Research Articles

Share     

Using Data to Fight Epilepsy


Epilepsy patients can take heart – a new invention by NTU's Electrical and Electronic Engineering (EEE) scientists could help doctors to ease or even cure their disease.

Professor Justin Dauwels and his team have created an automated program to help neurologists to locate seizure onset zones in epilepsy patients’ brains. Their breakthrough could lead to more effective surgical therapy that reduces or stops seizures.

Epilepsy, characterized by recurring, unprovoked seizures, affects about 65 million people globally, but only about 70% are successfully treated with drugs that are the main form of treatment. Even then, the medication can cause substantial side effects.

Surgical therapy, such as the implanting of stimulation devices, can help patients, but surgeons need to know the location of the seizure onset zones (SOZ). Currently, neurologists combine data from non-invasive procedures such as magnetic resonance imaging (MRI) and external electroencephalograms (EEG) to locate the zones, but this may not be sufficient.

Intracranial EEG monitoring – including stereo-EEGs, where depth electrodes are surgically implanted in brain tissue to capture neural signals – may be required, but since the seizures are hard to predict, it takes a long time to capture enough data. Such procedures are long, invasive, costly and carry infection risks.

Working with neurologists from the United States’ Massachusetts General Hospital and Harvard Medical School, the EEE team developed an automated program that identifies the seizure onset zone using just one hour’s recording of stereo-EEG, potentially reducing patients’ time in hospital.

The program looks at a combination of stereo-EEG features. Where most EEG studies explore the correlation between one EEG feature and the zones, the EEE researchers tested numerous possible combinations of EEG features – 255 in total – to find the one with the best localisation accuracy.

For example, electrodes in the seizure onset zone generally record more energy at low-frequencies, in addition to more high-frequency oscillations and interictal discharges. The EEE two-step program identifies electrodes suspected to be in the zone, and then “zooms in” to check and rank each contact point of those electrodes. At this contact-point level, the program has 74% accuracy.

Professor Justin Dauwels said: “The results demonstrate that stereo-EEG has great potential in localising the seizure onset zones, and a rigorous, quantitative method can be employed which moves significant clinical decisions from the subjective to objective.”


By Professor Justin Dauwels, School of EEE

 
Click here to find out more.
 
 
Published on: 19-Oct-2016