Abstracts

Rationale: We use multivariate analysis of structural MRI to predict the age of individuals with epilepsy and healthy subjects without seizures. Epilepsy subjects were stratified into two groups, comprising individuals with new onset focal epilepsy recruited as part of the Human Epilepsy Project, and individuals with intractable focal epilepsy being assessed for surgical resection at the NYU Comprehensive Epilepsy Center.Methods: Whole brain T1 weighted MRI was obtained on a single scanner using an MPRAGE volumetric acquisition (1.3 x 1 x 1.3 mm). The age of each individual was predicted using the PRONTO multivariate machine learning matlab toolbox [1]. Gray matter and white matter segmentations, derived using the SPM software package, were used as input data. The machine learning algorithm was trained on a large independent group of healthy control data obtained from a number of publicly available imaging databases. Age prediction was carried out using the white matter and gray matter segmentation images separately, yielding two age estimates per subject. The difference between the predicted age of the individual and the actual chronological age was compared between the three groups using a general linear model, with age and sex included as covariates. Individuals with a positive difference have an older predicted age than their chronological age.Results: Individuals with intractable epilepsy had a difference between predicted brain age and chronological age that was on average 8.8 years older than healthy controls (p = 0.003) when using the white matter segment as input to the brain age predictor model. No difference in predicted brain age was observed in new onset epilepsy cases (-1.5 years, p = 0.55). No differences in predicted brain age were observed when using the gray matter segment for either intractable cases (0.84 years, p = 0.72) or new onset cases (-1.06 years, p = 0.64).Conclusions: Intractable focal seizures are associated with an increased brain age of approximately 8.8 years when an individual's age is predicted using white matter images derived from structural MRI. The absence of similar changes in new onset focal epilepsy cases suggests that ongoing seizures may be responsible for the brain aging phenomenon observed in this study. A potentially useful clinical application of this technique will be to identify individuals with brains that resemble intractable cases (ie. larger difference between predicted vs chronological age) early in the course of their epilepsy. [1] Cole et al, Annals of Neurology, 77 (4): 571-581 Supported by The Epilepsy Study Consortium (ESCI), a non-profit organization dedicated to accelerating the development of new therapies in epilepsy to improve patient care. The funding provided to ESCI to support HEP comes from industry, philanthropy and foundations (UCB Pharma, Pfizer, Eisai, Lundbeck, Finding A Cure for Epilepsy and Seizures, The Andrews Foundation, Friends of Faces and others).