Yanagisawa Laboratory

Introduction

1) Development and clinical application of BMI using ECoG/MEG

A brain-machine interface (BMI) is a system that decodes brain activity to control an external device such as a robotic arm, PC or communication device. The BMI enables a severely paralyzed patient to restore motor function and communication ability through neuroprosthesis. This technology allows communication even for patients without any muscle activity (i.e., those in a totally locked-in state), making BMI one of the biggest hopes for such patients.
We have developed the BMI using human electroencephalographic (ECoG) signals recorded by intracranial electrodes placed on the surface of the human brain (Figure 1) and magnetoencephalography (MEG; Figure 2).

Figure 1: ECoG recording
Figure 1: ECoG recording
Figure 2: MEG recording
Figure 2: MEG recording

Video 1 shows a robotic hand controlled by the BMI using ECoG signals. The desired hand movements of the subject implanted with the electrodes were successfully decoded from the online ECoG signals. We have also developed a robotic hand controlled by MEG signalshttps://media.nature.com/original/nature-assets/srep/2016/160224/srep21781/extref/srep21781-s2.mov). 


Video 1. Left: The left hand of the patient recording ECoG signals; Right: The robotic hand controlled by BMI using online ECoG signals.

Also, we are developing a novel BCI using ECoG in the project of JST CREST [Symbiotic interaction] Creation and development of core technologies interfacing human and information environments, Construction of representational Brain-Computer Interaction technology (website).

Our task is to improve the BMI so that it is suitable for clinical application.

2) BMI neurofeedback to understand and treat neurological diseases

BMI is a useful tool to induce plastic alteration of the cortical representations to modulate neurological diseases. We applied MEG-based BMI for patients with phantom limb pain, who suffer intractable pain after amputation or other severe peripheral nerve damage(https://media.nature.com/original/nature-assets/ncomms/2016/161027/ncomms13209/extref/ncomms13209-s2.mov). We found that if a patient tried to control a robotic hand using the BMI and associating the movement with their missing arm, it increased their pain, but training the patients to associate the movement of the prosthetic with the unaffected hand decreased their pain (Figure 3). These results demonstrated that BMI training could control phantom limb pain depending on the induced sensorimotor plasticity. We are applying the BMI neurofeedback technique to various neurological diseases to understand the neural mechanism of the diseases and to treat their symptoms.

Figure 3: BMI neurofeedback for phantom limb pain
Figure 3: BMI neurofeedback for phantom limb pain

We reported that signals measured in the subthalamic nucleus could be modulated by the patient’s intention, altering deep brain activity.

Details of the dissertation(Publishd:17 December 2018

Real-time neurofeedback to modulate β-band power in the subthalamic nucleus in Parkinsons disease patients

http://www.eneuro.org/content/early/2018/12/14/ENEURO.0246-18.2018

https://medicalxpress.com/news/2018-12-real-time-feedback-parkinson-brainwaves.html

 

We are involving the JST ERATO Ikegaya Brain-AI Hybrid project that aims to develop a novel human ability through neurofeedback based on decoding of neural information (Web page here).

 

3) Computer-aided diagnosis of neurological diseases using MEG/EEG

We are developing a novel computer-aided diagnostic tool using artificial intelligence with big data from MEG/EEG signals.

By analyzing the waveforms of EEG and MEG, we can find the features of various diseases. We have previously identified that phase-amplitude coupling is characteristically found in cortical EEG during epileptic seizures. Also, we are currently researching automatically diagnosing diseases and finding new features by using artificial intelligence technology on the big data of our brain waves and magnetoencephalograms.
We have developed a neural network to identify epilepsy, spinal cord injury, and healthy people from resting magnetoencephalographic signals.

http://resou.osaka-u.ac.jp/ja/research/2019/20190326_1

Details of the dissertation(Published:

Automatic diagnosis of neurological diseases using MEG signals with a deep neural network

Original code; https://github.com/yanagisawa-lab

PAGE TOP
Copyright © Osaka University Institute for Advanced Co-Creation Studies, Yanagisawa laboratory. All Rights Reserved.