Scientists have discovered a way to integrate adaptive artificial intelligence (AI) machine learning with human motor learning in a non-invasive brain-computer interface (BCI) that reduces training time and improves performance. A new neuroscience milestone was achieved for non-invasive BCI by a team of researchers at Carnegie Mellon University, which published its recent study in Nature Communications.
“From a neuroscience perspective, this alignment mirrors known mechanisms of motor skill learning, in which consistent sensory feedback shapes synaptic plasticity along behaviorally relevant neural pathways,” wrote the study’s corresponding author Bin He, along with Carnegie Mellon University co-authors Hanwen Wang, Yisha Zhang, Maxim Karrenbach, and Yidan Ding.
Brain-computer interfaces are assistive technology that enable the user to control external electronic devices using their thoughts. To do so, BCIs need to be able to record brain activity and decode it to predict the user’s intentions.
BCI devices can be invasive, requiring a surgical implantation on the neural tissue of the brain. Alternatively, non-invasive BCIs, such as a wearable cap, uses sensors to measure brain activity, MEG (Magnetoencephalography), or EEG (Electroencephalography)
In general, invasive BCIs outperform non-invasive BCIs when it comes to signal accuracy. Recent advances in artificial intelligence and machine learning are helping to close this performance gap.
The way humans learn versus machines is a very different process, and one of the fundamental barriers to brain-computer interfaces. Neuroplasticity is the brain’s ability to learn, adapt, and change based on experience, such as through trial-and-error, also known as experiential learning. In contrast, AI machine learning is trained on massive datasets in order to identify patterns to make its predictions. If the datasets contain labeled data, this is called “supervised learning.” Unsupervised learning is when machine learning algorithms are trained on unlabeled data. Reinforcement learning is when AI algorithms learn and adjust future actions based on positive and negative feedback to its actions.
Up until now, the BCI training using motor imagery was largely focused on either user-centric or decoder-centric. According to the researchers at Carnegie Mellon University, they created the first joint learning framework that enables user and decoder adaptation at the same time within a combined learning loop.
“By aligning human learning and machine adaptation toward a shared control objective evoked by sensory cues, the system fosters a synergistic co-adaptive process,” the researchers wrote.
The Carnegie Mellon University research team enrolled 31 study participants of healthy, able-bodied adults, consisting of 13 males and 18 females. Each participant was outfitted with a 64-channel Neuroscan Quik-Cap with SynAmps 2/RT amplifiers by Compumedics Neuroscan to record brain activity non-invasively
The key to the joint learning framework is the creation of a closed-loop system with a decoder containing a sample-wise reweighting algorithm and user sensory guidance provided by touch stimulation. This method offers two-way adaptation to reduce noise and improve accuracy. The team’s solution resulted in an average of 77.5% accuracy for two-dimensional cursor control and 86% accuracy for one-dimensional cursor control.
“By coupling reinforcement-driven neural plasticity with adaptive algorithmic optimization, this framework advances BCI training from passive calibration to active human–machine joint learning, enabling practical and scalable neural interfaces for communication and rehabilitation,” the scientists concluded.
Artificial intelligence machine learning is rapidly transforming the brain-computer industry. This proof-of-concept demonstrates how BCI calibration and performance can be improved by a sensory-guided duo learning framework that cleverly combines human motor learning with AI can adapt. The discovery of innovative ways to improve the performance of non-invasive brain-computer interfaces gives hope to those who have lost the ability to move or speak in the future.
Brain Computer Interface Essential Reads
Breakthrough Brain-Computer Interface Is Minimally Invasive
AI Copilot Boosts Brain-Computer Interface’s Performance
