Disabled walker assistance

AI goes out on a limb for rehabilitation

Image credit: EPFL CHUV

When a body’s neural-muscle network systems are disrupted, electrical stimulation can re-establish communication. Artificial intelligence could provide further insight into how the human body works.

The simple action of picking up a ball and throwing it involves complex communication between the brain and motor neurons inside the spinal cord to the muscles in the arm. What happens when the nerves that transfer these signals are damaged, causing a roadblock in the path of the message?

Researchers at the Kentucky Spinal Cord Injury Research Center at the University of Louisville, USA, used a continuous electrical current at varying frequencies and intensities and locomotor training to restore brain-to-spine connectivity in some spinal injury patients. They were able to stand, regain trunk mobility, and walk a few steps without assistance when using the epidural stimulator.

In Europe, Stimulation Movement Overground (STIMO) is a clinical study by a team of scientists from the École Polytechnique Fédérale de Lausanne (EPFL) and the Lausanne University Hospital (CHUV), headed by Professor Grégoire Courtine and Professor Jocelyne Bloch.

Back in November 2018, the EPFL team observed pronounced growth of new nerve connections in areas that were targeted with electrical stimulation. Instead of applying continuous electrical stimulation, they targeted electrical pulses at specific locations to physiologically activate the spinal cord.

“After several months of training with electrical stimulation, our three participants were able to activate their previously paralysed muscles without electrical stimulation and they could even take a few steps, hands-free, without any support,” says Courtine.

An array of 16 electrodes, mapped to specific roots of the spinal cord for specific leg muscles, was surgically implanted. The electrode array is connected to an implantable pulse generator that is used for deep brain stimulation in people with Parkinson’s disease, but which has been adapted for this study to include real-time triggering capabilities, which can be used with a voice-controlled watch worn by the patient.

Based on the intended movement, detected by sensors on the patient’s feet, electrical stimulation bursts are delivered over the regions of the spinal cord that produce the movement. By thinking about activating the leg muscles, the residual connections in the brain activate muscles while electrical pulses activate the neural circuits associated with these muscles. The research shows that synchronised thought and excitation of the targeted circuits trigger the growth of new connections in the brain and spinal cord.

To train the muscles, the participant is suspended in a body-weighted harness that re-establishes the gravity-dependent gait muscle interaction when walking.

All three participants recovered voluntary control of leg muscles that had been paralysed; an effect which persisted beyond training sessions and even when the electrical stimulation was turned off.

Co-author Karen Minassian says: “Stimulation alone is not strong enough... the participant needs to actively engage all the time, and learn to recognise how voluntary contribution amplifies the input from the targeted electrical stimulation.”

“The next step is to start earlier, just after the injury, when the potential for recovery is much greater,” says Bloch.


Artificial neurons

Researchers at the University of Bath have created artificial neurons that can act like biological ones.

Artificial neurons can respond to electrical stimuli where the brain-nervous system communication has broken down, for example in spinal injury patients, where they can be used to repair the biological circuitry and respond to biological signals.

The team, led by Professor Alain Nogaret, a researcher at the University of Bath, modelled and derived equations to explain how neurons respond to electrical stimuli to recreate the neural responses. These calculations were used to model biological ion channels. The resulting silicon neurons mimicked biological hippocampal and respiratory neurons in response to a range of stimuli.

Nogaret says: “Until now neurons have been like black boxes, but we have managed to open the black box and peer inside,” and allow the researchers “to reproduce the electrical properties of real neurons in minute detail”.

The synthetic neurons require just 140nW, or one-billionth the power required by a microprocessor. This will enable them to be used in bio-electronic implant devices to treat chronic diseases where they can respond in real time to demands, in the same way a healthy body’s neurons do.

Nogaret summarises the breakthrough’s potential: “We can very accurately estimate the precise parameters that control any neuron’s behaviour with high certainty. We have created physical models of the hardware and demonstrated its ability to successfully mimic the behaviour of real living neurons”.

In 2018, researchers at the Center for Extreme Bionics at the Massachusetts Institute of Technology (MIT) Media Lab created a communication system to mimic muscle-tendon communication and to make disconnected biological pathways active again for use with prosthetic limbs. The two opposing muscle tendons are surgically connected so that one muscle contracts and shortens when activated and the other stretches to perform the movement.

The agonist-antagonist myoneural interface (AMI) sends movement signals from the central nervous system to the robotic prosthesis and relays data about the position, speed and torque of the joint via biological sensors, which translate the mechanical stretch into electrical signals that can be interpreted by the brain as sensations of position, speed and force.

At the Brigham and Women’s Faulkner Hospital in Boston, Massachusetts, a patient undergoing a below-the-knee amputation had two AMIs surgically implanted, one to control the prosthetic ankle joint and one to control the prosthetic subtalar joint below the ankle. A prosthetic limb was electrically linked to the peripheral nervous system using electrodes placed over each muscle.

When compared with other below-the-knee amputees, the AMI recipient had more control over the movement of the prosthetic limb and showed natural, reflexive behaviours, such as extending the toes.

Circuits around a spinal lesion often remain active. The Intelligent Spine Interface project, a collaboration between Brown University in Rhode Island and Intel, uses artificial neural networks to learn how to stimulate and communicate motor commands. Signals travelling down the spinal cord will be recorded and used to drive electrical spinal stimulation below the injury. Information coming up the cord will be used to drive stimulation to restore voluntary movements, feeling and sensation.

Surgeons at Rhode Island Hospital will implant an adapted version of Micro-Leads’ spinal cord stimulator, the HD64 electrode array, on both sides of a patient’s injury for communication that bypasses any damage.

The array will have 32 electrodes to double the surface area of the spinal cord that can be treated, compared with other devices, says Bryan McLaughlin, president of Micro-Leads. “We’re covering more surface areas, so there’s a greater probability that you are going to find the right nerve fibres, we’re reaching new fibres,” he explains.

Rather than use a pulse generator with extra wires, the array sends therapy pulses along what McLaughlin describes as a smart lead, which directs it to a particular electrode. Wires running out from the skin of the patient will be attached to Intel’s AI Accelerator hardware. A smart lead above and one below the injury site will record when the patient is attempting to move to help researchers understand brain signals to instigate a movement or residual neural signals received from sensory inputs from limbs. This data could be used to know when and how to stimulate the sensory circuitry.  

In the initial two-year phase of the project, an external computer will decode spinal signals but there are plans to create a fully implantable device for rehabilitation.

The project is supported by a $6.3m (£4.9m) grant from the US Defense Advanced Research Projects Agency.


How the brain communicates to the body

Voluntary movements are controlled by the motor cortex in the brain. This part of the brain is located behind the frontal lobe. It contains the primary motor cortex which receives information from other areas of the brain, such as the parietal lobe (position of the body), the frontal lobe (the intention and strategy to achieve the intention) and the temporal lobe (memory).

The co-ordination of the signals from the motor neurons to muscle fibres is controlled in the brain’s cerebral cortex, the outer layer of brain. This layer connects to the spinal cord and sends specific commands via nerve tracts to regulate motor control.

Each neuron connects to a single muscle. The motor neuron fires an impulse which travels along an axon, a thin extension of the nerve cell to the muscle.` When it reaches the end of the axon, a chemical is released that causes the muscle fibres to elongate and lock in position. When the nerve impulse stops, the muscle fibres return to their original positions.

The sequence of movements is regulated by the cerebellum. This small part of the brain houses more than half of the brain’s neurons (around eight to 10 billion). The cerebellum receives information from the sensory and motor cortexes and sends information to the motor cortex about direction, force and duration of the movement.


Sign up to the E&T News e-mail to get great stories like this delivered to your inbox every day.

Recent articles