BMI: Basic Introductions
It has become apparent of late, that whilst research into Brain Machine Interfaces continues at a steady pace, there is a certain amount of confusion in both academic and commercial circles about just what comprises a brain-machine interface.
This is thus a short disambiguation article, intended to make use of video files and imagery to explain to the lay person about the different types of BMI.
Implanted Computer Interface
The original concept, and usually first thought of when brain-machine interfaces are mentioned to a professional without exacting knowledge of the attempts to integrate with brain matter. Of the various attempts, the BrainGate interface is without doubt the most famouts, although there are other devices.
The video below is designed to illustrate the process undertaken by BrainGate in 2006. Whilst methods have changed, and electrode sensors have increased beyond the 10 x 10 array of the original, they have not changed significantly enough to warrant a new video: This method, including the suizable interface pedestal, is still current for implanted electrode arrays. The bulky hardware is there to affix to the skull, prevent infection on a permanently open wound, and prevent any of the compontents from shifting about.
Brain activity is read directly, picking up the firing signals from neuron(s) closest to each individual electrode.
Surface Scanning Computer Interface
Surface scanning works by attaching an electrode net onto the scalp. This is the same technique used for medical EEGs (electroencephalograms), except that the results of the electrode array placed on the scalp.are diverted in real-time to control an external device.
EEG based BMI works by reading the electrical field generated by the brain, detecting the tiny distortions in electrical field that leach through above the skull, and generalising their meaning from the rough area of the brain they came from. These surface nets are limited due to the presence of the skull, whose thick bone dampens electrical signals. All they can do is pick up fgeneral activity coming from a specific area of the brain where many neurons are firing. They lack the fidelity to detect somrthing as small as a single neuron firing, and thus are by nature imprecise.
The best a surface mounted interface can do is act as a probibility engine, deducing from general signals, what the user actually intended.
The demonstration video bvelow is of Emotiv's EPOC, the first commercially sold external net, in 2008.
Sub-Surface Scanning Computer Interface
The final BMI interface is another electrode array similar to the EEG method above. However, since the skull is such a problem, it is bypassed. The electrode net is unfolded under the skull, against the surface of the brain. In thuis manner, it can pick up individual signals too weak to make it through the skull. This results in several orders of magnitude increase in scanning fidelity, but pays the price of a permanent implant once again.
A hole is cut in the skull, rather a big hole, and a piece of the skull is removed. The net is then deployed over the brain. The net consists of a polymer sheet containing a grid of electrodes 2 millimetres in diameter and spaced 10 mm apart. This is known as electrocorticography or ECOG (sometimes known as intracranial EEG or icEEG, also referred to as subdural EEG or sdEEG).
The film demonstrated here is from Washington University in St. Louis, USA, where ECOG experiments have finally begun to move on to full human trials.