Improving Critical Congenital Heart Disease Screening and Detection of "Secondary" Targets

项目摘要/摘要 我们建议开发一种自动的危重先天性心脏病(CCHD)筛查算法 使用机器学习技术来结合非侵入性的灌流和氧合测量。 基于血氧饱和度(SpO2)的筛查是目前CCHD筛查的标准，但它不能 每年在美国检测高达50%的患有CCHD的无症状新生儿或近900名新生儿。 SpO2筛查漏掉的大多数新生儿都有主动脉阻塞的缺陷，如缩窄 在主动脉(CoA)，不会导致无氧血液进入循环。非侵入性测量 血流灌注指数(PIX)和脉搏血氧饱和度波形分析有望改善 检测患有CoA等缺陷的新生儿，这是SpO2目前最常遗漏的CCHD 放映。PIX和脉搏血氧饱和度测量波形都可以非侵入性地测量，并且使用相同的 用于SpO2筛查的设备。 我们团队的成员最近发现，添加PIX，一种非侵入性测量脉搏的方法 血流，有可能改善CCHD的检测，否则SpO2筛查会遗漏。然而， PIX在短时间(秒)内的变异性及其解释中的人为错误限制了其临床应用 能力。此外，对当前SpO2筛查算法的解释中的人为错误导致遗漏 健康新生儿的诊断和不适当的检测。因此，自动SpO2-PIX筛查 需要算法来简化筛选过程，并改进对遗漏缺陷的检测 通过SpO2筛查。为了实现这一点，我们将确定最佳的PIX波形，以创建 区分患有和不患有CCHD的新生儿。我们将进行脉搏血氧饱和度波形分析 确定其他对CCHD新生儿具有鉴别能力的非侵入性组件。此外，我们 将应用有监督的机器学习技术来自动执行算法解释。 这项拟议的研究意义重大，因为自动SpO2-PIX筛查算法可以节省 数百名未经SpO2筛查确诊的CCHD新生儿的生命。此外，这是 创新，因为它将是第一个在患有CCHD的新生儿中自动解释PIX测量和 自动PIX和SPO2的合并，这将允许在以后的步骤中轻松实施。穿过 与四个儿科心脏中心合作，我们将建立必要的基础设施和 多学科关系，以进行未来的多中心研究，以评估这种新的SpO2-PIX组合 大规模的算法，涉及数以千计的新生儿。改进CCHD的检测将需要 在所有参与新生儿护理和筛查的个人中采取多学科方法 CCHD。此外，与工程和计算机科学的合作将是必要的，以实现自动化 SpO2-Pix CCHD筛选算法。

PROJECT SUMMARY/ABSTRACT We propose to develop an automated critical congenital heart disease (CCHD) screening algorithm using machine learning techniques to combine non-invasive measurements of perfusion and oxygenation. Oxygen saturation (SpO2)-based screening is the current standard for CCHD screening, however it fails to detect up to 50% of asymptomatic newborns with CCHD or nearly 900 newborns in the United States annually. The majority of newborns missed by SpO2 screening have defects with aortic obstruction, such as coarctation of the aorta (CoA), that do not result in deoxygenated blood entering circulation. Non-invasive measurements of perfusion such as perfusion index (PIx) and pulse oximetry waveform analysis is expected to improve the detection of newborns with defects such as CoA, which is currently the most commonly missed CCHD by SpO2 screening. Both PIx and pulse oximetry waveforms can be measured non-invasively and with the same equipment used for SpO2 screening. Members of our team recently showed that the addition of PIx, a non-invasive measurement of pulsatile blood flow, has the potential to improve CCHD detection otherwise missed by SpO2 screening. However, variability of PIx over brief time periods (seconds) and human error in its interpretation limit its clinical capabilities. Additionally, human error in interpretation of the current SpO2 screening algorithm leads to missed diagnoses and inappropriate testing in healthy newborns. Therefore, an automated SpO2-PIx screening algorithm is needed to both simplify the screening process, and improve detection of defects that are missed with SpO2 screening. In order to achieve that, we will identify the optimal PIx waveforms to create a metric that discriminates between newborns with and without CCHD. We will perform pulse oximetry waveform analysis to identify other non-invasive components with discriminatory capacity for newborns with CCHD. Additionally, we will apply supervised machine learning techniques to automate the algorithm interpretation. The proposed research is significant because an automated SpO2-PIx screening algorithm could save the lives of hundreds of newborns with CCHD that are not diagnosed by SpO2 screening. Additionally, this is innovative as it will be the first automatic interpretation of PIx measurement among newborns with CCHD and merging of automated PIx and SpO2, which will allow for easy implementation at later steps. Through collaboration with four pediatric cardiac centers, we will establish the infrastructure and necessary multidisciplinary relationships to conduct future multicenter studies to evaluate this novel combined SpO2-PIx algorithm on a large scale involving thousands of newborns. Improving the detection of CCHD will require a multidisciplinary approach among all the individuals involved in the care and screening of newborns with CCHD. Additionally, collaboration with engineering and computer sciences will be necessary to automate the SpO2-PIx CCHD screening algorithm.