Predicting Pulmonary and Cardiac Morbidity in Preterm Infants with Deep Learning

通过深度学习预测早产儿的肺部和心脏发病率

• 批准号: 10646498
• 负责人: Andrew L. Beam
• 金额: $ 16.63万
• 依托单位: HARVARD SCHOOL OF PUBLIC HEALTH
• 依托单位国家: 美国
• 财政年份: 2019
• 资助国家: 美国
• 起止时间: 2019-07-01 至 2024-06-30
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/10646498
• 关键词: 37 weeks gestation; Accounting; Acute; Address; Adverse event; Affect; Area; Award; Big Data; Bioinformatics; Biometry; Birth; Birth Weight; Boston; Bronchopulmonary Dysplasia; Cardiac; Categories; Chronic; Clinical; Clinical Data; Clinical Medicine; Collaborations; Computational Technique; Conceptions; Data; Data Analyses; Data Scientist; Data Sources; Databases; Decision Making; Diagnosis; Doctor of Philosophy; Education; Electronic Health Record; Environment; Event; Future; Gestational Age; Goals; Graduate Degree; Grant; Health; Healthcare; Healthcare Systems; Heart; Heart Diseases; Hospitals; Human Pathology; Incidence; Infant; Informatics; Institutional Review Boards; Insurance Carriers; Life; Liquid substance; Lung; Lung diseases; Machine Learning; Measures; Mentors; Mentorship; Methodology; Modeling; Monitor; Morbidity - disease rate; Nature; Necrotizing Enterocolitis; Neonatal; Neonatal Intensive Care Units; Neonatology; Outcome; Patent Ductus Arteriosus; Patients; Pattern; Pediatrics; Performance; Perinatal; Physicians; Physiological; Population; Pregnancy; Pregnant Women; Premature Birth; Premature Infant; Premature Labor; Reproducibility; Research; Research Personnel; Research Support; Rest; Retinopathy of Prematurity; Risk Estimate; Scientist; Sepsis; Side; Signal Transduction; Source; Teaching Hospitals; Techniques; Time; Training; Translating; Update; Vulnerable Populations; Work; clinical practice; clinical predictors; data resource; deep learning; deep learning algorithm; deep learning model; design; education planning; electronic health data; experience; improved; insurance claims; model building; mortality; peer; portability; prediction algorithm; predictive modeling; premature; prognostic; respiratory distress syndrome; risk prediction; risk prediction model; social; statistics; structured data

RESEARCH SUMMARY The goal of this award is to provide Andrew Beam, PhD with research support and comprehensive mentoring designed to transition him to an independent investigator in perinatal and neonatal informatics. Preterm labor (PTL) is labor which occurs before 37 weeks of gestation and carries with it enormous health and financial consequences. Preterm infants have some of the highest levels of pulmonary and cardiac morbidity, yet machine-learning techniques for these important outcomes remains under developed. The research strategy is focused developing predictive models for two very important clinical scenarios using large sources of existing healthcare data. The focus of Specific Aim 1 develops a new form of machine learning known as deep learning for predicting PTL in pregnant women, while the focus of Specific Aim 2 investigates the use of deep learning for predicting clinical trajectories of preterm infants in the NICU. Currently, management and anticipation of both clinical scenarios is challenging and advancement in our predictive capacity could dramatically improve the quality and efficiency of the healthcare system. These models will be built using an existing database of 50 million patient-lives obtained through a partnership with a major US health insurer. Specific Aim 3 seeks to understand how the models constructed using this unique data resource translate and generalize to data from the electronic health records of Boston-area hospitals, which is a key concern for all healthcare data scientists. The education plan focuses on augmenting Dr. Beam’s graduate degrees in statistics and bioinformatics with additional training in clinical medicine and human pathology. This additional education will grant Dr. Beam a deeper understanding of the clinical problems faced by these populations and will allow for more fluid collaborations with clinicians in the future. The composition of Dr. Beam’s mentorship committee, which includes expertise in neonatology, biostatistics, and translational informatics, reflects his long-term desire to be quantitative scientist who works side-by-side practicing physicians so that quantitative research is translated into impactful clinical practice.

研究综述 该奖项的目标是为安德鲁梁博士提供研究支持和全面的指导 旨在将他转变为围产期和新生儿信息学的独立研究者。早产 (PTL)是在怀孕37周之前发生的分娩，伴随着巨大的健康和经济损失。 后果早产儿的肺部和心脏病发病率最高， 用于这些重要成果的机器学习技术仍处于开发阶段。研究策略是 重点开发两个非常重要的临床场景的预测模型，使用现有的大量来源， 医疗保健数据。Specific Aim 1的重点是开发一种新形式的机器学习，称为深度学习 用于预测孕妇的PTL，而具体目标2的重点是研究深度学习的使用 用于预测新生儿重症监护室中早产儿的临床轨迹。目前，管理和预测 这两种临床场景都具有挑战性，我们预测能力的进步可以显着提高 医疗保健系统的质量和效率。这些模型将使用现有的50个数据库建立 通过与美国一家大型医疗保险公司的合作，具体目标3旨在 了解使用这种独特的数据资源构建的模型如何转换和推广到数据， 波士顿地区医院的电子健康记录，这是所有医疗保健数据科学家的关键问题。 教育计划的重点是增加博士梁的统计学和生物信息学的研究生学位， 临床医学和人类病理学的额外培训。这一额外的教育将授予梁博士 更深入地了解这些人群所面临的临床问题， 未来与临床医生的合作。比姆博士导师委员会的组成， 包括生物学，生物统计学和翻译信息学的专业知识，反映了他长期以来的愿望， 与执业医师并肩工作的定量科学家，以便将定量研究转化为 转化为有效的临床实践

项目成果

Reporting guidelines for clinical trial reports for interventions involving artificial intelligence: the CONSORT-AI extension.

• DOI: 10.1038/s41591-020-1034-x
• 发表时间: 2020-09
• 期刊: Nature medicine
• 影响因子: 82.9
• 作者: Liu X;Cruz Rivera S;Moher D;Calvert MJ;Denniston AK;SPIRIT-AI and CONSORT-AI Working Group
• 通讯作者: SPIRIT-AI and CONSORT-AI Working Group

Sharpening the resolution on data matters: a brief roadmap for understanding deep learning for medical data.

• DOI: 10.1016/j.spinee.2020.08.012
• 发表时间: 2021-10
• 期刊: The spine journal : official journal of the North American Spine Society
• 作者: Schmaltz A;Beam AL
• 通讯作者: Beam AL

Time to reality check the promises of machine learning-powered precision medicine.

• DOI: 10.1016/s2589-7500(20)30200-4
• 发表时间: 2020-12
• 期刊: LANCET DIGITAL HEALTH
• 影响因子: 30.8
• 作者: Wilkinson, Jack;Arnold, Kellyn F.;Murray, Eleanor J.;van Smeden, Maarten;Carr, Kareem;Sippy, Rachel;de Kamps, Marc;Beam, Andrew;Konigorski, Stefan;Lippert, Christoph;Gilthorpe, Mark S.;Tennant, Peter W. G.
• 通讯作者: Tennant, Peter W. G.

Safe and reliable transport of prediction models to new healthcare settings without the need to collect new labeled data.

将预测模型安全可靠地传输到新的医疗保健环境，而无需收集新的标记数据。

• DOI: 10.1101/2023.12.13.23299899
• 发表时间: 2023
• 期刊: medRxiv : the preprint server for health sciences
• 作者: Tuwani,Rudraksh;Beam,Andrew
• 通讯作者: Beam,Andrew

A Review of Challenges and Opportunities in Machine Learning for Health.

• 发表时间: 2018-06
• 期刊: AMIA Joint Summits on Translational Science proceedings. AMIA Joint Summits on Translational Science
• 作者: M. Ghassemi;Tristan Naumann;Peter F. Schulam;A. Beam;I. Chen;R. Ranganath