Real-time neural spike sorting using dynamic graph network clustering

The human brain uses neural spikes to encode information, such as the images seen by the eyes, sound heard by the ears, smells sensed by the noise, as well as words to be spoken by the mouth.These neural spikes encoded information is processed by ~100 billion neurons in the brain in a highly parallel and nonlinear manner.If the firing times of neurons can be measured, it is possible to decode neural information and in turn apply interventional measures to alter abnormal brain activities to control neurological diseases. Recent technological advances are focused on development scientific tools, such as high channel count neural probes, to allow measurements of neural spike firing for many neurons. However, algorithms to allow analyzing these large numbers of neural spikes are also needed to be developed.In this talk, a real-time neural spike sorting algorithms – GEMsort (Graph nEtwork Multichannel neural spike sorting), which is based on dynamic graph network clustering will be discussed to allow sorting large number of neural spikes in real-time. This spike sorting technique may open up new opportunities to further analyze these sorted neural spikes to extract neural dynamics in real-time, which may lead to closed-loop neural modulation to manage neurological disease symptoms in the future.