GS: At the moment, we don't have a very good way to record from a lot of neurons over long periods of time. A sensor that could be implanted and stay in place and produce a good signal for a long time, while recording a large number of areas in the brain with high fidelity, currently does not exist. The ECoG electrodes are probably the closest to what I just described and have a clear technical trajectory to satisfying those requirements. It is important to understand that the technique does not record individual neurons but records the activity of a lot of neurons at once. However, ECoG can distinguish and detect the activity of much smaller regions of neurons than EEG. EEG, since it is much further away from the neurons, can record only fairly gross patterns of activity in the brain. With ECoG you can presumably pick up signals that are only a millimeter apart, a 1mm spatial resolution in detecting neuron activity. EEG's spatial signal resolution is two orders of magnitude worse since it is in the centimeter range. It's very similar to being in a football stadium and you are sitting next to six other people and you can hear what they say. If you go to a particular sector you can hear some people cheer for one team or the other. If you are outside the stadium you can't even hear that, but you can tell by the crowd when someone has scored a touchdown.
How much information about hand movements’ direction is encoded by ECoG signals in brain areas.
GS: All these devices are trying to apply the technology to a particular area: video gaming. I'm not sure that the number of electrodes is important. From an industry and venture capital perspective, what's important is how much money will it make and that is critically dependent on creating a product that people are willing to spend money on. The jury is still out, and it is not clear whether or not they can create a product that can sustain an industry and represents a product that has than just a geek or wow factor. Neurosky has some big partners like Mattel, which uses its headset with its Mindflex game. And then there's the Uncle Milton Star War Force Trainer. But it is too early to gauge their success. As long as the game is fun they'll be successful. Clearly, by adding the words "brain control" or "brain device" that's going to add to the wow factor, but what they actually record remains to be seen. Given the location of where the electrodes are placed, they are likely to pick up the muscle activity from grimacing or making other facial expressions while concentrating, rather than picking up brain activity. But it is possible that they could be picking up brain activity too. In the end, I feel the commercial success of these games will outweigh the technical details of what signals they are actually picking up.
ET: The videos I've seen of some of the Emotiv headsets show people gesturing while thinking in order to help signal detection. Also, it takes the system about six seconds to detect a response. Can EEG based headsets produce better results than these?
GS: It already does, as we have shown in the laboratory. As I mentioned earlier with the published thought control demonstration using a 3D cursor, the subject could hit one of four to eight targets in a second or two. You move the cursor towards the target with the cursor location being updated approximately every 50 milliseconds.
ET: Is that an EEG or ECoG system?
GS: EEG. ECoG of course can also do this, but that system use scalp-recorded EEG technology. However, that level of success requires a lot of training on the part of the subject. It also requires a very robust, reliable environment like a laboratory, where a highly trained technician takes a half hour to set up the subject and a conductive gel is applied to the sensor before putting it on the scalp. This is an environment quite different from and superior to what the average consumer will have; consumer devices typically use dry electrodes that do not perform as well. Commercially, if your game requires your customer to have a laboratory environment, and forces them to train for three months before they can play, then it's not going to sell very well. Another factor is the amount of preparation that the consumer has to do when setting up and tearing down the equipment for each session, which will also affect the consumer's enjoyment of the game. The reality is that you have to create a product that can perform in a noisy environment and with a minimal amount of training, preparation, and teardown time. Therefore the level of performance you can expect from a consumer device has to be a lot less.
(Note: The Emotiv headset requires you to moisten and then seat 16 sensor pads before wearing the headset and put them back in their storage case when you are done, a process that combined takes about 10 minutes. Note that although they are moistened, you do not have to apply a messy conductive gel.)
No comments:
Post a Comment