High-Resolution Transcranial Ultrasound Neuromodulation at Large Scale

大规模高分辨率经颅超声神经调节神经调节具有映射神经功能的潜力;增强我们的感知，运动和认知能力;并恢复因受伤或疾病而失去的感觉和运动功能。尽管进行了数十年的研究和开发，但最先进的非侵入性神经调节技术仍然遭受极差的空间分辨率（100-1000 mm 3）。该项目包括探索正交交叉超声束作为无创经颅手段的科学研究，用于大规模前所未有的0.1 mm 3空间分辨率神经调节。与其非侵入性对应物相比，这种交叉波束超声神经调节技术有可能将空间分辨率（焦点）提高几个数量级。因此，它将为一套全面的非侵入性神经接口提供一个独特的构建模块。它将为神经科学开辟新的机会，首先在动物中显着提高空间分辨率和非侵入性大脑神经调节的覆盖率。最终，它还将在人类的许多临床应用中具有巨大的转化潜力，例如治疗神经和精神疾病以及脑机接口。利用研究的多学科性质，该项目还包括围绕“机器学习启发的物理故障排除”框架创建的重要综合推广和教育组件，以影响K-12教师和学生，少数民族，本科生和研究生。故障排除框架将激发K-12学生对电气工程的兴趣，以招收更多学生（特别是女性）到这个专业，将教育从本科生到K-12教师和他们的学生的广泛受众（特别是大学预科女生）在这项研究的科学和应用，并透过有系统的疑难解答及解决问题的活动，提高教师及学生的研究技巧。电路和基于优化的机器学习的研究生课程也将通过多学科项目和客座讲座进行改造，以教育研究生设计和应用智能集成系统。该项目提出并探索高分辨率经颅超声刺激（HR-TUS）系统，其中颅外超声换能器阵列以电子方式引导≤ 1 MHz交叉聚焦超声束，在成像和机器学习模型的指导下，在不同的神经目标上使用0.1 mm 3的超声压力焦点。基于研究人员在集成电路，基于超声的系统，无线神经接口和图像分析机器学习方面的互补专业知识，该项目将为大规模HR-TUS建立基本基础，通过成像和机器学习模型引导正交交叉超声波束。该项目将通过开发基于波动方程的数值和计算模型来研究HR-TUS中人脑体积内的空间分辨率和覆盖范围的基本限制，以探索相控阵列的不同几何形状，频率和配置及其与正交交叉波束背景下的颅骨和脑组织的相互作用的影响。该项目还将探索成像和机器学习模型，用于在非卧床受试者中存在颅骨/组织对超声波束和位移的影响的情况下进行准确的解剖定位、聚焦和波束交叉。为了降低系统的复杂性，尺寸和功耗在大规模的三维刺激组织，该项目的新的解决方案是一个大的二维阵列的柔性基板上的最佳排列的模块化可选的线性阵列及其专用集成电路组成。本项目结束时的系统级演示将确定HR-TUS的可行性。具有机器学习模型的图像引导HR-TUS系统将为基于学习的声学引导经颅超声神经调节提供一流的平台（全声学）具有高空间分辨率（0.1 mm 3）大规模（整个大脑）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进行评估，被认为值得支持的搜索.

High-Resolution Transcranial Ultrasound Neuromodulation at Large ScaleNeuromodulation has the potential to map neural functions; enhance our perceptual, motor, and cognitive capabilities; and restore sensory and motor functions lost through injury or disease. Despite decades of research and development, state-of-the-art noninvasive neuromodulation techniques still suffer from extremely poor spatial resolution (100-1000’s of mm3). This project includes scientific research that explores orthogonal crossed beams of ultrasound as a noninvasive transcranial means for unprecedented 0.1 mm3 spatial resolution neuromodulation at large scale. Compared to its noninvasive counterparts, this crossed-beam ultrasound neuromodulation technology has the potential to improve the spatial resolution (focal spot) by several orders of magnitude. Therefore, it will yield a unique building block for a comprehensive set of noninvasive neural interfaces. It will open new opportunities in neuroscience with significant improvements in spatial resolution and coverage of noninvasive neuromodulation of the brain, initially in animals. Ultimately, it will also have huge translational potential for many clinical applications in humans, such as the treatment of neurological and psychiatric disorders and brain-machine interfaces. Leveraging the multidisciplinary nature of the research, this project also includes a significant integrated outreach and educational component created around a “Machine-Learning-inspired Physical Troubleshooting” framework to impact K-12 teachers and students, minorities, and undergraduate and graduate students. The troubleshooting framework will stimulate the interest of K-12 students in electrical engineering to recruit more students (particularly women) to this major, will educate a broad audience from undergraduate students to K-12 teachers and their students (particularly pre-college female students) in the science and applications of this research, and will enhance teachers’ and students’ research skills through systematic troubleshooting and problem-solving activities. Graduate curriculum on circuits and optimization-based machine learning will also be transformed with multidisciplinary projects and guest lectures to educate graduate students in the design and applications of smart integrated systems.This project proposes and explores high-resolution transcranial ultrasound stimulation (HR-TUS) system, in which extracranial ultrasound transducer arrays electronically steer ≤ 1 MHz crossed focused ultrasound beams, guided by imaging and machine learning models, at different neural targets with ultrasound pressure focal spots of 0.1 mm3. Building on the investigators’ complementary expertise in integrated circuits, ultrasound-based systems, wireless neural interfaces, and machine learning for image analysis, this project will establish the fundamental basis for large-scale HR-TUS with orthogonal crossed ultrasound beams guided by imaging and machine learning models. This project will investigate fundamental limits of spatial resolution and coverage within a human brain volume in HR-TUS by developing numerical and computational models based on wave equations to explore the effects of different geometries, frequencies, and configurations of phased arrays and their interactions with the skull and brain tissue in the context of orthogonal crossed beams. This project will also explore imaging and machine learning models for accurate anatomical targeting, focusing, and beam crossing in the presence of skull/tissue effects on ultrasound beams and displacements in ambulatory subjects. To reduce the system complexity, size, and power consumption in three-dimensional stimulation of tissues at large scale, the novel solution of this project is a large two-dimensional array on a flexible substrate consisting of optimally arranged modular selectable linear arrays and their application-specific integrated circuits. A system-level demonstration at the end of this project will establish the feasibility of the HR-TUS. The image-guided HR-TUS system with machine learning model will provide a first-in-class platform for learning-based acoustically guided transcranial ultrasound neuromodulation (all acoustic) with high spatial resolution ( 0.1 mm3) at large scale (over the whole brain).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.