Brain Machine Interfaces based on Subcortical LFP Signals for Neuroprosthetic Control and Neurofeedback Therapy

恢复上肢功能将为因脊髓损伤、截肢、中风等致残而瘫痪的人提供一定程度的独立性和自主感。“脑机接口”（BMIs）将大脑信号转换为控制信号，用于指导假肢或其他设备，并显示出恢复日常生活中重要功能的巨大潜力，如伸手和抓握。然而，将令人兴奋的研究进展转化为实际改善残疾人日常生活的临床应用才刚刚开始。现有BMI的临床应用中的关键挑战包括：1）难以确保数月或数年内稳定且令人满意的脑信号记录。随着时间的推移，信号的丢失导致神经假体装置的性能恶化和用户的挫折感。2)难以准确可靠地估计某些运动参数，例如简单抓取运动中的夹持力。迄今为止，BMI的最佳临床演示仍然无法准确地操纵机械手施加的力水平。迄今为止，对BMI的研究几乎完全集中在从大脑表面或上层（大脑皮层）获得的信号上。我以前的研究已经确定了来自“基底神经节”的大脑信号的重要作用，基底神经节是大脑深处的一个结构，在控制运动和代表抓握力方面。这些信号可以容易地从持续多年的电极记录下来，并且这种电极可以以相对安全的程序植入，这已经是运动障碍的常规治疗。因此，这些信号为BMI随时间的长期性能和可靠性提供了显著优势。我建议使用这些大脑深层信号来控制机器人手的握力，并研究患者如何学习使用假手。这将为使用大脑深处结构记录的信号来控制机器人手提供重要的原理证明。同时，理解和参与BMI技能学习的过程可能是进一步改善BMI表现的另一个重要机会。重要的是，BMI使用的核心是让受试者通过自愿改变他们的大脑活动来实现特定目标。但是，BMI是否可以用来训练患者改变他们自己导致问题的病理性大脑活动？对这个问题的积极回答可以导致新的治疗方法，以治疗那些已经确定了明确的病理性大脑信号的疾病。例如，基底神经节中的病理性脑活动与一系列疾病中的运动损伤密切相关，例如帕金森病（PD）。我将使用这里提出的BMI系统来训练PD患者减少目标大脑区域的病理信号，同时给他们关于这种病理信号水平的反馈（所谓的“神经反馈训练”）。我将检验这样的假设，即当给予反馈时，患者能够减少导致其疾病问题的病理信号，并且病理活动的自愿变化可以导致PD中运动相关症状的改善。这项工作也将有助于揭示神经反馈训练的潜在机制，这可能有助于该技术的其他有效临床应用。总之，这项工作将建立基于从脑深部区域记录的信号的新型脑机接口的基础，这些信号包含与运动意图相关的丰富信息，并且已被证明随时间推移是稳定的。我将使用新的框架来控制假肢手与分级握力，提供神经反馈训练，以减少PD症状，并研究基底神经节在运动的控制和学习中的作用。

Recovering upper limb function will offer a certain degree of independence and sense of autonomy to people with paralysis due to disabling spinal cord injury, amputation, stroke etc. The 'Brain machine interfaces' (BMIs) convert brain signals into control signals for guiding prosthetic arms or other devices, and have showed great potential to restore functions important for everyday life, such as reaching and grasping. However, the translation of the exciting research progress to clinical use that actually improves the daily lives of people with disabilities has barely begun. Key challenges in the clinical applications of existing BMIs include: 1) Difficulties in ensuring stable and satisfactory recordings of brain signals over months or years. Loss of signals over time leads to deterioration in the performance of the neuroprosthetic device and frustration in users. 2) The difficulty in accurately and reliably estimating certain movement parameters such as the gripping force in a simple grasp movement. To date, the best clinical demonstration of BMI has still not been able to accurately manipulate the force level that was applied by a robotic hand.Research into BMI has, to date, almost exclusively focused on signals obtained from the surface or the upper layer of the brain (the cerebral cortex). My previous research has identified the important role of brain signals from the 'basal ganglia', a structure deep inside the brain, in controlling movement and representing gripping force in a grasp. These signals can be readily recorded from electrodes that last many years and such electrodes can be implanted in a relatively safe procedure, which has been a routine therapy for movement disorders. Therefore, these signals offer significant advantages for long-term performance and reliability of the BMI over time. I propose using these deep brain signals to control the grip force of robotic hands, and to study how the patients learn to use the prosthetic hand. This will provide important proof-of-principle of using signal recorded from structure deep inside the brain to control a robotic hand. Meanwhile, understanding and engaging the process of BMI skill learning will potentially be another major opportunity for further improvement of the performance of BMIs.Importantly, what is central to BMI use is for a subject to achieve a specific goal by voluntarily changing their brain activity. But could the BMI be used to train patients to change their own pathological brain activity that is causing problems? Positive answer to this question can lead to novel therapies to diseases where a clear pathological brain signal has been identified. For example, pathological brain activity in the basal ganglia has been heavily associated with motor impairment in a range of diseases, such as Parkinson's disease (PD). I will use the BMI system proposed here to train patients with PD to reduce the pathological signals in the targeted brain area while giving them the feedback about the level of this pathological signal (so called 'neurofeedback training'). I will test the hypothesis that when given feedback, patients are able to reduce the pathological signal that is causing problem in their disease, and that voluntary change of the pathological activity can lead to improvement in movement related symptoms in PD. This work will also help to shed light on the underlying mechanisms of neurofeedback training, which may facilitate other effective clinical applications of this technique.In summary, this work will establish the foundations for novel brain-machine interfaces based on signals recorded from deep brain regions that contain rich information related to movement intention and have been proven to be stable over time. I will use the new framework to control a prosthetics hand with graded gripping force, to provide neurofeedback training to reduce symptoms in PD, and to study the role of basal ganglia in the control and learning of movements.