Being able to predict when a person could have a seizure would be an invaluable tool for the thousands of patients who live with seizure disorders. New research from Australia has found that using machine learning, or deep learning, to analyze brain wave patterns could improve the odds of detecting seizures before they strike. Using hundreds of EEG brainwave results from seizure patterns in patients, they compared those models to a dataset of normal brain activity. Being able to crunch the numbers on healthy brain function, alongside the data of seizures in patients, collected over the long-term, allowed a machine learning system to predict, with some success, when a seizure might be about to happen.
While the data is patient-specific and cant be used quite yet, the method of improving the prediction of seizures was the real gift of this work. Now that there is a set of algorithms to apply to each patient's specific condition, the rate of accurately predicting an episode is much improved.