Summary: New brain-machine interface technology allows immobile people to control their wheelchairs through mind control. The BMI allows users to traverse natural and cluttered environments after training.
Source: Cell Press
A mind-controlled wheelchair can help a paralyzed person gain new mobility by translating users’ thoughts into mechanical commands.
November 18 in the newspaper iScienceresearchers demonstrate that quadriplegic users can use mind-controlled wheelchairs in a natural, cluttered environment after prolonged training.
“We show that mutual learning of the user and the brain-machine interface algorithm is important for users to successfully use such wheelchairs,” says José del R. Millán, corresponding author of the study at the University of Texas at Austin. “Our research highlights a potential avenue for improved clinical translation of non-invasive brain-machine interface technology.”
Millán and his colleagues recruited three people with quadriplegia for the longitudinal study. Each of the participants followed training sessions three times a week for 2 to 5 months.
Participants wore a skullcap that detected their brain activities through electroencephalography (EEG), which would be converted into mechanical commands for wheelchairs via a brain-machine interface device.
Participants were asked to control the direction of the wheelchair by thinking about the movement of their body parts. Specifically, they had to remember to move both hands to turn left and both feet to turn right.
In the first training session, three participants had similar levels of accuracy (when device responses matched users’ thoughts) of around 43% to 55%. During the training, the brain-machine interface device team saw a significant improvement in the accuracy of Participant 1, which achieved over 95% accuracy by the end of their training.
The team also observed an increase in Participant 3’s accuracy to 98% halfway through their training before the team updated their device with a new algorithm.
The improvement seen in participants 1 and 3 correlates with the improvement in feature discrimination, which is the ability of the algorithm to discriminate the pattern of brain activity encoded for “go left” thoughts from that for ” turn right “.
The team found that the best feature discrimination is not only the result of the device’s machine learning, but also of the learning in the participants’ brains. The EEG of participants 1 and 3 showed clear changes in brain wave patterns as they improved the precision of mental control of the device.
“We see from the EEG results that the subject has consolidated an ability to modulate different parts of their brain to generate a pattern to ‘go left’ and a different pattern to ‘go right,'” says Millán. . “We believe there is a cortical reorganization that occurred as a result of the participants’ learning process.”
Compared to participants 1 and 3, participant 2 showed no significant changes in brain activity patterns throughout the training. His accuracy increased only slightly during the first sessions, which remained stable for the rest of the training period. This suggests that machine learning alone is insufficient to successfully maneuver such a mind-controlled device, Millán says.
At the end of the training, all participants were asked to drive their wheelchairs through a crowded hospital room. They had to work around obstacles such as a room divider and hospital beds, which are set up to simulate the real environment. Participants 1 and 3 completed the task while participant 2 did not.
“It appears that for someone to gain good control of the brain-machine interface that allows them to perform relatively complex daily activities like driving the wheelchair in a natural environment, it requires neuroplastic reorganization in our cortex” , explains Millán.
The study also emphasized the role of long-term user training. Although Participant 1 performed exceptionally well at the end, he also struggled in the first practice sessions, says Millán. The longitudinal study is one of the first to assess the clinical translation of noninvasive brain-machine interface technology in people with quadriplegia.
Next, the team wants to understand why Participant 2 did not experience the learning effect. They hope to conduct a more detailed analysis of the brain signals of all participants to understand their differences and possible interventions for people struggling with the learning process in the future.
About this neurotechnology research news
Author: Press office
Source: Cell press
Contact: Press Office – Cell Press
Image: Image is in public domain
Original research: Free access.
“Learning to control a wheelchair with BMI for people with severe tetraplegia” by José del R. Millán et al. iScience
Learning to control an IMC-driven wheelchair for people with severe tetraplegia
- Three participants learned to drive a non-invasive BMI-powered wheelchair
- Direct transfer of acquired BMI skills to wheelchair control
- Subject learning and robotic intelligence are key to IMC-powered translational robots
Mind-controlled wheelchairs are an intriguing assisted mobility solution applicable in cases of complete paralysis. Despite advances in brain-machine interface (BMI) technology, its translation remains elusive.
The main objective of this study is to probe the hypothesis that the acquisition of IMC skills by end-users is fundamental to controlling a non-invasive brain-powered smart wheelchair in real-world settings.
We demonstrate that three quadriplegic spinal cord injury users could be trained to operate a noninvasive, self-paced, thought-controlled wheelchair and perform complex navigational tasks. However, only the two users with increasing decoding performance and feature discrimination, significant changes in neuroplasticity, and improved IMC command latency, achieved high navigation performance.
Moreover, we show that dexterous and continuous control of robots is possible through low degree of freedom, discrete and uncertain control channels such as a motor imagery IMC, by mixing human and artificial intelligence through shared control methodologies.
We posit that subject learning and shared control are the key elements paving the way for noninvasive translational IMC.