2004 Conference Proceedings

Go to previous article 
Go to next article 
Return to 2004 Table of Contents 


CYBERLINK HANDS-FREE COMPUTER CONTROLLER: TWO NIH STUDIES - RESULTS AND MUSINGS

Presenters
Andrew Junker
Brain Actuated Technologies, Inc.
1350 President Street
Yellow Springs, Ohio 45387
Phone: 937-767-2674
Fax: 937-767-7366
Email: admin@brainfingers.com

Complete Paper: The feasibility of using the Cyberlink as a computer access solution for individuals with disabilities was investigated. Specific questions evaluated were: (1) Can the Cyberlink be used as a computer access solution for individuals with no other documented means of computer access? (2) Can Cyberlink control be made more accurate and reliable through the use of machine learning and adaptive control techniques? Clinical trials were conducted with twenty-five participants of varying ages and disabilities. Each participant had been unable to access technology due to physical limitations. Trials consisted of two tasks designed to measure response times and the ability to move a cursor to a target using the Cyberlink. Each task was completed by the participant a maximum of three times. History and performance data were collected and analyzed using Cyberlink data collection software developed for these tasks and now distributed with Cyberlink systems. Data was also obtained using questionnaires, videotapes and observation. All twenty five participants were able to exhibit some form of conscious computer control with the Cyberlink.

To provide a large range of data, it was decided that twenty-five participants who had no prior means of computer access would be studied. Participants were recruited from the Schiefelbusch Clinic, University of Kansas, in Lawrence KS. Providence Center in Portland, OR, Success for Kids in Loma Linda, CA, and the MDA/ALS Center of Hope in Philadelphia, PA. To obtain the twenty-five participants, the TAs used Cyberlinks to evaluate potential participants for the clinical trial. A total of 32 individuals were evaluated from which 25 participants were selected. Seven of the 32 potential participants were unable to exhibit consistent control of the Cyberlink.

Participants included female and male children and adults with CP, ALS, and brain injuries (BI). Participants were grouped into three populations: those diagnosed ALS, those diagnosed CP and the remaining participants with disabilities which included brain injuries. Participant grouping was clearly possible between ALS participants and others, CP and BI grouping was somewhat arbitrary.

Approximately 78% of the individuals tested, 25 out of 32, were able to achieve some form of access with the Cyberlink. These results are quite remarkable considering the profoundness of the disabilities of many of the participants. Three of the ALS participants are totally paralyzed except for a small amount of remaining jaw muscle control. Each of these participants started to use the Cyberlink as their means of computer access and communication when nothing else would work for them. They use the Cyberlink in their daily lives as their only means of computer access. Cyberlink data collection software was modified so they could start and run their own trials for themselves, which they did. These three participants exhibited Click reaction times that were comparable to reaction times obtained by able-bodied users.

Prior to working with the Cyberlink one of the BI participants had the diagnosis of "Persistive Vegetative State" (PVS). As a result of working with the Cyberlink the participant's PVS diagnosis was removed. The staff at Success for Kids, where the participant resides, reported they now have ways to interact and communicate with their brain injured kids. Before the Cyberlink, their options were to move the kids, read stories to the kids, or sit and watch TV with them. With the Cyberlink they have a communication link.

All but two of the participants controlled the Click and Acquire test programs with their muscle signals. One CP participant used a Beta brainwave for control. This participant exhibited no measurable facial muscle activity with the Cyberlink. One BI participant used an Alpha brainwave for control. This participant exhibited almost constant uncontrolled facial muscle activity. This activity was so strong that the participants muscle signal was usually at a maximum value. From the perspective of trying to understand the functionality and controllability of the various signals obtained with the Cyberlink it is noteworthy to mention that one participant exhibited a response that was at one extreme of the motor control functional level of no muscle activity and the other participant was at the other extreme of too much uncontrolled muscle activity. In the case of the CP participant they transitioned their brainwaves from alpha to beta causing a Cyberlink trigger event with no change in facial muscle activity. The other participant lowered their facial muscle signal and transitioned their brainwave signal from beta to alpha to cause a Cyberlink trigger event.

It is generally agreed that the brain generates more alpha resonance when a person relaxes. Further, when a person shifts into a more mentally focused state, alpha resonance will decrease. There may be an interaction between the alpha, beta and muscle frequency bands that can be used to facilitate training and control with the Cyberlink. From study results we would hypothesize that appropriate feedback and training of a user's dynamic alpha/beta relationship would be useful as feedback to enhance their Cyberlink control. The more informative and timely the information being fed back to the user in a closed-loop control situation the better chance the user will have to respond appropriately.

All the participants tested had problems with neural-muscular control. The ALS group performed the best and with the most consistency. It is hypothesized that the better performance is due to the higher cognitive level of the ALS participants and that the day to day variations in neural muscular control due to ALS degeneration was not as large as was experienced by the CP and BI participants. There was a great deal of within group participant differences in the BI and CP groups. It supports the notion that each participant would benefit from a predictive and adaptive control system that could compensate for these differences.

Variability in the data was affected by the disabilities of the participants and other contributing factors. Three of the participants were below the age of 4 years. One participant's vision was affected by their disability. For these participants the Click and Acquire tests were probably not the most appropriate tasks. Most of the participants used some form of medication because of their disability. The medication probably affected the participants' ability to use the Cyberlink. This was confirmed by the TAs and some of the ALS participants. The cognitive ability of the BI group was unknown or unclear for many of the participants. BI participants exhibited control comparable to a poorly trained able-bodied person one time and then exhibited no control or control with weak signals the next. Due to the expected user variability an adaptive controller will be invaluable.

It was insightful to observe the videos of the participants with CP as they controlled the Cyberlink. Spastic activity was exhibited for many. Their head and arms moved in what appeared to be uncontrolled motion. To facilitate the participants' ability to use the Cyberlink adjustments had to be made to baseline and sensitivity. This resulted in making the Cyberlink less sensitive to the participant's uncontrolled muscle signals or "background noise". Unfortunately it also made the Cyberlink less sensitive to the participants' consciously generated control inputs. CP participants achieved control but they may have had to expend excessive energy to bring their conscious signals above their self generated noise, which was or may have been quite fatiguing. We would hypothesize that if these participants were made aware of their self-generated noise in an informative way they would reduce some of the noise. Feeding back a measure of their alpha/beta-muscle ratios in a meaningful signal format would provide them with an informed ability to reduce their noise. For this to be successful it will probably be necessary for them to be engaged in a Cyberlink control task in which the feedback is intimately coupled to the task being performed. Feedback in this context would be more engaging and would have more relevance. When the Cyberlink is used as a switch the user has to move from a relaxed state to an active state to create a click. For someone with CP the lower the signal and background noise is during the relaxed state the less of a conscious signal they will have to create to transition to an active state. Thus if they can be made aware of the level of their relaxed state they may be able to shift down or reduce some of the "uncontrolled" background noise. Each time they were successful at doing this and experienced the reduced effort needed to transition to an active state would be effective feedback as well. We envision the paradigm where the user is engaged in control of a Cyberlink with an adaptive controller that tracks the user's changes and makes the user aware of the changes. In this case the resulting control system of user and Cyberlink would have a good chance of improving over time. The proposed predictive and adaptive Cyberlink controller will be able to provide feedback to create learning paradigms of this kind.


Go to previous article 
Go to next article 
Return to 2004 Table of Contents 
Return to Table of Proceedings


Reprinted with author(s) permission. Author(s) retain copyright.