International Business Machines (IBM) Corporation and University of Melbourne partner for a very unique but significant undertaking. They are trying to test if TrueNorth may be used to predict seizures.
Contrary to popular belief, seizures, a condition defined as the brain's electrical activity going haywire, don't usually occur suddenly. Many diagnosed with it can actually feel little changes in their behavior or motion prior to an attack.
However, the time difference between these seizures and the felt changes can be so small that they won't be able to plan their steps clearly, potentially putting their lives in danger.
So far, there's no known cure for seizures, but they can be managed, beginning with having more accurate way of predicting changes in brain waves long before the attacks.
This is what IBM and neurology researchers from the University of Melbourne have been working on [PDF] using a specially designed chip that works like a human brain.
This chip is called TrueNorth, whose architecture is composed of millions of digital synapses and neurons, which do not only communicate but also run multiple complex tasks with a very high degree of efficiency and a small amount of energy input.
TrueNorth is IBM's foundation to artificial neural networking and machine learning. A system with the chip can analyze similar or related sets of data until it can recognize different types of patterns such as speech, face, and, possibly, brain waves.
How will this technology be helpful to early seizure detection? Theoretically, IBM can develop a "wearable system that you put on a subject — on a patient — and have it do analysis in real-time, 24/7," said Stefan Harrer, IBM researcher in the university.
The wearable communicates with a brain implant, which records these brain waves. Over time, TrueNorth will be able to determine distinct patterns that may occur, say, a few hours or even a few weeks before a seizure.
The concept may also be used to "replace broken neural systems with machines — machines that can interact with the brain in a very natural way," says Freestone, University of Melbourne senior research fellow.
To test the probability of the concept, the researchers used the data collected through a hand-squeezing experiment using an electroencephalogram (EEG), which monitors abnormalities in the electrical activity of the brain.
After learning the patterns, the system with the chip is able to predict which hand was used for squeezing 76 percent of the time — less than what the researchers hoped but still proof that indeed systems can be trained.
Although the wearable system won't be available anytime soon, it's something we can all look forward to. Who knows, it can be used for other types of conditions like arrhythmia or heart attack.
