SCH: EXP: Monitoring Motor Symptoms in Parkinson's Disease with Wearable Devices

帕金森病（PD）对老年人构成严重威胁，影响多达100万美国人。没有治愈方法，药物只能缓解症状。此外，仅在美国，与PD相关的费用，包括治疗，社会保障支付和因无法工作而损失的收入，估计每年接近250亿美元。PD管理的现有技术存在几个缺点：（1）频繁的门诊是PD治疗的高成本的主要原因，并且对患者来说是不方便的，特别是在旅行困难的人群中;（2）不准确的患者自我报告和15-20分钟的门诊不足以让医生准确评估他们的患者，导致难以监测患者症状和药物反应;和（3）运动功能评估是主观的，使得难以监测疾病进展。此外，因为它们必须由训练有素的临床医生执行，所以频繁进行运动功能评估是不可行的。该项目旨在通过使用可穿戴加速度计和机器学习的家庭监测来促进PD管理的范式转变。开发新的算法和实验协议，以允许在日常生活环境中的PD运动症状的鲁棒检测和评估。 具体来说，该项目开发了弱监督学习，时间序列分析和分类器个性化的算法。在以前的研究中，数据是在受控环境中收集的，时间很短（1-4小时），并手动标记以进行全监督学习。相比之下，这个项目收集的是长期（几周）的家庭数据，参与者的行为是自然的，没有脚本。参与者使用手机应用程序标记自己的数据，将时间段标记为包含或不包含PD运动症状的发生。由于症状的确切时间是未知的，这构成了弱标记数据。该项目扩展了多实例学习算法，用于从时间序列中的弱标记数据中学习。其他主要技术挑战包括在优化期间检测细微的运动症状和局部最小值。为了进一步提高鲁棒性和泛化能力，该项目探索使用个性化算法从无监督数据中学习运动症状的个人特定模型。所提出的弱监督学习和个性化技术是通用的，它们可以应用于其他人类感知问题。

Parkinson's Disease (PD) poses a serious threat to the elderly population, affecting as many as one million Americans. There is no cure, and medications can only provide symptomatic relief. In addition, costs associated with PD, including treatment, social security payments, and lost income from inability to work, is estimated to be nearly $25 billion per year in the United States alone. The current state-of-the-art in PD management suffers from several shortcomings: (1) frequent clinic visits are a major contributor to the high cost of PD treatment and are inconvenient for the patient, especially in a population for which traveling is difficult; (2) inaccurate patient self-reports and 15-20 minute clinic visits are not enough information for doctors to accurately assess their patients, leading to difficulties in monitoring patient symptoms and medication response; and (3) motor function assessments are subjective, making it difficult to monitor disease progression. Furthermore, because they must be performed by a trained clinician, it is infeasible to do frequent motor function assessments. This project aims to promote a paradigm shift in PD management through in-home monitoring using wearable accelerometers and machine learning. Novel algorithms and experimental protocols are developed to allow for robust detection and assessment of PD motor symptoms during daily living environments. Specifically, this project develops algorithms for weakly-supervised learning, time series analysis, and personalization of classifiers. In previous studies, data was collected in controlled environments for a short amount of time (1-4 hours) and manually labeled for fully-supervised learning. In contrast, this project collects long-term (several weeks), in-home data where the participants' actions are natural and unscripted. Participants use a cell phone app to label their own data, marking segments of time as containing or not containing the occurrence of a PD motor symptom. Since the exact time of the symptom is unknown, this constitutes weakly-labeled data. This project extends multiple-instance learning algorithms for learning from weakly-labeled data in time series. Additional major technical challenges include detection of subtle motor symptoms and local minima during optimization. To further increase robustness and generalization, this project explores the use of personalization algorithms to learn person-specific models of motor symptoms from unsupervised data. The proposed techniques for weakly-supervised learning and personalization are general, and they can be applied to other human sensing problems.