PhD defense Giske Fiskarbekk Opheim: Utilizing 7 Tesla MRI and automated segmentation – A new era in the presurgical evaluation of patients with severe epilepsy
Friday, 5. February 2021, 14:00 - 17:00
Hits : 658
by NRUadmin
Place
Virtual: https://us02web.zoom.us/j/88428230690?pwd=djZVd2QwYUk3NkF4bHJSUnpUTU5EUT09 Meeting ID: 884 2823 0690, Passcode: 237283

Bedømmere
 
Professor Steen Hasselbalch  (chair, Danish Dementia Research Center, Department of Neurology, Rigshospitalet, University of Copenhagen)
Associate Professor Steffen Ringgaard (Department of Clinical Medicine - The MR Research Centre, Aarhus University, Denmark)
Professor Asta Haaberg (Norwegian University of Schience and Technology and St. Olav's Hospital, Trondheim, Norway)

Summary

Epilepsy is a common chronic neurological disease that affects 1% of people of all ages worldwide. About one third of patients suffer from drug-resistant seizures, and in those with focal seizure onset, surgery may be the only cure. With MRI, the clinicians look for epileptogenic lesions that help localize the seizure origin. Mesial temporal sclerosis (MTS) is the most common epileptogenic lesion found in MRI in drug-resistant epilepsy. The histopathological substrate of MTS is hippocampal sclerosis (HS). HS can be further divided into subtypes, that are desired to identify before surgery. Automated MRI segmentation and 7T MRI are emerging techniques with a great potential to improve detection of epileptogenic lesions, and also to add new and unique information to the epilepsy surgery evaluation process. Automated MRI segmentation tools may identify changes in volume and shape not possible to identify by visual assessment. 7T MRI may both help visually detect and subclassify epileptogenic lesions, and the high-resolution images may also improve automated MRI segmentations.

In the first article, we tested if automated hippocampal subfield segmentations in 3 Tesla (T) MRIs could identify specific patterns of hippocampal subfield volume loss that correspond to histopathological classification in patients with HS. We found excellent correspondence with HS diagnosis, but failed to identify subfield volume patterns that allow us to separate patients with two different HS subtypes. This study will be complemented with additional and independent data from the Norwegian epilepsy surgery program, which will allow us to further test whether patterns of MRI volume changes are better markers of seizure outcome and memory function than the histopathological classification.

In the second article, we compared standard quantitative hippocampal features between clinical 3T and 7T MRI in mesial temporal sclerosis (MTS) and non-MTS patients. Counter-intuitively, we found equally discriminative ability in the two MRI field strengths. Two main limiting factors were lack of histological ground truth and detailed understanding of differences in segmentation algorithm performance between 3T and 7T MRI.

The third article presents the first consensus-based recommendations for setting up and evaluating 7T MRI in epilepsy – a project I co-initiated and coordinated between teleconferences, and finalized manuscript draft in. This work was based on experiences from the 7T Epilepsy Task Force – an international group of 21 world-leading centers within this field. The article is intended as a handheld for centers that are new to 7T MRI in epilepsy.

In conclusion, the three studies provide information that is important to consider when implementing 7T MRI, automated segmentation and the combination thereof, in the presurgical evaluation of drug-resistant epilepsy patients. Our findings also demonstrate that in the cross-field between clinical and engineering considerations, more research is warranted to fully uncover characteristics in segmentation performance and potential of contributions from 7T MRI when compared to 3T. Lastly, having a consensus-based set of guidelines will hopefully help when setting up an epilepsy-specific 7T MRI protocol.