I-Corps: Artificial Intelligence (AI)-enabled point-of-care diagnostic ultrasound for injured children

I-Corps项目的更广泛影响/商业潜力是开发一种支持人工智能（AI）的超声诊断技术，以安全地诊断受伤儿童的内出血。目前的治疗标准是使用计算机断层扫描（CT）来诊断内出血，然而，每年有4000名儿童直接因电离辐射而患上癌症。此外，从急救人员下令进行CT扫描到放射科医生的最终报告可能长达24小时。提出的人工智能引导的超声诊断技术可能能够在没有致癌辐射的情况下，在几分钟内诊断出儿童的内出血。拟议中的技术还可以为那些不住在创伤中心附近的人提供专家级的紧急护理。目前，美国有3000万人居住在离创伤中心不到一小时路程的地方，也没有机会进行CT检查。I-Corps项目的基础是开发一种支持人工智能（AI）的软件，与现有的超声波技术一起使用，以安全诊断受伤儿童的内出血。所提出的软件可以在创伤检查期间指导超声集中评估的正确图像的获取，并可以通过识别游离液体来帮助解释结果，表明内部出血。已经开发出一种视图分类模型，可以准确识别儿科创伤检查的正确图像（准确率约为98%），并且目前正在开发一种新的机器学习算法来检测内出血。这些模型将集成到一个软件平台中，该平台可以指导超声图像采集，协助内出血诊断，并自动记录工作流程。该奖项反映了美国国家科学基金会的法定使命，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 an Artificial Intelligence (AI)-enabled, ultrasound diagnostic technology to safely diagnose internal bleeding in injured children. The current standard of care is to use a computerized tomography (CT) scan to diagnose internal bleeding, however, 4,000 of these children will develop cancer directly caused by the ionizing radiation each year. In addition, from the time the emergency provider orders the CT scan to the radiologist’s final report may be up to 24 hours. The proposed AI-guided ultrasound diagnostic technology may be able to diagnose internal bleeding in childrenwithout cancer-causing radiation and within minutes of arrival. The proposed technology also may give access to expert-level emergency care to those who do not live near a trauma center. Currently, 30 million people in the US do not live within an hour of a trauma center, and do not have access to CT. This I-Corps project is based on the development of an Artificial Intelligence (AI)-enabled software to use with existing ultrasound technology to safely diagnose internal bleeding in injured children. The proposed software may guide the acquisition of the correct images of the focused assessment with sonography during the trauma exam and may help interpret the results by identifying free fluid, indicating internal bleeding. A view classification model has been developed that accurately identifies the correct images for the pediatric trauma exam (accuracy of ~98%), and a novel machine learning algorithm to detect internal bleeding is currently in development. These models will be integrated into a software platform that may guide the ultrasound image acquisition, assist in the diagnosis of internal bleeding, and automate documentation workflow.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.