The headlines read, “Paralyzed patient feels sensation again.”
My first reaction was, well that’s like saying a blind man can hear again, since “paralysis” refers to the inability to move, not feel. But I knew what they meant – someone who lost both motor control and sensation from an injury was able to regain sensation.
But that is not what happened either. The patient in the study still cannot feel; they have not regained sensation. Rather, the researchers were able to produce sensation by directly stimulating the sensory cortex of the brain. That’s like physically moving the patient’s arm and saying, “Paralyzed patient moves again.”
The misleading headlines aside, this is an exciting, if incremental, advance. That is often the trick of good science communication – conveying the genuine excitement about an important incremental advance without overhyping it or sowing confusion. Science usually advances by baby steps, while headline writers want “breakthroughs.”
Brain-machine interface
So what is the genuinely interesting science here? This is part of research into brain-machine interface (BMI), the quest to control computers and robotic parts with the brain and to provide sensory information back to the brain.
The ultimate goal is what fiction writers envisioned for The Six Million Dollar Man in the 1970s – a prosthetic limb that can look, feel, and function naturally (with or without the super strength). In order to get there, however, we have to know that we can interface with the brain in such a way as to produce all the necessary phenomena.
This is an important and necessary step, and it was not a foregone conclusion that the idea of BMI could even work theoretically. The brain and body developed together, as a unit, and it was feared that adding an artificial piece to that elaborate network would simply not be able to interface effectively.
There has therefore been a lot of BMI research focusing on proof of concept – can we overcome any theoretical barriers? So far (spoiler) the research has shown that brain machine interfaces can work – the brain is plastic enough to adapt to a new interface, without any theoretical limitations.
So far BMI researchers have shown that subjects (monkeys or humans) can learn to control either a computer or a robotic limb with their brains. The brain can form the pathways necessary to produce functional control. How fine that control can be has yet to be determined, but researchers were able to train a monkey to feed itself Cheerios with a robotic arm.
Motor control has turned out to be the easiest aspect of BMI. The primary limitation is the number and size of the electrodes, and how close they are to the appropriate neurons in the brain. At this point the primary limiting factor is the technology of the electrodes.
Sensation has proved a harder nut to crack, but again all the proof-of-concept signs are positive so far. Sensation is more complex because there are different kinds of sensation, not just touch. The brain needs to feel where the limb is in three-dimensional space (proprioception), and needs feedback from the muscles to indicate that they are moving.
Further, the brain needs to process various sensory streams simultaneously to create the sensation of embodiment (that the limb is part of their body), of ownership (that the person owns the limb), and control (that the person controls the limb).
These more complex sensations are important for motor control. Without them the patient has to look at the robotic limb they want to control. With sensory feedback they don’t have to rely solely on vision, and the feedback improves their brain’s ability to control the limb.
Fortunately it is looking like all of these sensations can be produced artificially through BMI. Recently, for example, researchers were able to reproduce kinesthetic feedback of a robotic limb – the sense that the limb was moving, which added to the sense that it was under the patient’s control.
Where does this new study fit into the overall research? The researchers studied a subject with a C5 spinal cord lesion, resulting in paralysis and lack of sensation in all four extremities. They implanted an intracortical microstimulation (ICMS) array in primary somatosensory cortex (S1). They then looked at various types of stimulation (mainly adjusting frequency and amplitude of the electrical current) to see what sensations resulted.
Previously most subjects having their somatosensory cortex stimulated reported sensations of tingling or buzzing – artificial and not very useful sensations. The researchers wanted to know if they could adjust the stimulation in order to produce more natural sensations.
The subject reported:
It was quite interesting…It was a lot of pinching, squeezing, movements, things like that. Hopefully it helps somebody in the future.
The subject felt as if the limb were moving, or squeezing something, or pinching, not just a non-specific buzzing sensation. That is a pretty small advance when you think of the whole BMI research program, but it is also a very important demonstration of proof of concept. It is possible to produce functional, natural-feeling sensations just by shocking the sensory cortex.
In order for this finding to lead to something practical, however, researchers will need to explore the somatosensory cortex in detail, and create a map of how to simulate that cortex in order to produce certain sensations. This may not only have to be done generally but for each individual. That sensory algorithm would then have to be mapped to specific movements of a robotic limb.
This does seem like the hard way to do it, however. It is all top-down, using brute force to mimic sensory feedback which is painstakingly mapped between the artificial limb and the individual’s brain. It could work, but I have to wonder if there could be an easier way, a way for the mapping to be more automatic.
I suppose that is an area of possible future research.
In any case, it is good to know that we can theoretically produce natural-feeling and functional sensations through BMI. Again – so far all the proofs of concept have worked out. There does not seem to any theoretical limitation to our ability to interface machines and the human brain.
We have made many important baby-step advances in this research, and have to make many more. It’s likely that each encouraging but incremental advance will be announced through overhyped and misleading headlines.