Can a computer read finger movements just from the mind?

Imagine being able to control a computer just with your mind, without moving a muscle. This could become a reality thanks to new technology that decodes finger movements using magnetoencephalography (MEG). A team of researchers has developed a convolutional neural network (CNN) capable of distinguishing which fingers of the hand are moving, even if they are on the same hand! During an experiment, participants pressed buttons with the index and middle fingers of both hands while the MEG recorded brain activity. The CNN achieved 95% accuracy in distinguishing the hand used and 80-85% accuracy in recognizing individual fingers. This is a huge step forward for non-invasive brain-computer interfaces, which could revolutionize communication and rehabilitation. The CNN used is also fast and easy to interpret, making it perfect for future studies in neurophysiology.

The convolutional neural network used in this study, called LF-CNN (Linear Finite Impulse Response Convolutional Neural Network), was compared with other deep learning architectures such as EEGNet, FBCSP-ShallowNet, and VGG19. The LF-CNN stood out for its computational efficiency and ability to provide interpretable results in both spatial and spectral domains. This means it not only decodes finger movements with high accuracy but also helps better understand the neurophysiological patterns associated with motor cortex activity during cognitive tasks. MEG, used to record brain activity, is a non-invasive technique that measures the magnetic fields generated by neural activity. This study opens new possibilities for the development of more advanced brain-computer interfaces and for understanding the mechanisms underlying motor learning.