In theory, nerve connection is quite easy. Put two nerves of the same type right up next to each other and they will fuse and form a continuous signal transduction pathway. However, the success rate for this is somewhere on the order of 80%. Assuming there are ~100 connections to be made, 0.80^100 isn't very enticing. Perhaps using the compounds he talks about can increase the success rate, but this is still on fairly shaky grounds.

Actually, if I'm reading wikipedia correctly, there are only 62 connections that need to be made. While that doesn't help with the error rate it seems feasible to make that many connections individually in a short period of time, if that is in fact what they're doing.

The spinal nerves (by and large) originate in the spinal cord at their level of origin. That is to say that if you severed the spinal cord at some point, you could not 'split out' the connections corresponding to the lower spinal nerves from the overall bundle. Amongst other things, the spinal cord also contains neurons that form closed circuits without reaching the brain (these make up your reflexes).

So, if you transplanted the head and spinal cord, then use, you could somehow make '62 connections'. If you transplanted 'the head' you would have to join the entire spinal cords.

Finally, each nerve is really 'a cable bundle', so the process of joining them is kinda tricky, and beyond the scope of my knowledge. Needless to say, the authors of the paper think that it can work. Their references likely can point you in the right direction.

> if I'm reading wikipedia correctly, there are only 62 connections

For others, here's the Wikipedia link:

http://en.wikipedia.org/wiki/Spinal_nerves

I'm surprised that it's so few (31 pairs of spinal nerves). I would have imagined that there would be thousands and thousands of nerves.

Since we certainly have more than 62 sensory inputs and motor outputs in the trunk of our body, I assume that those 62 nerves hide quite a lot complexity.

0.8 ^ 100 simply means that it's extremely unlikely that there would be 100 out of 100 successful connections given that the probability of the success of each connection is 0.8. The expected success rate would still be around 80 out of 100.

Yes, but even if a few spinal nerves don't get connected, you'll have severe motility problems.

They explained that a few compounds including polyethylene glycol can help glue the nerves together and function (transmit electrical impulses), as they claim, at 10% efficiency. They then state that better compounds exist and that efficiency can also be improved with some other chemicals.

At least so I read it

Yeah I caught that. I'm wondering more about the alignment problem. You wouldn't, for example, just glue a severed bundle of fiber optics back together without reconnecting each strand individually.

Why not? Brains adapt very well. Monkeys given a brain implanted electrodes that control a mechanical arm learned to use it in days. I am pretty sure that with some therapy you could use all your limbs quite well if i swapped the neurons that go to them haphazardly. And it would probably take months, not years of therapy

http://en.wikipedia.org/wiki/Neuroplasticity

If you connected through a silicon interposer you could use an FPGA or something similar on the silicon to rearrange and possibly even transcode the signals as needed.

Some sort of precise robot similar to a wire-bonding machine could make bonds between nerve ends and pads on the silicon. And then maybe fill the whole area with something solid like epoxy so the bonds don't move or loosen?

In other words, an API?

More like a patch panel.