~ Data Mining - DIMACS REU 2006 At Rutgers University

Problem: Seizure Prediction

Seizures are the second-most-common neurological disorder, affecting millions of patients everyday. Patients suffering from seizures may suffer an attack at any time. Then severity of seizures ranges from unnoticable to inducing severe convulsing in the patient for minutes at a time.

Predicting seizures remains a difficult task. Because the brain is a chaotic system, and furthermore because we are limited in our capacity to extract information from the brain, making precise predictions based on simple observation of neurological data is all but impossible.

Being able to predict seizures minutes in advance would carry many benefits. Doctors would have more time to counteract or prepare the patient for the seizure; patients may even have more independence if they could know when a seizure was imminent. Such a predictive mechanism could also shine light on the forces that underlie and cause seizures, enabling further research and the development of new, more effective treatments.

Support Vector Machines

Support Vector Machines (SVMs)