Alright, I'll write something about science

After claiming to be a scientist (at least in training), it is probably time that I offered something of a scientific bent. I may still manage to fit sport in somewhere.

The world media was aflutter with reports of a paralysed man who controlled a cursor and a robotic hand using his brain, but not his muscles. Y'see, he had a sensor in his brain which sent signals to a processor, which decoded them and moved said cursor/robotic hand.

What to make of all this? First of all, it's worth checking out the article itself:
http://www.nature.com/nature/journal/v442/n7099/pdf/nature04970.pdf

It's nice that Nature has made this freely available, as like most scientific journals they have a way of keeping things to themselves and their subscribers. When faced with a journal's website telling me i'm not allowed to view an article, I go here: http://www.library.usyd.edu.au/ejournals/
and access is usually forthcoming.

Nature Publishing Group is also one of the most respected/highest impact/popular series of journals in scientific publishing. If you're a scientist with a paper that you think is white hot, you might try sending it to Nature.

Returning to our cybernetic tetraplegic, MN - for some reason, subjects are always referred to by their intials. Does it really protect their privacy when there's pictures of them in the article? And how hard can it be to find a tetraplegic with those initials? They also mention that his injury was from a knife wound that transected the spinal cord between cervicl vertebrae C3-C4. C'mon, you can't give those details but not explain any more!

One of my first thoughts was: how the heck do they transduce his brain signals? To me, it all came down to the signal processor. Measuring neuronal activity is not such a big deal, but converting that into movement of a cursor seemed to me more complicated. This probably reflects my background in neuroscience and I would like to now call upon anyone with an engineering background (you know who you are) to fill in the other details.

Without knowing anything, the following sketch may not be too far from the truth. The sensor uses its 100 electrodes to detect the activity of neurons - sometimes single ones, other times groups. This activity is measured in microvolts. What they then do is build up a database of firing patterns associated with certain movements. Imagine putting your hands together then apart, get pattern X. Open and close your hands, get pattern Y. Imagine manually tracking a technician-operated (in Australia we call them research assistants) cursor, get pattern Z.

This pattern was converted via a filter function into a two dimensional output signal - the neural cursor. This makes sense for imagining moving a cursor, but I'm not sure what a 2D output would mean for opening your hands or shrugging your shoulders. I still have no idea how the filter function works, but it seems to do the trick.

They then compared MN's neural cursor with the techinician's actual cursor. Over 6 sessions the correlation between the neural cursor and the technician's was 0.56 +/- 0.18 on the x axis and 0.45 +/- .15 on the y axis. It would be weird if there were a difference between x and y axes. This didn't seem to me to be a particuarly impressive correlation, but the results were 'similar or better than those seen in intact monkeys'. Unsurprisingly it turns out that they've done this in monkeys before. 'Preclinical animal studies' are a common story in neuroscience research.

It turns out the robotic control was also by using his cursor skills. Later on they got him to open some simulated email. Why was it simulated? I have no idea.They also got him to adjust the volume, channel and power to his television. Now that's research you can take to the bank! They also made movies of MN playing Neural Pong and Neural "HeMan" game. I'd love to watch these but damn MediaPlayer can't play the mov file.

Am I alone in finding many scientific papers hard to get through? This one had a lot of techniques I was unfamiliar with but still, it'd be nice if they converted it to a laywoman's version. Most newspaper reports just don't even try.

It seems like they had a few technical difficulties with the sensors, one of which required 'clinical protocol change', which meant cutting MN open and fiddling with the device. In a second subject (no initials for him) a technical issue caused an abrupt signal loss at most electrodes, the reason for which is being evaluated. I suppose this would be like getting disconnected at a critical time. Impressive research, though more reporting of difficulties and problems would better serve the public.

I can't think of a good way to sum up this paper, so I'll end with Go the Doggies!