Radiomics approach to engineering an artificial intelligence based echocardiography platform to predict cardiovascular surgery and heart failure outcomes.

放射组学方法设计基于人工智能的超声心动图平台来预测心血管手术和心力衰竭的结果。

• 批准号: 10367037
• 负责人: William Hiesinger
• 金额: $ 58.8万
• 依托单位: STANFORD UNIVERSITY
• 依托单位国家: 美国
• 财政年份: 2022
• 资助国家: 美国
• 起止时间: 2022-01-01 至 2026-12-31
• 项目状态: 未结题
• 来源: https://reporter.nih.gov/project-details/10367037
• 关键词: Accounting; Acute; Address; Affect; Apical; Artificial Intelligence; Artificial Intelligence platform; Benchmarking; Cardiac; Cardiovascular Diseases; Cardiovascular Surgical Procedures; Clinical; Cloud Computing; Color; Complex; Computer Systems; Data; Data Set; Decentralization; Decision Support Systems; Development; Devices; Diagnosis; Disease; Echocardiography; Effectiveness; Engineering; Event; Expert Systems; Failure; Framingham Heart Study; Future; Generations; Goals; Hand; Heart failure; Histopathology; Hospitals; Human; Individual; Informatics; International; Manuals; Measurement; Medical; Medical Imaging; Methods; Modernization; Motion; Myocardial; Nature; Operative Surgical Procedures; Outcome; Patients; Pattern; Performance; Persons; Pharmacotherapy; Postoperative Complications; Quality Control; Research; Research Personnel; Risk; Scanning; Screening procedure; Secure; Severities; Slide; Source; Supervision; Survival Analysis; System; Techniques; Thoracic Radiography; Time; Training; Transplantation; United States; Ventricular; Visual; Work; base; cardiogenesis; clinical diagnosis; clinical practice; cohort; community setting; computerized tools; cost effective; density; design; disease diagnosis; early screening; heart function; high dimensionality; high risk; imaging modality; implantable device; individual patient; left ventricular assist device; machine learning framework; mortality; multimodality; novel; outcome prediction; point of care; portability; pulmonary arterial hypertension; radiomics; right ventricular failure; screening; survival outcome; survival prediction; tool

SUMMARY In recent years, artificial intelligence has enabled automated systems to meet or exceed the performance of clinical experts across a wide variety of medical imaging tasks, in applications ranging from disease diagnosis using Chest X-Rays to survival analyses using histopathology slides. All current automated echocardiography systems – much like human echocardiography reads – are inherently reductionist in nature; a complex sequence and pattern of cardiac contraction is reduced to an outline of one or more chambers, from which a few global metrics of heart function are then calculated. Despite the staggering increase in usable data, the vast majority of information contained in time-resolved echocardiography videos remain woefully underutilized. As opposed to treating echocardiography studies as videos intended solely for visual interpretation, the ‘radiomics’ approach treats medical images as high-dimensional datasets to be mined with advanced computational tools. The overall goals of this project are to further develop and validate our novel, generalizable, multi-modal artificial intelligence (AI) platform for analyzing time resolved echocardiography studies, to address this underutilization. The impact of such an ECHO AI system is immediately perceptible in the field of heart failure. An estimated 6.5 million people suffer from heart failure in the United States. Across the spectrum of severity in this disease, echocardiography remains the cornerstone of screening and clinical diagnosis, a guide for medical management and pharmacotherapy, and an essential tool for planning acute lifesaving surgical interventions. We propose to build on our preliminary research and ready access to high quality paired echocardiographic and clinical datasets to achieve the following goals: 1) Develop a surgical decision support system for end-stage heart failure patients considered for left ventricular assist device (LVAD) implant. 2) Expand and generalize our ECHO AI tools to enable downstream prediction of long-term survival and development of heart failure, in both asymptomatic individuals and patients with pulmonary arterial hypertension 3) Cloud and hardware integration of our ECHO AI platform. The end result of our research will be a powerful ECHO AI tool with that is translatable, and integrated into clinical practice.

总结 近年来，人工智能使自动化系统能够达到或超过 临床专家在各种医学成像任务，在应用范围从疾病诊断 使用胸部X光片进行生存分析，使用组织病理学切片。所有当前自动超声心动图 系统--很像人类的超声心动图读数--本质上是固有的还原主义者;一个复杂的序列 心脏收缩的模式被简化为一个或多个腔室的轮廓， 然后计算心脏功能的度量。尽管可用数据的数量惊人地增加，但绝大多数 时间分辨超声心动图视频中包含的信息仍然远远没有得到充分利用。而不是 “放射组学”方法将超声心动图研究视为仅用于视觉解释的视频 将医学图像视为要使用高级计算工具挖掘的高维数据集。整体 这个项目的目标是进一步开发和验证我们的新颖的、可推广的、多模态的人工智能 (AI)分析时间分辨超声心动图研究的平台，以解决这一利用不足的问题。 这种ECHO AI系统在心力衰竭领域的影响是立即可察觉的。估计6.5 在美国有100万人患有心力衰竭。在这种疾病的严重程度范围内， 超声心动图仍然是筛查和临床诊断的基石，是医疗管理的指南 和药物治疗，以及规划急性救生手术干预的重要工具。我们建议 建立在我们的初步研究和高质量的配对超声心动图和临床数据集的基础上 目的：1）开发一个针对终末期心力衰竭患者的手术决策支持系统 考虑植入左心室辅助装置（LVAD）。2)扩展和推广我们的ECHO AI工具， 使下游预测长期生存和发展的心力衰竭，在无症状和 3）ECHO AI的云和硬件集成 平台我们研究的最终结果将是一个强大的ECHO AI工具，它是可翻译的，并且是集成的。 临床实践。