I-Corps: Artificial Intelligence-Enabled Shoe Insoles to Assess Walking Function in Real Life Environments

该I-Corps项目更广泛的影响/商业潜力是开发敏感的定量评估，重点关注现实生活环境中的行走功能，以评价临床试验和上市后监测中神经系统疾病（如神经肌肉疾病、帕金森病、亨廷顿病和多发性硬化症）患者新治疗的安全性和有效性。步行是日常生活中最常见的活动，也是功能独立的关键因素。新的药物治疗正在改变几种神经系统疾病的治疗前景，改变它们的自然病程，在行走功能方面取得了令人鼓舞的结果。步行相关的数字移动性结果有可能客观地捕获日常表现，以补充临床评估和患者报告的结果，从而提供更全面的患者状况。在临床研究中使用可穿戴设备来评估新干预措施的有效性的兴趣正在上升。据估计，到2025年，近70%的临床试验将涉及可穿戴设备，而制药公司通过采用更客观、更敏感和更精细的数字移动结果，平均每年可以节省1亿美元的试验开发支出。然而，大多数用于现实生活步态监测的设备仅限于体积数字移动性结果，其易于捕获但缺乏逐步粒度。这项创新可以在患者的生活环境中纵向收集准确的体积和步幅步态指标，以帮助了解疾病轨迹如何受到新治疗的影响。通过在较短的时间内捕获细微但有临床意义的功能变化，这项创新可以降低临床试验的成本，为患者带来更快、更实惠的有效治疗。Corps项目基于抽象模型的开发，该抽象模型将用于可穿戴步态监测系统的联合收割机传统信号处理方法与机器学习回归的巨大表达能力和上级转换推理的个性化属性。这项创新的机器学习模型没有取代传统方法，而是使用一组优化的输入特征来校正它们的输出，从而生成准确的逐步时空和动态数字移动结果，而没有困扰端到端机器学习模型并阻碍其在嵌入式系统中实现的计算负担。通过利用转导推理的优势，创新的模型在延长时间的测量中提供了前所未有的准确性，而不需要特定于受试者的标记数据。这种新方法为下一代可穿戴技术提供了算法支持，以填补黄金标准实验室设备和新兴可穿戴步态监测设备之间的准确性差距，这阻碍了这些系统在临床研究中的广泛使用。通过扩大可穿戴步态监测设备的适用范围，该创新将促进对健康和临床人群的实际行走功能及其随时间变化的轨迹的理解。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 sensitive, quantitative assessments that focus on ambulatory function in real-life settings to evaluate safety and effectiveness of new treatments in clinical trials and post-marketing surveillance in individuals with neurological conditions, such as neuromuscular disorders, Parkinson’s disease, Huntington disease, and Multiple Sclerosis. Walking is the most frequent activity of daily living and a key contributor to functional independence. New pharmacological treatments are changing the therapeutic landscape for several neurological conditions, altering their natural history, with encouraging results in ambulatory function. Walking-related digital mobility outcomes have the potential to objectively capture daily performance to supplement clinical assessments and patient-reported outcomes, thereby providing a more comprehensive picture of a patient’s condition. Interest in the use of wearables in clinical research to assess the efficacy of new interventions is on the rise. It is estimated that by 2025 nearly 70% of all clinical trials will involve wearables, while the average pharma company can save $100 million/year in trial development spending by adopting more objective, sensitive, and granular digital mobility outcomes. Yet, most devices for real-life gait monitoring are limited to volume digital mobility outcomes, which are easy to capture but lack stride-by-stride granularity. This innovation could enable the collection of accurate volume and stride-by-stride gait metrics longitudinally, in patients’ living environments, to help understand how disease trajectories are affected by new treatments. By capturing subtle but clinically meaningful functional changes over shorter time periods, the innovation could reduce the cost of clinical trials and bring effective treatments to patients faster and more affordably.This I-Corps project is based on the development of abstraction models that synergistically combine conventional signal processing methods for wearable gait monitoring systems with the vast expressive capability of machine learning regression and the superior personalization properties of transductive inference. Instead of replacing conventional methods, this innovation's machine learning models correct their outputs using an optimized set of input features, thereby generating accurate stride-by-stride spatiotemporal and kinetic digital mobility outcomes without the computational burden that plagues end-to-end machine learning models and hinders their implementation in embedded systems. By leveraging the strengths of transductive inference, the innovation's models provide unprecedented accuracy over extended-time measurements without the need for subject-specific labelled data. This novel approach provides algorithmic support for next-generation wearable technology to fill the accuracy gap between gold-standard laboratory equipment and emerging wearable gait monitoring devices, which hampers the widespread use of these systems in clinical research. By widening the range of applicability of wearable gait monitoring devices, the innovation will promote the understanding of real-life ambulatory function and its trajectories over time in healthy and clinical populations.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.