CAREER: Contextually Informed Autonomous Robotic Surgery

及时和广泛获得医疗服务已成为当今最大的社会问题之一，并被Covid-19疫情推向了最前沿。使问题恶化的是，美国的普通外科医生越来越短缺，数百家农村医院预计将在未来几年内关闭。这个教师早期职业发展（Career）项目为未来实现自主手术提供了急需的研究，以解决这些广泛的社会问题，即医疗保健获取和公平挑战。具体来说，该项目涉及到实现自主手术机器人的一大步，即以转录的形式向它们灌输人体解剖学和生理学知识，这些知识可以被人工智能利用来自主执行手术。作为项目的一部分，我们制定了一个与研究计划紧密结合的教育计划。我们将为高中学生开发一个互动的课程游戏，并计划在课堂环境中体验，让他们在视频游戏中探索编程和生物学。为了实现该项目的总体目标，将研究用于表示具有语义属性的多模态、可变形的组织环境的密集和降阶模型，并使用神经网络来近似正反问题。定义了将人体作为语义导航问题的解剖路线图，并利用局部上下文进行语义定位和环境映射。对于语义定位，将定义在可变形场景和方法中利用语义标签的理论和策略，利用几何先验和主动操作来识别特征和分类场景对象。最后，给定情境信息观察模型，解剖情境可用于评估和优化安全计划和稳健轨迹，以及行为树等更高层次的决策。涉及编程和生物学教学的教育活动将侧重于从生物学角度教授解剖背景和从块编程角度教授基本顺序逻辑。作为互动课程的一部分，该活动包括动手编程部分，包括构建一个自主代理来解决教育外科模拟器视频游戏。在互动课程之后，将收集和分析有关他们对STEM以及生物学和计算机科学潜在跨学科性质的看法的调查数据。然后，通过将活动开源并结合我们所学到的经验教训，我们的目标是让我们的电子游戏接触到代表性不足的少数群体，并为他们提供体验计算机科学，工程学和生物学所提供的更广泛，多学科机会的机会。该项目由跨部门机器人基础研究项目支持，由工程（ENG）和计算机与信息科学与工程（CISE）联合管理和资助。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。

Timely and widespread access to medical care has become one of the greatest societal concerns today, pushed to the forefront by the Covid-19 outbreak. Exacerbating the problem is that there is a growing shortage of general surgeons in the US, and hundreds of rural hospitals are projected to close in the coming years. This Faculty Early Career Development (CAREER) project produces much-needed research towards realizing autonomous surgery in the future to address these broad societal issues of healthcare access and equity challenges. Specifically, the project involves taking a big step of realizing autonomous surgical robots is imbuing them with knowledge of human anatomy and physiology in a transcribed form that can be leveraged by artificial intelligence for performing procedures autonomously. As part of the project, an educational plan that integrates closely with the research plan is planned. An interactive lesson-game for high school students is to be developed, and an experience in a classroom setting is planned in which they will explore programming and biology in a video game. To achieve the overall goal of this project, dense and reduced-order models for representing multimodal, deformable tissue environments with semantic attributes will be investigated and neural networks to approximate the forward and inverse problem. An anatomical roadmap is defined that presents the human body as a semantic navigation problem, and local context can be leveraged to perform semantic localization and environment mapping. For semantic localization, theory and strategies for leveraging semantic labels in deformable scenes and methods will be defined, utilizing geometric priors and active manipulation to identify features and classify scene objects. Finally, given contextually informed observation models, anatomical context may be used to evaluate and optimize safe plans and robust trajectories as well as higher-level decision making such as behavior trees. The educational activity involving teaching students about programming and biology will focus on teaching anatomical context from a biological perspective and basic sequential logic from a block-programming perspective. As part of an interactive lesson, the activity includes a hands-on programming portion that involves building an autonomous agent to solve an educational surgical simulator videogame. Following the interactive lesson, survey data on their perceptions regarding STEM and the potential interdisciplinary nature of biology and computer science will be collected and parsed. Then, by making the activities open-sourced and incorporating our lessons learned, we aim for our videogame to reach underrepresented minorities and provide them an opportunity to experience the broader, multi-disciplinary opportunities that computer science, engineering, and biology have to offer. This project is supported by the cross-directorate Foundational Research in Robotics program, jointly managed and funded by the Directorates for Engineering (ENG) and Computer and Information Science and Engineering (CISE).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.