Radiomics in Temporal Lobe Epilepsy: Revealing Unseen Lesions
Abstract number :
3.247
Submission category :
5. Neuro Imaging / 5A. Structural Imaging
Year :
2022
Submission ID :
2204952
Source :
www.aesnet.org
Presentation date :
12/5/2022 12:00:00 PM
Published date :
Nov 22, 2022, 05:27 AM
Authors :
Christopher Veys, MD – Advocate Aurora Health; Ivy Sullivan, B.S – Rush University Medical Center; Rebecca O'Dwyer, MD – Assistant Professor, Epilepsy, Rush University Medical Center; Travis Stoub, PhD – Associate Professor, Epilepsy, Rush University Medical Center
Rationale: Focal onset epilepsies account for up to 60% of all adult epilepsies, of which the majority are temporal lobe epilepsy (TLE). Approximately 30% of patients with EEG confirmed drug resistant temporal lobe epilepsy have normal MRI scans. These patients are less likely to undergo surgery and have a significantly lower likelihood of seizure freedom (51%) as compared to those with hippocampal sclerosis (75%), often requiring additional costly and at times invasive evaluations. Radiomics is an emerging advanced computational analysis using machine learning and neural networks to extract a high number of qualitative and quantitative characteristics/data known as “features” from clinical images. This study examines the use of radiomics as a means of finding subtle MRI abnormalities not identified by visual inspection.
Methods: This is a retrospective analysis of epilepsy monitoring unit admissions at Rush University Medical Center between 2020 and 2022 for patients with EEG confirmed temporal lobe epilepsy and non-lesional 3T MRI brain epilepsy protocol read by a neuro-radiologist. Freesurfer was used to segment images into regions of interest. Manual corrections were then performed as needed. Specific regions of interest (ROI) were then selected based on their involvement in temporal lobe epilepsy, such as the hippocampus, amygdala, and entorhinal cortex. Radiomic features were extracted from each ROI using an in-house program. These features were selected based on prior research showing their ability to differentiate temporal lobe epilepsy patients from controls.
Results: Three radiomics features (volume, grey level nonuniformity, and run length nonuniformity sigma) were statistically significant when comparing a patient’s entorhinal cortex, unilateral to their epileptiform discharges, to the entorhinal cortex of control patients.
Conclusions: Radiomics has the ability to identify subtle MRI abnormalities within the mesial temporal structures in patients with medically intractable, non-lesional temporal lobe epilepsy. These findings could be used in the future to further identify good candidates for a curative resective surgery.
Funding: None
Neuro Imaging