Information from Lay-Language Summaries is Embargoed Until the Conclusion of the Scientific Presentation
374—Brain-Machine Interface: Limb Control
Monday, November 11, 2013, 8:00 am - 12:00 noon
374.08: Real-time prosthetic arm control using MEG signals of paralyzed patients
Location: Halls B-H
*T. YANAGISAWA1,2,3, R. FUKUMA4, K. MATSUSHITA1, M. HIRATA1, H. KISHIMA1, Y. SAITOH1, R. KATO5, T. SEKI5, H. SUGATA1, H. YOKOI5, Y. KAMITANI2, T. YOSHIMINE1; 1Dept. of neurosurgery, Osaka Univ., Osaka, Japan; 2ATR Computat. Neurosci. Labs., Kyoto, Japan; 3Div. of Functional Diagnos. Sci., Osaka Univ. Grad. Sch. of Med., Osaka, Japan; 4Nara Inst. of Sci. and Technol., Nara, Japan; 5The Univ. of Tokyo Interfaculty Initiative in Information Studies Grad. Sch. of Interdisciplinary Information Studies, Tokyo, Japan
Abstract Body: Objective: Prosthetic arm control using signals from intracranial electrodes has been shown to afford restoration of upper limb functions to paralyzed patients. For clinical application of the invasive BMI system, we need to know which patient is suitable for the therapy. Or, we need a non-invasive method to test the patients whether they will be able to use the invasive system after training. Here, we developed a prosthetic arm controlled by signals of magnetoencephalography (MEG) to train the patients to control the prosthetic arm by attempting to move their paralyzed limbs. Methods: A 160-channel Yokogawa MEG system with a real time output port was used. The entrained patients had severe paresis on their upper limbs due to trauma and stroke. All participants gave written informed consent to participate in the study, which was approved by the Ethics Committee of Osaka University Hospital. First, the MEG signals were obtained while the patients attempted to grasp or open their paralyzed hands with auditory and visual cues given every 5.5 s. The attempted movements were inferred by support vector machine and gaussian process regression using time-averaged MEG signals. The prosthetic arm was controlled to mimic the inferred movements. Then, for patient training session, the patients controlled the prosthetic arm to grasp and release an object at the instructed timing of every 7s. Electrical stimulation was applied to the patients’ hands as a sensory feedback when the prosthetic hand grasped an object. Results: The accuracy to infer the attempted movement type was varied among patients from approximately 60 to 80%. Statistical analysis revealed that the movement-related signals around the motor area significantly contributed to the prediction. Repeating the patient training sessions, the accuracy was improved for some patients. Interpretation: MEG signal has been shown to be applied to control the prosthetic arm even for the paralyzed patients. The developed system may be useful to train the patients to control the prosthetic arm and to evaluate the applicability of the motor restoration using invasive BMI.
Lay Language Summary: We have developed a prosthetic arm controlled by magnetoencephalographic (MEG) signals from paralyzed patients. Using this system, we have shown that severely paralyzed patients are able to control a prosthesis by imagining to move their paralyzed limbs. Although patients with electrodes implanted in their brain have been shown to be able to control prosthetic arms, there was previously no non-invasive technique to let patients learn how to control a prosthetic arm using their brain signals. Our achievement is significant for the clinical use of prosthetic arms controlled by brain signals to test and train patients considering invasive therapy, using a non-invasive method. Prosthetic arm control using brain signals is being studied in an effort to restore upper limb function in paralyzed patients. The efficacy of controlling a prosthetic arm smoothly enough to be used in the daily lives of patients has been demonstrated using invasive methods for recording brain signals, such as needle electrodes implanted in the brain or planar electrodes on the surface of the brain. However, the efficacy of controlling a prosthesis is different among patients. To clinically apply the invasive system for the patients, we need to know if the patient is suitable for the treatment before the implantation. Therefore, we need a non-invasive method to test whether patients will be able to use the invasive system after the implantation, so we developed a non-invasive system to control a prosthetic arm using MEG signals. MEG is a non-invasive neurophysiological recording technology that measures the magnetic fields generated by neuronal activity in the brain. It has a very high temporal and excellent spatial resolution. Moreover, because MEG is completely non-invasive, patients can use the system as many times as desired without any limitation. We entrained the patients with complete paresis on the upper limbs due to trauma, stroke and severe paresis on the whole body due to amyotrophic lateral sclerosis. They were asked to attempt to perform grasping or opening movements with their paralyzed hands, although their hands did not move at all. The attempted movements were inferred from the MEG signals using machine learning techniques. The accuracy to infer the attempted movement varied among patients from approximately 60 to 80%, though no statistical relationship was found between the severity of paresis and predicted movement accuracy. Some patients successfully controlled the prosthetic arm to catch and release a ball by imagining grasping and opening their paralyzed hands. The success rate to control the prosthesis was improved for some patients after repeating the training to catch the ball. Our findings show that a non-invasive system can be useful for improving a patient’s capability to control a prosthetic arm and evaluate the applicability of motor restoration using invasive methods. For a further study, we are developing and testing a novel rehabilitation therapy using this system to improve patients’ motor function itself.
Neuroscience 2013 (43rd annual meeting of the Society for Neuroscience)Exit