Authors :
Presenting Author: Suyi Ooi, MBBS – The Florey Institute of Neuroscience and Mental Health, Melbourne Brain Centre, Heidelberg, Melbourne, Victoria, Australia
Chris Tailby, PhD – The Florey Institute of Neuroscience and Mental Health, Melbourne Brain Centre, Heidelberg, Melbourne, Victoria, Australia
Heath Pardoe, PhD – The Florey Institute of Neuroscience and Mental Health, Melbourne Brain Centre, Heidelberg, Melbourne, Victoria, Australia
Patrick Carney, MBBS PhD – Eastern Health Clinical School, Monash University, Box Hill, Victoria, Australia
Moksh Sethi, MBBS – Faculty of Medicine, Dentistry and Health Sciences, The University of Melbourne, Melbourne, Victoria, Australia
Jonas Haderlein, PhD – The Florey Institute of Neuroscience and Mental Health, Melbourne Brain Centre, Heidelberg, Melbourne, Victoria, Australia
Graeme Jackson, MD PhD – The Florey Institute of Neuroscience and Mental Health, Melbourne Brain Centre, Heidelberg, Melbourne, Victoria, Australia
Rationale:
Predicting seizure recurrence following a first unprovoked seizure (FUS) remains a significant clinical challenge, especially when routine clinical magnetic resonance imaging (MRI) and EEG do not reveal epileptiform abnormalities. Here, we incorporate quantitative structural MRI-derived biomarkers into prediction models for seizure recurrence at 12 months and identify brain structural features that are predictive of seizure recurrence.
Methods:
We analysed a retrospective, multicentre cohort of 197 adult patients with FUS, comprising 83 with seizure recurrence, and 114 with no seizure recurrence at 12 months. All participants had normal or non-diagnostic MRI and EEG findings. Morphometric features were extracted from clinical 3T T1-weighted MRI using FreeSurfer. Machine learning algorithms were trained on combined imaging and clinical features, using nested cross-validation for model selection. Performance was compared to a logistic regression model based on clinical features only.
Results:
The best performing model, a support vector machine trained on a combination of imaging features and clinical factors, achieved an AUC of 0.65 (95% CI: 0.59–0.73), significantly better than chance (p = 0.01 when compared to an AUC of 0.5). In contrast, the logistic regression model trained on clinical factors alone yielded an AUC of 0.57 (95% CI: 0.49–0.65), not statistically different to chance (p = 0.28). Direct comparison between these two models was not statistically significant (95% CI for the difference in AUC: –0.019 to 0.173, p = 0.11). The most important imaging features for prediction were inter-hemispheric asymmetry of subcortical and cortical grey matter volumes and regional gyral curvatures, particularly in fronto-parietal and limbic regions.
Conclusions:
Quantitative structural MRI biomarkers add additional information for machine learning models for prediction of seizure recurrence. Grey matter cortical and subcortical asymmetries, most likely developmental, are associated with seizure recurrence and precede the diagnosis of epilepsy.
Funding:
S.O. is supported by a National Health and Medical Research Council (NHMRC) post-graduate scholarship (project ID 2022072) and the Australian New Zealand Association of Neurologists (ANZAN) Education and Research Fund. D.N.V. and C.T. are supported by an NHMRC project grant (APP1157145). C.T., H.R.P., J.H., G.D.J. and D.N.V. are supported by the Australian Epilepsy Project, funded by the Australian Government Medical Research Future Fund, Grant/Award Number: MRFF75908 and RFRHPSI000008; Victorian-led Frontier Health and Medical Research Program.