” said Cécile Brosset, CEO of Sonio. “Our technology is designed to help health care providers detect issues early and streamline processes, ultimately improving the care every patient receives.”

Using the Sonio Suspect AI model, health care providers may detect fetal anomalies in their patients as early as 11 weeks’ gestation. This increases the time families and providers have to act, plan interventions, and improve maternal and fetal health outcomes.

The 22-point improvement in anomaly detection was highlighted in a multicenter reader performance study. There were 47 sites included in the analysis, 37 of which were in the United States. In the analysis, the area under the curve rose from 69% to 91% when using the Sonio Suspect AI model.

Multiple patient demographics were included in the analysis, highlighting the model’s efficacy across a varied population. Additionally, consistency was reported regardless of clinician background or experiencing, including maternal-fetal medicine specialists, obstetricians and gynecologists, and radiologists.

Poor image quality has been linked to 49% of missed cases, and misinterpretation to 31%, highlighting the need for early and accurate detection of congenital malformations to reduce fetal and maternal mortality. Sonio Suspect provides cutting-edge technology that addresses quality control and diagnostic accuracy to mitigate these issues.

Sonio’s model bridges the gap between technology and clinical application through the use of both real-time quality control and AI-driven anomaly detection. This allows for health care providers to employ consistent, efficient, and accurate prenatal screening across diverse health care settings.

The efficacy of an AI algorithm for fetal ultrasound was proven in a Sonio study posted in Ultrasound in Obstetrics & Gynecology.²The analysis included 17,169 screening fetal images performed in all 3 trimesters, and the algorithm’s blinded assessments were compared to expert assessments to evaluate the software’s performance.

A mean sensitivity of 0.928 was reported when labelling 4 standard planes from the first trimester, alongside a mean sensitivity of 0.911 for labelling of 12 standard planes from the second and third trimester.When identifying 6 brain structures, the mean sensitivity was 0.852 and the mean specificity was 0.942.

When identifying 12 heart structures, these values were 0.900 and 0.982, respectively. This data indicated effective management of quality control during fetal ultrasound using an AI model.

“Such a software could improve the workflow of practitioners and imaging quality by ensuring completeness of examinations and that expected anatomical structures are visible on standard planes,” concluded investigators.

References

1. Sonio announces FDA clearance of Sonio Suspect breakthrough AI algorithm delivers 22-point improvement in anomaly detection and enables early prenatal diagnosis. Sonio. February 24, 2025. Accessed February 24, 2024.
2. Stirnemann J, Besson R, Debavelaere V, et al. OC01.01: Performance of an AI algorithm for quality control of routine fetal ultrasound. Ultrasound in Obstetrics & Gynecology. 2023. doi:10.1002/uog.26321