We are conducting the following research with the aim of elucidating neural information expression, developing technologies for decoding and controlling neural information, and applying them to medical treatment.
1. Brain-Machine Interface
By applying machine learning (artificial intelligence) to big data on physiological signals such as electroencephalograms, it is possible to elucidate neural information related to human intentions, visual perception, thought states, etc. (Nat. Comm. 2024)
We have developed a technology to decode brain information from intracranial electroencephalograms measured from electrodes placed inside the human skull, and have developed a Brain-Machine Interface (BMI) that uses this to control robots and display imagined images on a screen (Ann. Neurol., 2012, Comm. Biol., 2022, Fig. A, B). This technology will be applied in medical applications to restore motor function and assist communication in patients who cannot move their bodies.
We are also developing a BMI that uses intravascular EEG as a minimally invasive method of obtaining intracranial EEG.
Figure A. Robot control using intracranial EEG BMI
Figure B. BMI that visualizes recalled content
2.Neurofeedback
By using BMI to connect a robot or avatar to brain activity instead of the body, we can induce plastic changes in brain activity and clarify the relationship between brain activity and function related to motor control.
We have also shown that for phantom limb pain, a condition in which the missing arm is painful after the loss of an arm, pain can be alleviated by training the patient to move a BMI robot (Nat. Comm. 2016, Neurology 2020, Figure C). This is called neurofeedback (NF) therapy. We will develop disease treatments using NF by decoding brain information and using that information.
Figure C. Treatment of phantom limb pain using magnetoencephalography BMI
3. Decoding neural information from EEG
Applying AI to big EEG data can enable automated diagnosis of conditions such as dementia and epilepsy (Watanabe et al., Neural Networks. 2024).
We can mathematically elucidate the type of brain information expressed in EEG and obtain new EEG characteristics (Fukuma et al., Communications. Biology. 2023).
From large volumes of intracranial EEG data, we will identify EEGs that characterize human thought (Iwata et al., Nature Communications, 2024).
We will use AI and new mathematical methods to elucidate neural information representation.
Multicenter cohort of electroencephalograms and magnetoencephalograms
Healthy people Approximately 300 people
Dementia/MCI EEG: Approximately 600 people
(Joint research with the Department of Psychiatry and Neurosurgery, The University of Osaka, Kochi University, Nissei Hospital, and Kansai Medical University)
Epilepsy Intracranial EEG: approx. 100 people
Electroencephalogram: approx. 1,300 people
Magnetoencephalography: approx. 300 people
(Joint research with the Department of Neurosurgery and Pediatrics at The University of Osaka, Osaka Prefecture Maternal and Child Center, and Suita Municipal Hospital)
Aoe et al., Scientific Reports, 2019
4. Neuroinformatics
We aim to use biosignals measured daily using wearable sensors to manage and promote health.
Neuroinformation includes a wide range of physiological signals that can be used to infer a person’s condition, such as electroencephalograms, electromyograms, and motion analysis data, as well as brainwaves. We aim to convert a wide range of biosignals, from invasive intracranial EEG to wearable sensors, into big data as neural information, and by decoding and controlling the information, we aim to achieve a variety of medical applications, from health promotion to neural function reconstruction. To this end, we aim to develop human resources who can conduct research that combines different fields, not only medicine, but also information, engineering, and neuroscience.