Artificial intelligence (AI) Machine learning is transforming assistive technologies that help paralyzed people regain movement. New research published in the Journal of the American Physical Society APL bioengineering We demonstrate how AI has the potential to restore lower extremity function in patients with severe spinal cord injury (SCI) by identifying patterns in brain signals captured non-invasively by electroencephalography (EEG).
“These findings establish a baseline for EEG decoding of lower limb movement attempts in severely paralyzed patients and pave the way for the development of brain-controlled neuroprosthetic systems,” wrote corresponding author Silvestro Micera, along with co-authors Laura Toni, Valeria De Seta, Luigi Albano, Daniele Emedoli, Aiden Xu, Vincent Mendez, Filippo Agnesi, Sandro Iannaccone and Pietro. Mortini and Simone Romani.
Using technology to improve the quality of daily life for people with paralysis is a worthy cause. According to the World Health Organization, an estimated 15 million people worldwide have spinal cord injuries. In the United States, more than 308,600 Americans live with traumatic spinal cord injuries, and men account for 78% of new cases since 2015, according to the National Spinal Cord Injury Statistics Center. According to the Christopher and Dana Reeve Foundation, the lifetime cost for a 25-year-old with severe quadriplegia (also known as severe quadriplegia) can exceed $4.7 million.
The study’s researchers are from the Federal Institute of Technology of Lausanne (EPFL) in Lausanne, Switzerland, the University Hospital of Lausanne, Switzerland, the Scuola Superiore Sant’Anna in Pisa, Italy, and the IRCCS Ospedale San Raffaele in Milan, Italy.
Four male patients with spinal cord injuries between the ages of 18 and 33 participated in this proof-of-concept study. Each participant performed four experimental sessions consisting of multiple trials separated by several weeks. While seated in a wheelchair equipped with a 64-channel EEG cap by ANT-Neuro, each participant was asked to visualize a movement without actually attempting the movement, and to later attempt to perform the movement. There are four types of movements: one that bends the left and right hip joints, and one that stretches the left and right knees.
After static data analysis of brain activity recordings, we used the AI machine learning classifier XGBoost (eXtreme Gradient Boosting) to analyze all sessions and Monte Carlo repeated sampling for each participant.
“Our results suggest that EEG signals are often able to distinguish between lower limb movement attempts and rest attempts, but decoding left and right, hip and knee movements is more elusive,” the researchers reported.
With this proof of concept, scientists demonstrated that AI can decipher brain activity non-invasively recorded with an EEG cap and distinguish between a severely paralyzed patient’s resting state and attempted leg movement. This favorable feasibility evaluation serves as a starting point for future new non-invasive neuroprosthetics that allow paralyzed people to move again through thought alone.
Copyright © 2026 Kami Rosso All rights reserved.
