Mind-controlled exoskeleton to help disabled people walk again

Every year thousands of people in Europe are paralysed by a spinal cord injury. Many are young adults, facing the rest of their lives confined to a wheelchair. Although no medical cure currently exists, in the future they could be able to walk again thanks to a mind-controlled robotic exoskeleton being developed by EU-funded researchers.

The system, based on innovative ‘Brain-neural-computer interface’ (BNCI) technology - combined with a light-weight exoskeleton attached to users’ legs and a virtual reality environment for training - could also find applications in the rehabilitation of stroke victims and in assisting astronauts rebuild muscle mass after prolonged periods in space.

In the United Kingdom, every eight hours someone suffers a spinal cord injury, often leading to partial or full lower-body paralysis. In the United States, more than 250.000 people are living with paralysis as a result of damage to their spinal cord, usually because of a traffic accident, fall or sporting injury. Many are under the age of 50, and with no known medical cure or way of repairing damaged spinal nerves they face the rest of their lives in a wheelchair.

But by bypassing the spinal cord entirely and routing brain signals to a robotic exoskeleton, they should be able to get back on their feet. That is the ultimate goal of researchers working in the ’Mind-controlled orthosis and VR-training environment for walk empowering’ (Mindwalker) project, a three-year initiative supported by EUR 2.75 million in funding from the European Commission.

'Mindwalker was proposed as a very ambitious project intended to investigate promising approaches to exploit brain signals for the purpose of controlling advanced orthosis, and to design and implement a prototype system demonstrating the potential of related technologies,’ explains Michel Ilzkovitz, the project coordinator at Space Applications Services in Belgium.

The team’s approach relies on an advanced BNCI system that converts electroencephalography (EEG) signals from the brain, or electromyography (EMG) signals from shoulder muscles, into electronic commands to control the exoskeleton.

The Laboratory of Neurophysiology and Movement Biomechanics at the Université Libre de Bruxelles (ULB) focused on the exploitation of EEG and EMG signals treated by an artificial neural network, while the Foundation Santa Lucia in Italy developed techniques based on EMG signals modelled by the coupling of neural and biomechanical oscillators.

One approach for controlling the exoskeleton uses so-called 'steady-state visually evoked potential’, a method that reads flickering visual stimuli produced at different frequencies to induce correlated EEG signals. Detection of these EEG signals is used to trigger commands such as 'stand’, 'walk’, 'faster’ or 'slower’.

A second approach is based on processing EMG signals generated by the user’s shoulders and exploits the natural arm-leg coordination in human walking: arm-swing patterns can be perceived in this way and converted into control signals commanding the exoskeleton’s legs.

A third approach, 'ideation’, is also based on EEG-signal processing. It uses the identification and exploitation of EEG Theta cortical signals produced by the natural mental process associated with walking. The approach was investigated by the Mindwalker team but had to be dropped due to the difficulty, and time needed, in turning the results of early experiments into a fully exploitable system.

Regardless of which method is used, the BNCI signals have to be filtered and processed before they can be used to control the exoskeleton. To achieve this, the Mindwalker researchers fed the signals into a 'Dynamic recurrent neural network’ (DRNN), a processing technique capable of learning and exploiting the dynamic character of the BNCI signals.

'This is appealing for kinematic control and allows a much more natural and fluid way of controlling an exoskeleton,’ Mr Ilzkovitz says.

The team adopted a similarly practical approach for collecting EEG signals from the user’s scalp. Most BNCI systems are either invasive, requiring electrodes to be placed directly into brain tissue, or require users to wear a 'wet’ capon their head, necessitating lengthy fitting procedures and the use of special gels to reduce the electrical resistance at the interface between the skin and the electrodes. While such systems deliver signals of very good quality and signal-to-noise ratio, they are impractical for everyday use.

The Mindwalker team therefore turned to a 'dry’ technology developed by Berlin-based eemagine Medical Imaging Solutions: a cap covered in electrodes that the user can fit themselves, and which uses innovative electronic components to amplify and optimise signals before sending them to the neural network.

'The dry EEG cap can be placed by the subject on their head by themselves in less than a minute, just like a swimming cap,’ Mr Ilzkovitz says.

Keep reading