Sonio Suspect

Performance Metrics

Clinical Evidence

The evidence base is two retrospective studies, both run by the manufacturer for the FDA submission. The standalone validation in the 510(k) summary used 8,745 fetal ultrasound images from 1,115 examinations across 75 sites, 64 of them in the United States, held out from model development. Across the eight findings the device reached a mean sensitivity of 93.2 percent (95 percent bootstrap CI 91.6 to 94.6) and a mean specificity of 90.8 percent (89.5 to 92.0). Performance varied by finding. Abdominal situs inversus scored 99.3 percent on both measures. Malposition of the great vessels had the lowest sensitivity, 87.7 percent, and absence or unusual size of one of the three vessels the lowest specificity, 81.5 percent. Disequilibrium or absence of a ventricle sat at 90.0 percent sensitivity and 81.7 percent specificity. The summary says subgroup results by indication, race, ethnicity, BMI, gestational age, maternal age, scanner manufacturer, geography and finding status are in the user manual, not in the public document. Validation covered GE, Philips, Samsung and Canon scanners only. The reader study, also retrospective, had 13 readers (five maternal-fetal medicine specialists, six obstetrician-gynaecologists, two radiologists) each read 750 images from 287 examinations and 47 sites, with and without the software, separated by a washout of at least 28 days. The primary endpoint was reader accuracy. Mean reader AUC rose from 0.689 (95 percent CI 0.65 to 0.73) unassisted to 0.909 (0.886 to 0.931) assisted, a difference of 0.219 (0.185 to 0.253). The gain was largest for absence of the cavum septi pellucidi, where unassisted readers scored 0.567, and smallest for the stomach finding, where they already scored 0.84. The summary's prose gives the assisted AUC as 90.0 percent while its table gives 0.909; the table value is used here. The study is image-level, so it measures recognition of a finding on a given frame, not detection across a whole examination. No peer-reviewed publication of the reader study was found on PubMed. The standalone dataset is published as Morisset et al., Obstetrics and Gynecology 2026 (PMID 41505754). The abstract reports 6,452 images from 1,115 examinations in 942 patients at 75 sites in five countries, with 6,094 images contributing to the estimates, and the same 93.2 percent sensitivity and 90.8 percent specificity. The image count differs from the 8,745 in the FDA summary; the full text, which should explain the exclusions, could not be read for this entry. Ground truth came from a multidisciplinary panel using a three-step majority consensus. The authors report performance generally consistent across manufacturers, regions, BMI categories and trimesters. All nine authors disclose a Sonio relationship: four are employees (Morisset, Logé-Munerel, Debavelaere, Besson), three are co-founders (Besson, Ville, Stirnemann) and three are consultants (Fries, Oyelese, Turan). No comment, erratum or retraction is linked to the record. Other Sonio-linked papers in PubMed concern different products: the 2023 Ultrasound in Obstetrics and Gynecology paper by Stirnemann et al. validates a phenotype-based diagnostic support tool, and the Combs et al. quality-review series and the 2025 visual-checklist study concern Sonio Pro and exam completeness. A 2026 letter in the same journal by Besson et al. (PMID 42175650) tests two research models of fetal heart structures, stated as not approved for clinical use. A model trained mostly on normal hearts fell to 14 percent sensitivity on transposition of the great arteries; adding abnormal cases fixed most of this. It bears on the great-vessel finding but is not a test of the product. None of these papers evaluates Sonio Suspect. AurisMark: one mark. The device holds an FDA clearance and an EU MDR register entry, and has one peer-reviewed study. Two marks need prospective-grade evidence of this device, and both studies are retrospective, manufacturer-run, and silent on what happens to referral rates or anomaly detection in live screening.

Inside the Auris+ Listing

Regulatory Approvals

Safety Record

No adverse-event report names Sonio Suspect in the FDA MAUDE database, and no recall or enforcement action names Sonio, checked via openFDA on 1 October 2026, nineteen months after clearance. The one report under the Sonio name is a June 2025 malfunction for the sibling product Sonio Detect, with no injury. The harm surface is indirect. A missed flag can let a finding pass unremarked, and a false flag can prompt referral; the labelling keeps the interpreting physician responsible for the read and bars management decisions on the software alone. The 510(k) summary states that the predicate had no design-related recall; neither FDA document has a clinical safety section, so no statement about adverse events in the studies can be made here.

Intended Use & Indications

From the FDA indications for use (K243614, 21 February 2025): Sonio Suspect is intended to assist interpreting physicians, during or after fetal ultrasound examinations, by automatically identifying and characterizing abnormal fetal ultrasound findings on detected views, using machine learning techniques. The device is intended for use as a concurrent reading aid on acquired images, during and/or after fetal ultrasound examinations. The device provides information on abnormal findings that may be useful in rendering potential diagnosis. Patient management decisions should not be made solely on the results of the Sonio Suspect analysis. Prescription use. The 510(k) summary limits the scope to eight findings: malposition of the great vessels; absence or unusual size of one of the three vessels; disequilibrium or absence of one of the two ventricles; thoracic situs inversus; abdominal situs inversus; non-visibility of a single stomach bubble or an abnormally big stomach; absence of the cavum septi pellucidi; and absence of the corpus callosum. The corpus callosum view is not detected automatically and must be assigned by the user.