AI-driven low-cost ultrasound for automated quantification of hypertension, preeclampsia, and IUGR

AI 驱动的低成本超声可自动量化高血压、先兆子痫和 IUGR

• 批准号: 10567313
• 负责人: Gari David Clifford
• 金额: $ 65.53万
• 依托单位: EMORY UNIVERSITY
• 依托单位国家: 美国
• 财政年份: 2022
• 资助国家: 美国
• 起止时间: 2022-09-20 至 2027-08-31
• 项目状态: 未结题
• 来源: https://reporter.nih.gov/project-details/10567313
• 关键词: Address; Algorithms; Artificial Intelligence; Birth; Blood Pressure; Caring; Cellular Phone; Childhood; Code; Computer Systems; Country; Data; Detection; Development; Devices; Dimensions; Doppler Ultrasound; Fetal Development; Fetal Growth; Fetal Growth Retardation; Fetal Heart Rate; Foundations; Frontline worker; Funding; Gestational Age; Goals; Gold; Guatemala; Hand; Health; Health system; High-Risk Pregnancy; Hypertension; Image; Improve Access; Income; Infant; International; Knowledge; Label; Life; Machine Learning; Maternal Health; Maternal Mortality; Medical; Mission; Monitor; Morbidity - disease rate; National Institute of Child Health and Human Development; Neonatal Mortality; Observational Study; Outcome; Patients; Perinatal mortality demographics; Population; Pre-Eclampsia; Pregnancy; Pregnancy Complications; Prenatal care; Process; Reading; Research; Resource-limited setting; Resources; Running; Rural; Savings; Signal Transduction; System; Techniques; Technology; Time; Training; Transducers; Triage; Underserved Population; Validation; Work; base; clinical risk; cohort; cost; cost effective; cost efficient; deep learning algorithm; diagnostic accuracy; diagnostic platform; diagnostic strategy; diagnostic tool; diagnostic value; empowered; fetal; fetal medicine; field study; heart rate variability; high risk; improved; innovation; long short term memory network; low and middle-income countries; maternal hypertension; maternal risk; mortality; neonatal health; novel; novel strategies; perinatal outcomes; point of care; prediction algorithm; pregnancy disorder; prevent; prospective; public health relevance; risk prediction; screening; standard of care; stillbirth; technology validation; tool; tv watching; two-dimensional; ultrasound

PROJECT SUMMARY/ABSTRACT Life-saving advances in medical care in recent decades have reduced global mortality rates but have underperformed in addressing maternal mortality, stillbirth, and neonatal mortality. A key reason for these disparities in both low- and high-income settings is the lack of systematic screening with appropriate and affordable) technology for high priority conditions such as maternal hypertension and preeclampsia and fetal growth restriction. The development of new low-cost diagnostic tools to improve access to detection of these conditions by front-line workers would change outcomes for the most underserved populations, which is our long-term goal. In an NICHD-funded study, we collected point of care Doppler ultrasound recordings and developed a preliminary machine learning approach for detecting intrauterine growth restriction (IUGR) and maternal hypertension. The overall objective of this proposal is to prospectively validate these findings in two large underserved pregnancy cohorts in rural Guatemala and urban Georgia. Our general hypothesis is that our low-cost artificial intelligence will perform as well in detecting maternal hypertension, preeclampsia, and IUGR as standard-of-care high-cost diagnostic approaches. In Aim 1, we will validate our ultrasound-based IUGR detection algorithm against the standard of care (2-dimensional fetal imaging). In Aim 2, we will validate maternal hypertension and preeclampsia algorithms against gold-standard blood pressure devices and clinical risk prediction tools. In Aim 3, we will implement real-time versions of the algorithms validated in Aims 1 and 2 and implement them on an edge-computing system for field testing. Successful completion of this proposal will result in a novel and cost-effective approach to screening for maternal hypertension, preeclampsia, and IUGR using point-of-care Doppler connected to a low-cost, AI-enabled edge-computing system, suitable for wide use in low-resource settings. This proposal is innovative because it uses an artificial intelligence approach and widely-available point-of-care Doppler devices to provide new approaches to timely detection of high-impact maternal-fetal conditions. Our results will provide a strong basis for wide-scale deployment of new maternal and fetal screening technology which is expected to have a significant impact on maternal and fetal morbidity by improving access to timely screening. This research aligns with the NICHD's mission to advance knowledge of pregnancy, fetal development, and birth by promoting strategies that prevent maternal, infant, and childhood mortality and morbidity through lost-cost high-impact screening technology.

项目总结/摘要 近几十年来，医疗保健方面的救生进步降低了全球死亡率，但 在解决孕产妇死亡率、死产和新生儿死亡率问题方面表现不佳。一个关键原因是， 低收入和高收入环境中的差异是缺乏适当和 负担得起的）技术，用于高优先级的条件，如产妇高血压和先兆子痫和胎儿 增长限制。开发新的低成本诊断工具，以改善这些疾病的检测 一线工人的工作条件将改变最缺乏服务的人群的结果，这是我们的 长期目标。在NICHD资助的一项研究中，我们收集了床旁多普勒超声记录， 开发了一种初步的机器学习方法来检测宫内生长受限（IUGR）， 母亲高血压本提案的总体目标是在两个方面前瞻性地验证这些发现， 在危地马拉农村和格鲁吉亚城市，有大量未得到充分服务的怀孕妇女。我们的一般假设是， 低成本的人工智能将在检测孕产妇高血压、先兆子痫和IUGR方面表现良好 作为高成本诊断方法的标准护理。在目标1中，我们将验证我们基于超声的IUGR 检测算法与护理标准（二维胎儿成像）。在目标2中，我们将验证 针对金标准血压装置的孕产妇高血压和先兆子痫算法和临床 风险预测工具。在目标3中，我们将实现目标1中验证的算法的实时版本， 2并在边缘计算系统上实现，进行现场测试。成功完成此提案将 为筛查母体高血压、先兆子痫和IUGR提供了一种新的、具有成本效益的方法 使用连接到低成本、支持AI的边缘计算系统的即时多普勒，适合广泛使用 在低资源环境中。这一提议具有创新性，因为它使用了人工智能方法， 广泛使用的床旁多普勒设备，为及时检测高冲击提供新的方法 母胎情况我们的研究结果将为大规模部署新的孕产妇和新生儿保健服务提供强有力的基础。 胎儿筛查技术，预计将对孕产妇和胎儿发病率产生重大影响， 改善及时筛查的机会。这项研究符合NICHD的使命，以提高对 通过促进预防孕产妇、婴儿和儿童疾病的战略， 通过低成本高影响筛查技术降低死亡率和发病率。