neurodudes

The July 13 issue of Nature included some neural prosthetics papers, one of which was the paper reporting 9 months of stimulation of Matthew Nagle, a tetraplegic who received the first trial of the 96-electrode BrainGate implant in his right precentral gyrus (motor cortex (MI) for arm). The authors were Leigh R. Hochberg, Mijail D. Serruya, Gerhard M. Friehs, Jon A. Mukand, Maryam Saleh, Abraham H. Caplan, Almut Branner, David Chen, Richard D. Penn and John P. Donoghue.

Details about Matthew Nagle’s experiment

In summary, the paper reported that:

• the implant works
• Nagle learned to use it to control a cursor and various other things
• it can be learned really quickly compared to some noninvasive approaches (like using eye saccades to control things)
• it is not as accurate as saccades; although he can control the direction of the cursor pretty well there are instabilities and oscillations in the cursor’s movement (see the neural pong video for an example, and also figure 5)
• unlike eye saccades and some most other noninvasive approaches, Nagle can use the implant to control the cursor while carrying on a conversation and while looking around the room normally — it doesn’t require exclusive concentration or the cessation of other movement
• the patient has “bad days” and “good days” — sometimes it’s harder for him to accurately control the cursor than others (I got the idea that the researchers had to retune the system daily, also, although I wasn’t quite sure about exactly what was involved there)
• a “notable” number of electrodes do lose signal after awhile (months), although based on impedence measurements and the timing of the losses (abrupt lossage of a number of units in a short timespan) the team seems to think this is a problem with the device, rather than the death of cells near the electrode
• the paper also briefly reports some information about the second trial subject, a 55-year old. Apparently the second subject cannot achieve as accurate control as Nagle can
• Nagle had been injured for years prior to the implant; so it’s nice to know that his motor cortex is still usable for this sort of thing

They used a linear regression algorithm to decode the neural activity. Spooky videos of neural control of various things are available here.

Also, this article from Brown’s alumni magazine was a neat read that goes a little more into the circumstances around the experiment and the reactions of some of the people involved. Sadly, that article states that regulations require Nagle to have his implant removed at the end of the trial.

Paper on alternate decoding method

Another paper in this issue was

Gopal Santhanam, Stephen I. Ryu1, Byron M. Yu, Afsheen Afshar and Krishna V. Shenoy. A high-performance brain–computer interface. Nature 442, 195-198(13 July 2006)

This paper uses a 96-electrode array in premotor cortex in monkeys to move a cursor. Rather than continuously, smoothly moving the cursor, they attempted to teleport the cursor immediately to the destination. The authors found that this method permits faster cursor movement and hence high information throughput (6.5 bits per second, about 4x previous systems). They used multivariate Gaussians or Poissons to decode the data.

Editorials

There were also two editorials. The first is Is this the bionic man?. The second is Neuroprosthetics: In search of the sixth sense by Alison Abbott which reports on ideas to use invasive BCIs to introduce sensory input into the CNS. She notes that if you were going to rehabilitate a paralyzed person, it would be important to give them artificial proprioception.

That commentary also included a great quote from Shenoy stating an observation about the brain’s adaptability that I personally think is very important to keep in mind: “When the concept of stimulating the auditory nerve emerged in the 1970s, people said it would be impossible to generate the right electrical signal to the brain. But it turned out that you don’t have to get it perfect, just close enough for the brain to do its own fine-tuning.”

Intention decoding algorithms

I started a (very incomplete) list of available algorithms for decoding movement intentions from neural data. Please add to it if you know of more. It’s on NeuroWiki, at NeuroWiki:MovementIntentionDecodingAlgorithms.

This entry was posted on Friday, September 1st, 2006 at 8:59 pm by Bayle Shanks and is filed under Neural prosthetics. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.