Speeding Peripheral Nerve Machine Interfaces
The peripheral nervous system consists of everything that descends from the brainstem and disperses across the body. Hundreds of millions of nerve fibre strands, the vast majority of which are travelling to and from the outer layers, including the muscle, surface fat and skin. Here, they provide the bandwidth that allows muscle movement, and the sensory perceptions of pressure, texture, heat, cold, pain and proprioception to function, by connecting the millions of corpuscles dotted under the skin into the brain at the top of the skull. Each corpuscle is directly connected to its own individual nerve, relaying its own individual signal back.
The corpuscles, regardless of the type of data they are attuned to, function
as sub-processors, analysing and classifying the data en situ, and simply transmitting
the result back, as a data stream.
The nerves, again, with the exception of the pain pathway, are mylinated. This means that a myelin sheath coats the axon - the long thin strand that pulls back from the central cell body, and runs as a single thin thread for all those metres. It is akin to a network cable; at its core that is what it is. Electrical signals pass up and down this organic cable, as individual ions are passed slowly along its length. The myelin sheath acts as an insulator, allowing more energy to be used, creating a stronger electrical signal and allowing the ions to be transported more swiftly, without interference spreading to the electrical pathway of a neighbouring nerve, which may be only nanometers away. With the help of the sheath. The fastest neural pathways thus may reach the blinding speeds of 100 to 200 metres per second. This means there is a perceptual delay on even the fastest nerves travelling from the fingertips, of over 200 nanoseconds.
For neuroprosthetics, interfacing with the nervous system, and providing artificial sensory input to augment or replace the natural, this time-delay is a godsend. It facilitates a time-window whereby such signals may be overlaid on the natural, without perceptual time delay. Since signals propagated by radio wave or laser, or fibre optic cable, have a much greater maximum propagation speed - on the order of 200,000,000 metres per second, there is the potential for artificial signals to travel from a point up to 1,400 kilometres away, in the same time it takes for signals to travel up your arm, from you striking a key.
In order to maximise this distance to its full potential, the interface with natural nerve impulses has to be as brief as possible. Every nanosecond counts. There are two ways of regaining such time. One method uses software, the other utilises hardware and the actual interface itself.
On the software side of things, there are many efforts to decipher the coding language used by the brain to talk to and from the central nervous system, in the hope that we will one day be able to talk in exactly the same language with artificial devices, ramping up the accuracy of such signals to the same level as those of a natural origin.
Likewise, this will shave full seconds off of the current best interface times, which may take 2-5 full seconds to register with current methods. The speed gain achievable via software is both impressive and extremely necessary. However, it possesses a finite limit, as when we are using the exact same language as the brain - only a matter of time - we will still need to integrate with the peripheral nerves in the most speed-efficient manner. This is where hardware comes in.
Common sci-fi literature and films talk of a neural jack in the base of the
skull where the artificial signals are pumped into the base of the brainstem.
This approach is actually rooted in practicality. The further from the sensory
nerve endings, and closer to the brain you interface an artificial signal, the
less natural axon your signals have to pass through, and thus the faster they
can travel. Mere centimetres of difference in distance from the brain, translate
into kilometres of distance with augmented signals.
In order to understand how this is possible, perhaps a short overview of how the nerves propagate information is necessary. As has been stated above, the bulk of any nerve is the axon, a long, thin thread that may travel millimetres as easily as metres. It is connected at one end, by the main cell body of the nerve, including the nucleus, and a great many dendrites that sprout out like hairs from all angles of the cell, each no thicker than the axon itself.
At the other end the axon tends to spread out, forming a tree like structure itself, with dozens to hundreds of micro axon threads. Around the axon forms a network of mylin cells, in some cases millions of them, which function like a cable shield, but are not part of the nerve itself. The axon branches break free of this sheath, and exist to connect to the dendrites of other nerve cells.
Where an axon branch meets a dendrite, something called a synapse is formed. This synapse facilitates the transfer of signal from one nerve to the next, by means of electrochemical signals, where the dendrites of one cell use the internal cell processes to release a flow of chemical markers to be detected by the axon of the next. These chemical markers travel a good deal more slowly than the electrical pathways within axons. Thus, even though they are the easiest kind of transfer to detect, they are not ideal candidates for time-critical augmented interfaces.
An opposing approach, much more difficult to achieve, given current knowledge and surgical ability, but which would certainly produce better results, is to slice and splice the axon itself, prior to exit into the main cell body. That is to sever it completely, then rejoin the two severed ends, using an electrode-incorporating median, that could then seamlessly propagate the original signal, or replace it with the augmented/artificial signal, as a switch mechanism.
This would be accomplished by means of two electrodes per axon, one imbedded into the transmitting half, the other into the receiving half, and a switch mechanism between them, for rerouting the signal. A third, artificial path would lead out of this substrate layer, to a larger neural jack assembly likely placed behind the brainstem itself, where relatively relaxed constraints on operational size exist.
By bypassing the chemical exchange, and essentially slicing the entire spinal cord in one go, it would become possible to relatively simply use a dense layer of nanoprobe electrodes, of sufficient density that it could be 99.9999% guaranteed that each nerve would be interfaced by an electrode. Thus, it would not matter at implantation stage, assessing which electrode did what task. This could be worked out afterwards by the embedded microprocessor(s) monitoring the connection. Once sufficient data as to which nerve was carrying what, had been gathered, the results could be blasted into PROM, preventing the more dangerous potential forms of 'mind hacking'.
The interface type suggested above is of course, far beyond the state of the art capability at time of writing this paper. It is the author's hope that the directions for hardware interface expressed herein be utilised to help guide the direction of neuroprosthetics research as the continued over-emphasis on software-only methods of improving neuroprosthetic responsiveness begin to deliver decreased return on investment.
NB: It is acknowledged that at time of writing, no way yet exists to reconnect a severed spinal cord. This is a current and longstanding bottleneck in neuroprosthetic hardware advancement.