I-Corps: Artificial Intelligence Models to Improve Heart Failure Management

该I-Corps项目的更广泛影响/商业潜力是开发可以改善许多利益相关者心力衰竭管理的技术，包括医院，综合健康组织（IHO），医疗保险公司，电子健康记录（EHR）供应商，医疗器械制造商，制药公司和合同研究组织（CRO）。医院、IHO和医疗保险公司可能会看到该技术在改善患者预后、延长生产寿命和降低医疗成本方面的潜力。制药公司、医疗器械制造商和CRO也可能看到该技术在加快和改善临床试验患者招募方面的潜力。此外，该技术最终可以应用于邻近的慢性疾病，如慢性阻塞性肺病（COPD）和肾衰竭。I-Corps项目的愿景是开发一个人工智能（AI）平台，整合孤立的患者数据，以改善慢性病的管理。该项目基于机器学习（ML）算法的开发，对心力衰竭患者进行分层和表型，以识别有不良事件风险的患者，并推荐适当的干预措施。ML模型是在数百万个去识别的患者记录上训练的，包括胸部X光照片和数字数据（例如，生命体征和实验室检查）。经过训练的人工智能（AI）模型从临床数据中识别出细微的指标和模式，并在粒度级别上对患者进行分层。深度表型分析有可能个性化心力衰竭管理的干预计划。该团队开发了机器学习（ML）算法，以从胸片中定量评估患者液体过载，发现胸片包含可用于预测心力衰竭患者再入院的信息。30天的读取使命率与AI模型根据出院前胸部X光图像估计的肺水肿严重程度密切相关。该奖项反映了NSF的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

The broader impact/commercial potential of this I-Corps project is the development of technology that could improve heart failure management for many stakeholders, including hospitals, integrated health organizations (IHOs), medical insurers, electronic health records (EHRs) vendors, medical device manufacturers, pharmaceutical companies, and contract research organizations (CROs). Hospitals, IHOs and medical insurers may see value in the technology’s potential to improve patient outcomes, increase productive lifespans, and reduce healthcare costs. Pharmaceutical companies, medical device manufacturers, and CROs may also see value in the technology’s potential to expedite and improve patient recruitment for clinical trials. Furthermore, the technology could ultimately be applied to adjacent chronic conditions, such as chronic obstructive pulmonary disease (COPD) and kidney failure. The vision is to develop an artificial intelligence (AI) platform that integrates siloed patient data to improve management of chronic conditions.This I-Corps project is based on the development of machine learning (ML) algorithms that stratify and phenotype heart failure patients to identify those at risk of adverse events and recommend appropriate interventions. The ML models are trained on millions of de-identified patient records including chest radiographs and numerical data (e.g., vital signs and lab tests). The trained artificial intelligence (AI) models identify subtle indicators and patterns from the clinical data and stratify patients on a granular level. The deep phenotyping has the potential to personalize intervention plans for heart failure management. The team developed machine learning (ML) algorithms to quantitatively assess patient fluid overload from chest radiographs, finding chest radiographs contain information that can be used to predict readmissions of heart failure patients. The 30-day readmission rate correlates well with the pulmonary edema severity that the AI model estimates based on the pre-discharge chest x-ray images.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.