Population-level Pulmonary Embolism Outcome Prediction with Imaging and Clinical Data: A Multi-Center Study

项目摘要 家长奖的目标是开发一种自动化医疗人工智能(AI-HC)，以实现护理点风险 使用融合深度学习策略对肺栓塞(PE)患者的预后进行分层 同时分析健康记录和成像数据。一个理想的私募股权投资风险评分系统不仅会 预测死亡率，但也评估急性PE造成的许多衰弱的长期后果的风险。是这样的 因此，系统将促进最佳管理，并且可能需要智能地使用临床， 实验室和成像数据结合在一起，以便为多个PE提供准确的患者特定风险评分 结果衡量标准。为了建立一个稳健的模型，家长奖将应用深度的分布式训练 美国四家大型医疗机构的学习模式。分发算法而不是数据 避免共享可单独识别的患者信息。如果成功，这个家长奖将是第一个 努力利用诊断成像(像素)数据与结构化和非结构化电子产品相结合 用于预测结果的医疗记录(EMR)数据。使用强大的临床、实验室和 成像数据，该系统将为多个PE结果测量提供特定于患者的风险评分。此外， 家长奖促进多中心合作，包括调查 针对不同人群的PE患者进行培训、测试并最终部署自动化 在各种临床环境中的预测模型。 与家长奖合作提供了一个独特的机会来解决两个紧迫的伦理问题：如何 在AI-HC造成危害之前，您是否预测、识别并解决了AI-HC的伦理问题？你怎么知道？ 确定AI-HC后，记录并向多个用户传达重要的道德约束 (包括人工智能的开发者)？补充剂团队与家长奖密切合作 研究人员对人工智能-HC的一般伦理和开发检查AI-HC的方法进行了研究。在这 补充：我们将试行一种方法来：1)确定可能随着发展和 为PE多地点部署AI-HC；以及2)就如何解决这些伦理问题达成共识。我们 3)还将就道德“标签”达成共识，以传达已确定和解决的道德问题 约束条件。在做1、2和3的过程中，我们将提炼一种确定和解决道德问题的通用方法 AI-HC面临的挑战以及如何沟通AI-HC已确定的伦理关切的路线图。

Project Summary The goal of the parent award is to develop an automated healthcare AI (AI-HC) to achieve point-of-care risk stratification for pulmonary embolism (PE) patient outcomes using a fusion deep learning strategy that can simultaneously analyze health records and imaging data. An ideal PE risk-scoring system would not only predict mortality, but also assess the risk for the many debilitating long-term consequences of acute PE. Such a system would, therefore, facilitate optimal management and would likely require intelligent use of clinical, laboratory, and imaging data together in order to provide accurate patient -specific risk scoring for multiple PE outcome measures. In order to build a robust model, the parent award will apply distributed training of deep learning models across four large US healthcare institutions. Distributing the algorithm rather than the data avoids sharing individually identifiable patient information. If successful, this parent award will be the first endeavor to leverage diagnostic imaging (pixel) data in combination with structured and unstructured electronic medical record (EMR) data to predict outcomes. Using a powerful combination of clinical, laboratory, and imaging data, this system will provide patient-specific risk scoring for multiple PE outcome measures. Further, the parent award fosters multi- center collaborations, including investigation of the generalizability of the approach to different populations of PE patients and to train, test, and ultimately deploy the automated predictive model in a variety of clinical environments. Partnering with the parent award presents a unique opportunity to address two pressing ethical questions: How do you anticipate, identify, and address ethical problems with AI-HC before they cause harm? How do you document and communicate important ethical constraints with AI-HC, once identified, to multiple users (including the developers of the AI)? The supplement team has worked closely with the parent award investigators on ethics of AI-HC generally and on developing approaches to examine AI-HC. In this supplement we will pilot an approach to: 1) identify ethical issues that may emerge with development and multi-site deployment of AI-HC for PE; and 2) develop consensus on how to address these ethical issues. We will also 3) develop consensus on an ethics “label” to communicate identified and addressed ethical constraints. In doing 1, 2 & 3 we will refine a generalizable approach for identifying and addressing ethical challenges with an AI-HC and a roadmap for how to communicate identified ethical concerns for AI-HC.