# AI reading of blind-sweep ultrasound to date pregnancies where sonographers are scarce Idea · id `maternal:idea-ai-blind-sweep-ultrasound` · Maternal & Newborn Atlas · release `fc5a600ec7387e71` (public) > General information from published sources, not medical advice, and not reviewed by clinicians. Check the cited source before relying on anything here. The atlas's own text and curation may be reused under Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0, https://creativecommons.org/licenses/by-nc/4.0/): credit the atlas, and do not use it commercially. See the atlas About page. Text quoted from a source stays under that source's own terms, given as the source's reuse\_basis. The atlas's reuse terms do not cover it. The downloads list each quotation with its source in third\_party\_quotations. Browse, life stage: Pregnancy Browse, care area: The baby before birth ## In plain language Ultrasound is the best way to date a pregnancy, but it needs trained sonographers and costly machines. Researchers trained a computer program to estimate how many weeks pregnant a woman is from simple sweeps of a low-cost probe over the abdomen. In a study in Zambia and North Carolina, untrained users got dating about as accurate as trained sonographers. It is still a research tool and not routine care. ## Technical summary Hypothesis with a prospective test: AI estimation of gestational age from blind sweeps by novice users on a low-cost device is as accurate as standard biometry by credentialed sonographers. Results: development study (NEJM Evidence 2022; 4695 volunteers; mean absolute error 3.9 against 4.7 days) and prospective diagnostic accuracy study (JAMA 2024; 400 participants; 3.2 against 3.0 days, difference 0.2 days, 95% CI -0.1 to 0.5, within a prespecified 2-day margin). The first paper reports funding from the Bill and Melinda Gates Foundation. Dating accuracy only; effects on care and outcomes are not tested. ## Claims, with the population each applies to ## Context: Pregnant volunteers in North Carolina and Zambia, development and test sets of an AI gestational age model Recipient: Fetal; Population: 4695 pregnant volunteers recruited from September 2018 through June 2021 who had blind ultrasound sweeps of the abdomen alongside standard fetal biometry. The model was tested in three test data sets, including women who conceived by in vitro fertilisation and sweeps collected by untrained users with low-cost devices in Zambia.; Care setting: Research ultrasound; some sweeps collected by untrained providers with low-cost battery-powered devices; Geography: North Carolina (United States) and Zambia; Life stage: Pregnancy; Gestational age: Not reported; Postnatal age: Not reported; Follow-up: Not stated in the parts of the source that were read. ### In a model-development study, 4695 pregnant volunteers in North Carolina and Zambia were recruited from September 2018 through June 2021 and had blind ultrasound sweeps (cineloop videos) of the abdomen alongside standard fetal biometry. A neural network trained to estimate gestational age from the sweeps had a mean absolute error of 3.9 days against 4.7 days for biometry in the main test set (difference -0.8 days, 95% CI -1.1 to -0.5; P<0.001). In the set of sweeps collected by untrained users with low-cost, battery-powered devices, the error was 4.9 days for the model against 5.4 days for biometry (difference -0.6, 95% CI -1.3 to 0.1). - Category: Description - Evidence stage: Observational - Certainty: Certainty not assessed - Applies to: Recipient: Fetal; Population: 4695 pregnant volunteers recruited from September 2018 through June 2021 who had blind ultrasound sweeps of the abdomen alongside standard fetal biometry. The model was tested in three test data sets, including women who conceived by in vitro fertilisation and sweeps collected by untrained users with low-cost devices in Zambia.; Care setting: Research ultrasound; some sweeps collected by untrained providers with low-cost battery-powered devices; Geography: North Carolina (United States) and Zambia; Life stage: Pregnancy; Gestational age: Not reported; Postnatal age: Not reported; Follow-up: Not stated in the parts of the source that were read. - Limitations: A development study with test sets drawn from the same project, before the prospective study above. The in vitro fertilisation and untrained-user comparisons have intervals that include no difference. Dating accuracy only. Only the abstract was read. The abstract states that the Bill and Melinda Gates Foundation funded the work. - Last checked: 2026-10-14 - Sources: 1. Supports: AI Estimation of Gestational Age from Blind Ultrasound Sweeps in Low-Resource Settings (Publication, edition PubMed record (abstract and bibliographic data) from NCBI E-utilities, retrieved 2026-10-14. The full text and the version of record were not read.) Locator: Abstract, METHODS, sentence 1, fragment 1 of 2. Retrieved 2026-10-14T18:18:40Z. Source state: Current. Snapshot hash: d1398ce4cfb7ce5ac00348751fd01da51e0ef49c72f4df2c8b63ee2c778f24a6. How it applies: PubMed abstract. 2. Supports: AI Estimation of Gestational Age from Blind Ultrasound Sweeps in Low-Resource Settings (Publication, edition PubMed record (abstract and bibliographic data) from NCBI E-utilities, retrieved 2026-10-14. The full text and the version of record were not read.) Locator: Abstract, METHODS, sentence 1, fragment 2 of 2. Retrieved 2026-10-14T18:18:40Z. Source state: Current. Snapshot hash: d1398ce4cfb7ce5ac00348751fd01da51e0ef49c72f4df2c8b63ee2c778f24a6. How it applies: PubMed abstract. 3. Supports: AI Estimation of Gestational Age from Blind Ultrasound Sweeps in Low-Resource Settings (Publication, edition PubMed record (abstract and bibliographic data) from NCBI E-utilities, retrieved 2026-10-14. The full text and the version of record were not read.) Locator: Abstract, METHODS, sentence 2, fragment 1 of 2. Retrieved 2026-10-14T18:18:40Z. Source state: Current. Snapshot hash: d1398ce4cfb7ce5ac00348751fd01da51e0ef49c72f4df2c8b63ee2c778f24a6. How it applies: PubMed abstract. 4. Supports: AI Estimation of Gestational Age from Blind Ultrasound Sweeps in Low-Resource Settings (Publication, edition PubMed record (abstract and bibliographic data) from NCBI E-utilities, retrieved 2026-10-14. The full text and the version of record were not read.) Locator: Abstract, METHODS, sentence 2, fragment 2 of 2. Retrieved 2026-10-14T18:18:40Z. Source state: Current. Snapshot hash: d1398ce4cfb7ce5ac00348751fd01da51e0ef49c72f4df2c8b63ee2c778f24a6. How it applies: PubMed abstract. 5. Supports: AI Estimation of Gestational Age from Blind Ultrasound Sweeps in Low-Resource Settings (Publication, edition PubMed record (abstract and bibliographic data) from NCBI E-utilities, retrieved 2026-10-14. The full text and the version of record were not read.) Locator: Abstract, RESULTS, sentence 1, fragment 1 of 2. Retrieved 2026-10-14T18:18:40Z. Source state: Current. Snapshot hash: d1398ce4cfb7ce5ac00348751fd01da51e0ef49c72f4df2c8b63ee2c778f24a6. How it applies: PubMed abstract. 6. Supports: AI Estimation of Gestational Age from Blind Ultrasound Sweeps in Low-Resource Settings (Publication, edition PubMed record (abstract and bibliographic data) from NCBI E-utilities, retrieved 2026-10-14. The full text and the version of record were not read.) Locator: Abstract, RESULTS, sentence 1, fragment 2 of 2. Retrieved 2026-10-14T18:18:40Z. Source state: Current. Snapshot hash: d1398ce4cfb7ce5ac00348751fd01da51e0ef49c72f4df2c8b63ee2c778f24a6. How it applies: PubMed abstract. 7. Supports: AI Estimation of Gestational Age from Blind Ultrasound Sweeps in Low-Resource Settings (Publication, edition PubMed record (abstract and bibliographic data) from NCBI E-utilities, retrieved 2026-10-14. The full text and the version of record were not read.) Locator: Abstract, RESULTS, sentence 3, fragment 3 of 3. Retrieved 2026-10-14T18:18:40Z. Source state: Current. Snapshot hash: d1398ce4cfb7ce5ac00348751fd01da51e0ef49c72f4df2c8b63ee2c778f24a6. How it applies: PubMed abstract. 8. Supports: AI Estimation of Gestational Age from Blind Ultrasound Sweeps in Low-Resource Settings (Publication, edition PubMed record (abstract and bibliographic data) from NCBI E-utilities, retrieved 2026-10-14. The full text and the version of record were not read.) Locator: Abstract, CONCLUSIONS, sentence 2. Retrieved 2026-10-14T18:18:40Z. Source state: Current. Snapshot hash: d1398ce4cfb7ce5ac00348751fd01da51e0ef49c72f4df2c8b63ee2c778f24a6. How it applies: PubMed abstract. 9. Background: AI Estimation of Gestational Age from Blind Ultrasound Sweeps in Low-Resource Settings (Publication, edition PubMed record (abstract and bibliographic data) from NCBI E-utilities, retrieved 2026-10-14. The full text and the version of record were not read.) Locator: Abstract, CONCLUSIONS, sentence 3. Retrieved 2026-10-14T18:18:40Z. Source state: Current. Snapshot hash: d1398ce4cfb7ce5ac00348751fd01da51e0ef49c72f4df2c8b63ee2c778f24a6. How it applies: Funding statement in the abstract. ## Context: Pregnant individuals scanned by novice users with an AI-enabled ultrasound device, 14 0/7 to 27 6/7 weeks (Zambia and North Carolina) Recipient: Fetal; Population: Individuals with viable, single, nonanomalous, first-trimester pregnancies (400 enrolled), whose gestational age was set by transvaginal crown-rump length. The person scanned is the pregnant woman; the quantity estimated is the gestational age of the pregnancy.; Care setting: Low-cost battery-powered device used by novice users with no prior training in sonography, at follow-up visits; Geography: Lusaka, Zambia, and Chapel Hill, North Carolina (United States); Life stage: Pregnancy; primary evaluation window 14 0/7 to 27 6/7 weeks of gestation; Gestational age: 14 weeks 0 days (98 days) to 27 weeks 6 days (195 days), measured at the time of measurement; Postnatal age: Not reported; Follow-up: Random follow-up visits throughout gestation ### A prospective diagnostic accuracy study enrolled 400 individuals with viable, single, nonanomalous, first-trimester pregnancies in Lusaka, Zambia, and Chapel Hill, North Carolina. At random follow-up visits, novice users with no prior training in sonography collected blind sweeps of the maternal abdomen with an AI-enabled low-cost, battery-powered ultrasound device, while credentialed sonographers did fetal biometry with a high-specification machine. In the primary window (14 0/7 to 27 6/7 weeks) the device met the prespecified equivalence margin of 2 days: mean absolute error 3.2 days against 3.0 days for biometry (difference 0.2 days, 95% CI -0.1 to 0.5). The share of assessments within 7 days of the reference gestational age was 90.7% for the device and 92.5% for biometry. - Category: Description - Evidence stage: Observational - Certainty: Certainty not assessed - Applies to: Recipient: Fetal; Population: Individuals with viable, single, nonanomalous, first-trimester pregnancies (400 enrolled), whose gestational age was set by transvaginal crown-rump length. The person scanned is the pregnant woman; the quantity estimated is the gestational age of the pregnancy.; Care setting: Low-cost battery-powered device used by novice users with no prior training in sonography, at follow-up visits; Geography: Lusaka, Zambia, and Chapel Hill, North Carolina (United States); Life stage: Pregnancy; primary evaluation window 14 0/7 to 27 6/7 weeks of gestation; Gestational age: 14 weeks 0 days (98 days) to 27 weeks 6 days (195 days), measured at the time of measurement; Postnatal age: Not reported; Follow-up: Random follow-up visits throughout gestation - Limitations: Tested at two sites and in one gestational window (14 to 27 weeks), with first-trimester crown-rump length as the reference. It measures dating accuracy, not any effect on care or outcomes. It is a research tool; the sources do not show routine use. Only the abstract was read; the registry number printed there (NCT05433519) was not fetched. - Last checked: 2026-10-14 - Sources: 1. Supports: Diagnostic Accuracy of an Integrated AI Tool to Estimate Gestational Age From Blind Ultrasound Sweeps (Publication, edition PubMed record (abstract and bibliographic data) from NCBI E-utilities, retrieved 2026-10-14. The full text and the version of record were not read.) Locator: Abstract, DESIGN, SETTING, AND PARTICIPANTS, sentence 1. Retrieved 2026-10-14T18:18:29Z. Source state: Current. Snapshot hash: 983cc77f9baeba4430473c25ed0df98926e093f8a2207d36ac68c5cc45c2ce94. How it applies: PubMed abstract. 2. Supports: Diagnostic Accuracy of an Integrated AI Tool to Estimate Gestational Age From Blind Ultrasound Sweeps (Publication, edition PubMed record (abstract and bibliographic data) from NCBI E-utilities, retrieved 2026-10-14. The full text and the version of record were not read.) Locator: Abstract, DESIGN, SETTING, AND PARTICIPANTS, sentence 2. Retrieved 2026-10-14T18:18:29Z. Source state: Current. Snapshot hash: 983cc77f9baeba4430473c25ed0df98926e093f8a2207d36ac68c5cc45c2ce94. How it applies: PubMed abstract. 3. Supports: Diagnostic Accuracy of an Integrated AI Tool to Estimate Gestational Age From Blind Ultrasound Sweeps (Publication, edition PubMed record (abstract and bibliographic data) from NCBI E-utilities, retrieved 2026-10-14. The full text and the version of record were not read.) Locator: Abstract, DESIGN, SETTING, AND PARTICIPANTS, sentence 3, fragment 1 of 3. Retrieved 2026-10-14T18:18:29Z. Source state: Current. Snapshot hash: 983cc77f9baeba4430473c25ed0df98926e093f8a2207d36ac68c5cc45c2ce94. How it applies: PubMed abstract. 4. Supports: Diagnostic Accuracy of an Integrated AI Tool to Estimate Gestational Age From Blind Ultrasound Sweeps (Publication, edition PubMed record (abstract and bibliographic data) from NCBI E-utilities, retrieved 2026-10-14. The full text and the version of record were not read.) Locator: Abstract, DESIGN, SETTING, AND PARTICIPANTS, sentence 3, fragment 2 of 3. Retrieved 2026-10-14T18:18:29Z. Source state: Current. Snapshot hash: 983cc77f9baeba4430473c25ed0df98926e093f8a2207d36ac68c5cc45c2ce94. How it applies: PubMed abstract. 5. Supports: Diagnostic Accuracy of an Integrated AI Tool to Estimate Gestational Age From Blind Ultrasound Sweeps (Publication, edition PubMed record (abstract and bibliographic data) from NCBI E-utilities, retrieved 2026-10-14. The full text and the version of record were not read.) Locator: Abstract, DESIGN, SETTING, AND PARTICIPANTS, sentence 3, fragment 3 of 3. Retrieved 2026-10-14T18:18:29Z. Source state: Current. Snapshot hash: 983cc77f9baeba4430473c25ed0df98926e093f8a2207d36ac68c5cc45c2ce94. How it applies: PubMed abstract. 6. Supports: Diagnostic Accuracy of an Integrated AI Tool to Estimate Gestational Age From Blind Ultrasound Sweeps (Publication, edition PubMed record (abstract and bibliographic data) from NCBI E-utilities, retrieved 2026-10-14. The full text and the version of record were not read.) Locator: Abstract, MAIN OUTCOMES AND MEASURES, sentence 1, fragment 2 of 3. Retrieved 2026-10-14T18:18:29Z. Source state: Current. Snapshot hash: 983cc77f9baeba4430473c25ed0df98926e093f8a2207d36ac68c5cc45c2ce94. How it applies: PubMed abstract. 7. Supports: Diagnostic Accuracy of an Integrated AI Tool to Estimate Gestational Age From Blind Ultrasound Sweeps (Publication, edition PubMed record (abstract and bibliographic data) from NCBI E-utilities, retrieved 2026-10-14. The full text and the version of record were not read.) Locator: Abstract, MAIN OUTCOMES AND MEASURES, sentence 1, fragment 3 of 3. Retrieved 2026-10-14T18:18:29Z. Source state: Current. Snapshot hash: 983cc77f9baeba4430473c25ed0df98926e093f8a2207d36ac68c5cc45c2ce94. How it applies: PubMed abstract. 8. Supports: Diagnostic Accuracy of an Integrated AI Tool to Estimate Gestational Age From Blind Ultrasound Sweeps (Publication, edition PubMed record (abstract and bibliographic data) from NCBI E-utilities, retrieved 2026-10-14. The full text and the version of record were not read.) Locator: Abstract, RESULTS, sentence 1, fragment 1 of 2. Retrieved 2026-10-14T18:18:29Z. Source state: Current. Snapshot hash: 983cc77f9baeba4430473c25ed0df98926e093f8a2207d36ac68c5cc45c2ce94. How it applies: PubMed abstract. 9. Supports: Diagnostic Accuracy of an Integrated AI Tool to Estimate Gestational Age From Blind Ultrasound Sweeps (Publication, edition PubMed record (abstract and bibliographic data) from NCBI E-utilities, retrieved 2026-10-14. The full text and the version of record were not read.) Locator: Abstract, RESULTS, sentence 1, fragment 2 of 2. Retrieved 2026-10-14T18:18:29Z. Source state: Current. Snapshot hash: 983cc77f9baeba4430473c25ed0df98926e093f8a2207d36ac68c5cc45c2ce94. How it applies: PubMed abstract. 10. Supports: Diagnostic Accuracy of an Integrated AI Tool to Estimate Gestational Age From Blind Ultrasound Sweeps (Publication, edition PubMed record (abstract and bibliographic data) from NCBI E-utilities, retrieved 2026-10-14. The full text and the version of record were not read.) Locator: Abstract, RESULTS, sentence 2, fragment 1 of 2. Retrieved 2026-10-14T18:18:29Z. Source state: Current. Snapshot hash: 983cc77f9baeba4430473c25ed0df98926e093f8a2207d36ac68c5cc45c2ce94. How it applies: PubMed abstract. 11. Supports: Diagnostic Accuracy of an Integrated AI Tool to Estimate Gestational Age From Blind Ultrasound Sweeps (Publication, edition PubMed record (abstract and bibliographic data) from NCBI E-utilities, retrieved 2026-10-14. The full text and the version of record were not read.) Locator: Abstract, RESULTS, sentence 2, fragment 2 of 2. Retrieved 2026-10-14T18:18:29Z. Source state: Current. Snapshot hash: 983cc77f9baeba4430473c25ed0df98926e093f8a2207d36ac68c5cc45c2ce94. How it applies: PubMed abstract. 12. Supports: Diagnostic Accuracy of an Integrated AI Tool to Estimate Gestational Age From Blind Ultrasound Sweeps (Publication, edition PubMed record (abstract and bibliographic data) from NCBI E-utilities, retrieved 2026-10-14. The full text and the version of record were not read.) Locator: Abstract, CONCLUSIONS AND RELEVANCE, sentence 1, fragment 1 of 2. Retrieved 2026-10-14T18:18:29Z. Source state: Current. Snapshot hash: 983cc77f9baeba4430473c25ed0df98926e093f8a2207d36ac68c5cc45c2ce94. How it applies: PubMed abstract. 13. Supports: Diagnostic Accuracy of an Integrated AI Tool to Estimate Gestational Age From Blind Ultrasound Sweeps (Publication, edition PubMed record (abstract and bibliographic data) from NCBI E-utilities, retrieved 2026-10-14. The full text and the version of record were not read.) Locator: Abstract, CONCLUSIONS AND RELEVANCE, sentence 1, fragment 2 of 2. Retrieved 2026-10-14T18:18:29Z. Source state: Current. Snapshot hash: 983cc77f9baeba4430473c25ed0df98926e093f8a2207d36ac68c5cc45c2ce94. How it applies: PubMed abstract. ## Connected pages - Uses low-cost probe sweeps: Pregnancy ultrasound (`maternal:obstetric-ultrasound`) - The NEJM Evidence paper reports funding from the Bill and Melinda Gates Foundation: Bill & Melinda Gates Foundation (`maternal:gates-foundation`) - Filed under this front: Imaging and monitoring (`maternal:front-imaging-and-monitoring`) ## Provenance Content hash: 44b873d76853c121699d5a1114670ec012803fb645dc59a7429fdb69dc91b261 Last substantive edit: 2026-10-14 Evidence last checked: 2026-10-14