International Assessment of Brain-Computer Interface Research
A low-resolution version of the final report of this study is available for download in Adobe Acrobat (.pdf) format [~6 MB].
Advances in computational technology, component miniaturization, biocompatibility of materials, and sensor technology will lead to improved feasibility of useful brain-computer interfaces in the next five years. Since the 1970s there has been increasing interest among agencies of the Federal government, such as NSF, DARPA, ONR, AFOSR, U.S. Army, NIST, and NIH, state agencies, universities, and private industry in improving human-computer interaction and developing a BCI system.
Judging from scientific papers published in technical journals and at conferences, BCI has seen increasing interest since 2000, when the First International Meeting on Brain-Computer Interface Technology was held and reported in the IEEE Transactions on Rehabilitation Engineering. The literature of the field has doubled since 2002.
Experiments with animals and demonstrations by a few human quadriplegics have shown that useful neural signals from the brain can be sensed, interpreted and used to drive a computer or simple prosthetic device. The Defense Advanced Research Projects Agency has an initiative to fully develop a brain signal interfaced prosthetic arm within five years. Similar commercially developed prosthetics are expected to follow rapidly.
Hardware, software and devices for BCI research are available and being developed at universities and are being spun-off intcommercial enterprises, for example, the University of Utah Array, the University of Michigan Array and biomimetic VSLI chips at the University of Southern California. These technologies are now enabling rapid advancement of neural signal recording and interpretation for prosthetic devices.
Neuron firing in the brain may be detected through electrodes normally inserted in the cortex, singly or in multiple electrode arrays, or through electrodes placed non-invasively in contact with the scalp using electroencephalographic methods (EEG). Magnetoencephalographic activity (MEG), thermography, functional MRI interpretation and analysis of near infrared spectrum (NIRS) activity are being considered as auxiliary sensing methods.
Need for an International Study
Significant activity in BCI research is evident overseas and in Canada. Japanese research in the use of near infrared spectrum (NIRS) sensing and interpretation may be leading the world. The University of Tuebingen, Germany, Lund University, Sweden, Fraunhofer Institute, Berlin, Korean Research Institute, Brain Science Institute, RIKEN, Japan, and Swiss Federal Institute of Technology, Lausanne, among others, have ongoing research programs in BCI. China and Brazil are emerging with active research programs in BCI. The Defense Evaluation Research Agency of the U.K. also has programs in BCI research.
Understanding the status and trends in BCI research abroad will inform program managers in U.S. research agencies and the researchers in the field to enable more effective scientific exchanges, direct more focused research in promising areas, and produce international collaboration.
Purpose and Scope
The goal of this study is to gather information on the worldwide status and trends in brain-computer interface research and to disseminate it to government decision makers and the research community.
The study panelists will gather information on BCI research abroad, which will be useful to the U.S. Government in its own programs. The study will critically analyze and compare the research in the United States with that being pursued in Japan, Europe, or other selected countries. This information will serve the following purposes:
- Identify good ideas overseas worth exploring in U.S. R&D programs
- Clarify research opportunities and needs for promoting progress in the field generally
- Identify specific opportunities (persons and institutions) for international collaboration
- Evaluate the position of foreign research programs relative to those in the U.S.
The study will review the status and trends of research and development with respect to BCI that are important for achieving successful implementation of BCI systems. The study will emphasize the neural engineering and systems engineering aspects of BCI, including computational algorithms and control methods, to effect synthetic human motor movement through neuroprosthetic systems, i.e. manipulation of a prosthesis or tele-operated device, in response to planned motor movement (PMM) activity in the applicable area of the cortex.
The sponsors of the BCI study in consultation with the chair person will specify the scope of the study. The discussion below is meant to aid in the determination of the desired priorities of the assessment of BCI abroad.
In a recent informal discussion at a BCI workshop at MITRE Corporation, Tysons Corner, VA, leading researchers suggested that the following areas were important to advancing achievements in the field.
- Improved implantable components: biocompatibility of electrodes for long term use; arrays conforming the gyri and folds of the cortex; finer resolution of neural activity
- Mathematical modeling methods for large amounts of data produced by arrays of 100 electrodes or more
- Sensor development and new approaches to delivery to cortical sites of interest, perhaps through nanodevices
- Improved probability modeling for neural spike train signals
- Solving dexterous manipulation with feedback and control systems
- Providing actuator technology, component miniaturization, and energy storage, light enough for human wearability of the systems
- Better understanding of electro-physical and systems activities of the cortex
- Creation of standard data sets for evaluating various newly developed algorithms and modeling assumptions
Among the significant elements of research are:
- Signal detection of PMM, which may be through direct probing of the cortex, or remotely through associated e-m or other signals at the scalp or e-m signals along other neural pathways which convey PMM
- Signal Processing and Control
- Noise filtering
- Recognition of PMM signals
- Production of effector command signals
- Control and tuning of effector commands through appropriate sensing mechanisms for the activity
- Measuring effectiveness of processing and control through use of standard data sets and other methods
- Efficiency and effectiveness of processing algorithms
- Bioengineering of multielectrode sensing arrays
- May be computer controlled prosthesis or other devices
- May have embedded intelligence with appropriate distributed control system
- May include successful animal research with transfer potential to humans
- Will include both invasive and non-invasive signal sensing systems
- May include work on wireless sensor transmitters embedded in the cortex or
- Other neural pathways (Since cochlear implants and synthetic vision implants are extensive research areas in their own right, this study will NOT assess those fields.)
The following lists issues that may be of interest in assessing the field.
- Theories of cognitive and neural operations for motor control from a bioengineering perspective
- Cortical plasticity impacted by BCI and impacting on BCI
- Neural signals dependence (or independence) of normal neuromuscular control channels
- Sensors for neural activity
- Research on capacities and limitations of non-muscular communication channels
- Analytical methods, algorithms and systems for neural sensor input
- Computer software, hardware, and useful peripherals
- Design of the computer interface for direct neural input
- Implant retrieval/patient follow-up
- Feedback control; control algorithms
- Sensing and measuring techniques
- Limitations/advances of cognitive science in this field
- Special requirements for different injuries and impairments of the users
- Training methods for the user and the artificial intelligence of a computer interface
- Subject training (operand conditioning) vs. machine training
- Invasive vs. non-invasive measurements of neural signals
- Evoked potentials vs. spontaneous central nervous system (CNS) responses
- Signal processing and machine learning techniques for BCI
- Successful applications of BCI
- Implant retrieval and follow up with patients
- Metrics for measuring and sensing
- Tactile sensing for feedback, especially in artificial limbs
Since the result of the assessment of BCI abroad will inform U.S. government research support and policies, topics related to policy, research direction, new education programs, and technology transfer may be of interest:
- Higher education curriculum advances to facilitate BCI
- Government programs and policies with respect to BCI
- Technology transfer programs in support of BCI
- Cultural, ethical, and political considerations for use of brain activity information and prosthetic devices
- What are the regulatory issues? How are they handled abroad?
- What are some of the obstacles and limitations to further progress in this area?
- What are the suggested future areas of study? What are the opportunities?
Greg A. Gerhardt
Jose C. Principe
Dawn M. Taylor
Patrick A. Tresco