Butterfly Gestational Age Tool

Performance Metrics

Clinical Evidence

Two peer-reviewed studies cover the model and one manufacturer validation covers the cleared product. The model was developed in the Fetal Age Machine Learning Initiative (FAMLI) at the University of North Carolina. Pokaprakarn and colleagues (NEJM Evidence 2022) enrolled 4,695 pregnant volunteers in North Carolina and Zambia between September 2018 and June 2021 and trained a neural network on 109,806 blind sweeps, or 21.3 million frames, from 3,509 of them. In the main test set of 716 women the model's mean absolute error (MAE) against a previously established gestational age was 3.9 days, better than same-day sonographer biometry at 4.7 days (difference minus 0.8 days, 95 per cent CI minus 1.1 to minus 0.5). In 249 Zambian women whose sweeps were taken by nurse midwives with no ultrasound training on a Butterfly iQ, the model's MAE was 4.9 days against 5.4 for biometry (difference minus 0.6 days, 95 per cent CI minus 1.3 to 0.1, so not shown to be better), and 17.4 days for the last menstrual period. The IVF set of 47 women also had a confidence interval crossing zero. The authors note that the model underestimated gestational age beyond 37 weeks in the novice set and cannot estimate below 9 weeks. Stringer and colleagues (JAMA 2024, NCT05433519) then ran a prospective diagnostic accuracy study of the model built into the software of a Butterfly iQ+. Four hundred women with viable singleton pregnancies were dated by transvaginal crown-rump length in the first trimester in Lusaka and Chapel Hill, then scanned at randomly timed follow-up visits by clinicians with no sonography training after a one-day course. In the primary window of 14 to 27 weeks the tool's MAE was 3.2 days against 3.0 days for biometry on a high-specification machine (difference 0.2 days, 95 per cent CI minus 0.1 to 0.5), inside the prespecified two-day equivalence margin. Assessments within 7 days of the true age were 90.7 per cent for the tool and 92.5 per cent for biometry. Both methods lose precision later. At 28 to 36 weeks the MAE was 6.1 days for the tool and 7.1 for biometry (difference minus 1.1 days, 95 per cent CI minus 1.7 to minus 0.4). At 37 to 40 weeks the tool did worse than biometry: 11.5 days against 9.1 (difference 2.4 days, 95 per cent CI 1.2 to 3.7), with 28 per cent of tool estimates within 7 days against 46 per cent for biometry. Subgroups including the Zambian and North Carolina cohorts and high body-mass index were consistent. The Gates Foundation funded the study, Butterfly donated the probes and integrated an encrypted model into its software, and neither had a role in design or analysis. No conflicts were reported. The 510(k) summary (K252148) cites the 2022 paper as the tool's development study and does not cite the 2024 trial, so whether the cleared build is the same model weights as the one tested in JAMA is not stated. The clearance validation is a separate, manufacturer-run prospective study: 111 pregnant women at 16 to 37 weeks, enrolled March 2025 to February 2026 at four US sites (58 at Butterfly's Burlington office, 36 at its New York office, 14 at Thomas Jefferson University, 3 at a primary-care practice in Flagstaff), scanned by 13 trained clinicians (7 physicians, 6 sonographers) with both an iQ+ and an iQ3, with sonographer biometry at the same visit. The FDA summary reports Bland-Altman bias and limits of agreement, not MAE. With the iQ+, bias against biometry was 0.7 days at 16 to 21 weeks, minus 0.9 days at 22 to 27 weeks and minus 1.3 days at 28 to 37 weeks, each with a confidence interval crossing zero, and limits of agreement of about plus or minus 7 days early and minus 12 to plus 10 days late. With the iQ3 the tool ran consistently low: bias minus 2.5, minus 2.5 and minus 4.9 days across the three windows, each interval excluding zero, with the third-trimester limits from minus 14.5 to plus 4.7 days. The acceptance criterion was agreement with the last-menstrual-period date within 10, 14 or 30 days by trimester window. The summary does not say how the criterion was applied or whether it was met. At 16 to 21 weeks the upper limit of agreement with that date was 12.2 days for the tool on the iQ+ and 10.8 days for biometry, both beyond 10 days. Subgroup results by window, site, BMI and provider type are described as showing no appreciable difference from biometry, without figures. A Google blind-sweep model reported in JAMA Network Open in 2026 is a different system and is not evidence for this device. AurisMark reasoning: one FDA clearance and peer-reviewed prospective evidence on the model earn one mark. A second is withheld because the strongest study tested the research model in an iQ+ with novice users in 2022 to 2023 rather than the cleared build in the cleared user population, the manufacturer validation is 111 patients with a systematic underestimate on the iQ3, and after five months on the market there is no deployment or independent evidence for the cleared tool in any jurisdiction.

Inside the Auris+ Listing

Regulatory Approvals

Safety Record

No recall, enforcement action or MAUDE adverse-event report naming the Gestational Age Tool was found in the openFDA device databases, queried directly on 9 September 2026. The platform's history is relevant: in February 2020 Butterfly recalled two earlier AI features, Auto Ejection Fraction and Auto Bladder Volume, because they had shipped in app updates without a 510(k); both recalls terminated in June 2020. The safety question for this tool is a confident-looking date that is wrong late in pregnancy. Limits of agreement against biometry in the third trimester span about 22 days on the iQ+ and 19 days on the iQ3, the iQ3 reads about five days low there, and the research model underestimates beyond 37 weeks. The indication answers this by making the output adjunctive and excluding prenatal management and delivery planning. Using the estimate to manage a pregnancy or plan a delivery is outside the cleared use.

Intended Use & Indications

Per the FDA indications for use (K252148): The Butterfly Gestational Age Tool is indicated to provide an output of gestational age (GA) of a singleton intrauterine pregnancy presumed to be between 16-37 weeks gestation. It is for use by qualified and trained healthcare professionals in environments where healthcare is provided. This adjunctive information is not intended to be used for prenatal management and/or delivery planning. The Butterfly Gestational Age Tool is to be used with Butterfly's ultrasound systems (iQ+ or iQ3).