Skip to navigation Skip to main content Skip to footer

Approved Research

Multi-Spectral Structural Post-Processing for Lesion Detection in Epilepsy

Principal Investigator: Professor Niels Focke
Approved Research ID: 58506
Approval date: July 20th 2020

Lay summary

Epilepsy is a severe neurological disease that can be difficult to control with medication in about 1/3 of patients. For these patients epilepsy surgery, i.e. the surgical removal of the epileptogenic focus is a very successful option. However, before any surgery can be considered, the origin of the seizures needs to be found. MRI is an important tool in this endeavor and can be used for detecting changes of brain structures that are associated to the epilepsy. This can encompass malformations, scars or tumors. Many of these changes can be very subtle and easily be overlooked given the complex anatomy of the brain. To help the epilepsy specialists and radiologists in detecting such lesions, computer-based processing can be utilized.

Within this project we want to study, if the huge dataset of UK Biobank can help to improve the performance of computer-based processing ("post-processing") in epilepsy and which methods can be applied so make brain better comparable between different MRI scanners. Specifically, we plan to use the MRI scans acquired by the UK Biobank in subjects without a neurological or brain-affecting general condition to serve as a control dataset against which each individual epilepsy patient is then compared. This procedure can highlight areas of the brain where this patient differs from the "norm", i.e. areas which may be related to the structural cause of the epilepsy and seizures. We will use the huge dataset of UK biobank to test different strategies and techniques for selecting this control cohort, e.g. using age and gender matched controls and assess if this improves the classification. We also want to assess different strategies to make MRI from different scanners comparable with each other and test the impact of this on epilepsy lesion detection and general biological features (e.g. age and gender differences). Finally, will also use machine learning methods, i.e. a computer program that can learn which variations of brain morphology are normal and which are not, to further enhance this process.

We believe this project will improve MRI processing methods in general and finding the causes and origin of epilepsy in particular.