A peer review study from Frontiers in Systems Neuroscience, this post looks to break down the study into an executive summary for the busy neuroscience professional. It involves two experiments involving motor skill receptivity and how performance is impacted.
Brain-computer interfaces (BCIs) are the brain communications devices and control units that are being used and tested in labs that utilize brain activity to control an external device. Test subject control and perform certain tasks and learn to control movement or selections of the objects using devices or caps that are wired with sensors. The sensors pick up the mu and beta rhythms over the sub-regions of the sensor-motor cortex and translate the changes in those rhythms to become control commands for manipulation of devices or for interaction with computer systems. Many studies focus on the computer side at improving the performance of the interaction but very few have looked at the human side as to whether the subject using them can begin to feel ownership and thus help in building the skill acquisition for the subject. Feedback design is essential and influences the processing of motor learning when comparing a movement of a cursor to grasping a ball. Some study suggests that manipulation of feedback can enhance results by positive biasing and the use of inaccurate feedback can impede performance. This study looks to probe into the influence of feedback to improve the user’s performance and interaction with brain control interfaces — if the illusion of body ownership transfer (BOT) for the operators can have a positive effect. Would the owner of the computer-generated hands improve the ability to in subjects to abilities in the process? Two experiments were used.
This study used 40 participants (38 right-handed, 2 left-handed). While sitting in a chair with an EEG electrode cap (27 electrodes, a reference electrode mounted in the right ear and a ground on the forehead), study participants were asked to imagine squeeze motions to record brain activities that reflected the difference of right and left-hand imagery. The 40 responses then recorded with 20 trials for both each right and left-hand experience. From this assessment, a Common Spatial Pattern (CSP) could be detected for weighing the electrode importance in a discriminating task. This difference was enlarged to show variance and discrimination between the right and left hands such that negative values were transferred to the robot’s left-hand grasp and positive values were transferred to the right hand. The subjects then performed motor imagery tasks while they watched first-person images of the robot’s hands. They were also instructed to “look down to their body” as if they were watching their own hands that were sitting on their lap. In order to measure each individual’s physiological reaction to a threatening stimulus (in this case the illusion of shot), skin conductance electrodes were also placed on the palms of the hand. Prior to the experiment, participants watched an injection of a syringe into the robot’s hand to test response. After this, the testing sessions commenced in random order to test:
• Still — no response throughout the session with the robot hand
• Match — the robot hand moved in trials that classified the correct response and on cue
• Raw — the robot’s hand moved randomly, not on cue, or utilized the wrong hand.
The still was designed to be the control condition and expected to raise not ownership over the hand. In all conditions, participants were to perform trials that were identical to trials provided in a training session. Each session was 3 minutes long. At the end of the last trial, an injection was given to the robot hand to examine if ownership could cause a pain response stimulus. Two questions were asked to follow-up:
1. If it feels like it was your own hand being injected?
2. Throughout the sessions, did you interact and feel as if the hands were like your own?
The data showed that during the match condition, there was a higher/significant positive correlation found between brain-computer interface performance and the intensity of the illusion. With a test trial involving the successful grasping of a light bulb, the result was a higher mean value more so than in the other two conditions. The match condition, when moving strongly in agreement with the operator’s intentions (evolved more of reliably and precisely the mind-control of the robot’s hands) could more easily induce a transfer of ownership response. No significant difference was confirmed when compared to raw or control. There was a wide dispersion of the result values due to the performance of the subject. This was reflected in the wide ownership illusion and the illusion was augmented when the negative feedback of the subject’s miss-performance was eliminated. Thus, BOT could be affected by the subjects’ brain control interface performance and feedback design. BCI performance caused a stronger perception of the illusion in subjects the higher the BOT-motivated the subjects were.
Because of the performance enhancement of the perception of the participants, self-evaluation and the inducement of the body ownership transfer response was tested in four different feedback conditions including two that the subject’s performance was positively and negatively biased in the first half of the session. This provided a comparison of the person’s online performance in the first portion of the session AND compared the time-variant sequences or distribution of the EEG with regards to left or right-hand imagery in each portion of the session. In experiment 2, sixteen participants were selected — none from the previous experiment. Each did 4 sessions consisting of 40 imagery trials for approximately 5 minutes. During the first half of each session (20 trials), random conditions included:
• Raw — where the participant’s performance was not biased in how the robot hand grasped a ball in reference to a classification result
• Match — where performance again was not biased but the robot hand only corresponded correctly to the lighting of a bulb when the classification was met.
• Positive (Fake-P) feedback — in which the performance was positively biased and lit correctly in 90% of the trials regardless of the subjects real performance
• Negative (Fake-N) feedback — in which the performance was negatively biased and displayed the correct lighting of the ball in only 20% of the trials regardless of the subject’s real performance.
The assumption was made that the presence of negative would affect the subject’s perception regardless of the accuracy of the performance. The second half of each performance was performed as they did under the Raw condition. The goal was to seek changes in the Brain-computer interface’s performance and skills in order to see whether positive or negative bias had any effect. There was also speculation as to whether under bias feedback, test subjects would consciously or subconsciously modify the generation of brain activity patterns. Original brain signals were used to challenge the processing of the changes and clarify the data.
There were two outliers that were detected in the Fake-N condition and one in the Raw condition thus their data was discarded leaving 13 subjects. A one-way ANOVA test revealed significant results and differences between the Fake-P and Raw conditions and also between the Match and the Raw conditions. There were no realized time/performance enhancements resulting from the experiments. Results revealed a class separation between the two halves of the experiment sessions. This was a more significant separation in results in the Fake-P condition trials and indicated that the subjects could generate brain/motor patterns that were more classifiable by receiving positive feedback than in comparison to the performance in the Fake- condition. This result was also confirmed in the Match and Raw condition that seemed to show that the Match condition (when they did not receive negative feedback on their failing performance), performance did improve and subjects could produce more separable work in both right and left hands. Both results implied that positive feedback hands an enhancing effect on the body ownership transfer AND again, that the stronger the transfer was induced by positive feedback with the facilitation of the imagination of the movements. There was also the assumption that the biasing effect was more closely related to the subject’s own personality and influence on motivational behaviors — there are those who respond more favorable to positive feedback/encouragement and those subjects who respond better to negative/try harder feedback. To further enhance the assimilation of the experiment, they suspended questioning as it was an interruption into the assimilation process for the subjects. An additional study was asserted to verify whether the measurement of intensity to the enhancement of motor imagery to positive feedback could be enhanced further.
The first study showed that negative feedback on the subject’s mistakes/errors impeded ownership and the body ownership transfer. This correlation between the ownership correlated in performance and how subjects were able to control their hands. The second experiment realized the feedback could immediately boost the effect BUT with the analysis of the brain patterns, it could change the trends in how motor imagery learning is performed. They thought it was conceivable that more realistic feedback presentations can assist a user to train and manipulate a system faster and more efficiently. Personality needs to be taken into account as whether people respond to positive (good job) or negative (try harder) feedback. Subject’s motor imagery skills dramatically change in a session based on the state of their mind and this needs to be classified in further studies to see the effects. This was a fascinating study that showed how rehabilitation could be enhanced using feedback but also how the ownership of an illusion (as well as motivation surrounding it) to make a computerized body part one’s own can make progress in stepping forward. It displays an interesting dynamic on how people can more quickly perhaps learn to used computerized limbs in the future.