Tremor decomposition: A method for determining which muscles are most responsible for a patient's tremor

震颤是最常见的运动缺陷之一。它影响患有多种疾病的患者，包括原发性震颤，帕金森病，肌张力障碍，小脑共济失调（身体运动失控）等。对于这些患者，震颤使日常生活活动（进食，穿衣，写作等）困难或不可能。虽然药物治疗和手术干预已经显著减少了患者的痛苦，但它们只推荐给一部分患者。即使在这些患者中，干预措施也只是部分有效，使许多患者没有有效的治疗选择。令人惊讶的是，几乎没有震颤抑制设备可供患者使用。例如，可以设想设计用于抑制震颤的可穿戴上肢设备（例如，支架）。然而，开发有效的震颤抑制设备的一个重大障碍是，目前没有办法知道在哪里（哪些肌肉/关节）进行干预，因为没有确定哪些肌肉对患者的震颤最负责的关键方法。这项工作的目的是开发一种“震颤分解方法”，可以确定哪些肌肉对患者的震颤最负责任。与分析肌肉活动以预测关节处的运动的“合成方法”相反，“分解方法”在相反的方向上工作，即，分析关节处的运动以预测涉及哪些肌肉。 该方法可以应用于所有类型的震颤，无论震颤是病理性的（例如，与损伤或疾病有关的震颤）或生理的（例如，在健康成人中常见的震颤）。一旦识别出震颤的起源（或治疗目标），该方法就可以应用于多种震颤抑制策略。最后，震颤是一个跨学科的问题，震颤抑制需要工程和神经科学方面的专业知识。 因此，来自工程和神经科学的研究助理（本科生和研究生）将在团队中共同完成这项研究的各个方面，从而培养出习惯于在跨学科团队中共同工作以解决影响问题的毕业生。为了帮助震颤研究社区，将创建一个网站，提供数据集和开发的计算方法，包括与数据交互的能力，以查看哪些肌肉有助于选定的运动。 该项目的重点是开发和验证一种称为“震颤分解方法”的方法，以确定哪些肌肉对患者的震颤最负责任（以便将来能够有针对性地抑制震颤）。 这项工作建立在联合研究者最近发现的数学上相似的问题的解决方案的基础上：表面肌电（sEMG）分解。虽然sEMG生成和震颤传播在物理上是不同的，但它们在数学上是相似的。因此，第一个目标是通过应用用于求解sEMG分解的方法来开发震颤分解方法，即，收件人：1）将震颤传播建模为通过未知脉冲响应过滤的源的混合，2）确定是否充分满足单独从输出（关节位移）的测量执行分解所需的唯一性要求，以及3）针对震颤，调整被证明成功用于sEMG分解的两种方法（空间滤波和盲源分离）。 第二个目标是使用从诊断为原发性震颤的患者收集的现有数据集，在震颤严重程度、运动条件、肢体配置和受试者特征的范围内验证分解方法。 数据集包括输入（sEMG）和输出（关节位移），因此可以用于确定是否可以单独从关节位移执行震颤分解。 如果没有，则将使用与sEMG分解工作中使用的方法类似的方法来改变该方法。 第三个目标是将分解方法应用于现有数据集以确定震颤起源的模式，即，以确定时不变或甚至患者通用震颤抑制装置的可行性。 虽然该方法是在原发性震颤的背景下开发的，但所提出的方法适用于所有类型的震颤，并有可能提高几乎所有外周震颤抑制策略的有效性。该奖项反映了NSF的法定使命，并被认为值得通过使用基金会的智力价值和更广泛的影响审查标准进行评估来支持。

Tremor is one of the most common movement deficits. It affects patients suffering from multiple disorders, including Essential Tremor, Parkinson's disease, Dystonia, Cerebellar Ataxia (loss of control of bodily movements) and others. For these patients, tremor makes activities of daily living (eating, clothing, writing, etc.) difficult or impossible. Although medication and surgical interventions have significantly reduced patient suffering, they are only recommended for a subset of patients. And even in these patients, interventions are only partially effective, leaving many patients without effective treatment options. Surprisingly, there are few tremor-suppressing devices available to patients. One might envision, for example, a wearable upper-limb device (e.g. a brace) designed to suppress tremor. However, a significant obstacle to developing effective tremor-suppressing devices is that, currently, there is no way to know where (which muscles/joints) to intervene because there is no key way of determining which muscles are most responsible for a patient's tremor. The purpose of this work is to develop a "tremor decomposition method" that can determine which muscles are most responsible for a patient's tremor. In contrast to a "composition method" that would analyze muscle activities to predict a movement at a joint, a "decomposition method" works in the opposite direction, i.e., analyzes the movement at a joint to predict what muscles are involved. The method can be applied to all types of tremor, whether the tremor be pathological (e.g., tremor related to injury or disease) or physiological (e.g., tremor common in healthy adults). Once the origin of the tremor (or treatment target) is identified, the method can be applied to a wide diversity of tremor-suppressing strategies. Finally, tremor is an interdisciplinary problem, and tremor suppression requires expertise in both engineering and neuroscience. Thus, research assistants (undergraduate and graduate students) from both engineering and neuroscience will work together in teams to complete all aspects of this research, leading to graduates who are used to working together on interdisciplinary teams to solve problems of impact. To assist the tremor research community in general, a website will be created that will provide access to datasets and the computational methods developed, including the ability to interact with the data to see which muscles contribute to selected movements. This project is focused on developing and validating a method, termed "tremor decomposition method," to determine which muscles are most responsible for a patient's tremor (to enable targeted tremor suppression in the future). The work builds on the Co-Investigator's recent discovery of a solution to a problem that is mathematically similar: surface electromyogram (sEMG) decomposition. Although sEMG generation and tremor propagation are physically different, they are mathematically analogous. Thus the first objective is to develop the tremor decomposition method by applying the method used to solve sEMG decomposition, i.e., to: 1) model tremor propagation as a mixture of sources filtered by unknown impulse responses, 2) determine if the uniqueness requirement necessary to perform decomposition from measurements of the output (joint displacement) alone is sufficiently satisfied, and 3) adapt, for tremor, two methods (spatial filtering and blind source separation) that proved successful for sEMG decomposition. The second objective is to validate the decomposition method across a range of tremor severities, movement conditions, limb configurations, and subject characteristics using an existing data set collected from patients diagnosed with Essential Tremor. The data set includes both input (sEMG) and output (joint displacement), thus can be used to determine if tremor decomposition can be performed from joint displacement alone. If not, the method will be altered using a method similar to that used in the sEMG decomposition work. The third objective is to apply the decomposition methods to the existing dataset to determine patterns in tremor origin, i.e., to determine the feasibility of time-invariant or even patient-generic tremor-suppressing devices. Although the method is developed in the context of Essential Tremor, the proposed method applies to all types of tremors and has the potential to increase the efficacy of almost all peripheral tremor-suppressing strategies.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.