Structural brain MRI drives a prediction model for seizure recurrence after first unprovoked seizure
Poster No:
350
Submission Type:
Abstract Submission
Authors:
Suyi Ooi1,2, Chris Tailby1,3, Jonas Haderlein1, Heath Pardoe1, Patrick Carney2,4, Moksh Sethi5, Graeme Jackson1,2, David Vaughan1,2
Institutions:
1Florey Institute of Neuroscience and Mental Health, Victoria, Australia, 2Department of Neurology, Austin Health, Victoria, Australia, 3Department of Clinical Neuropsychology, Austin Health, Victoria, Australia, 4Eastern Health Clinical School, Monash University, Victoria, Australia, 5Department of Neurology, Northern Health, Epping, Victoria
First Author:
Suyi Ooi
Florey Institute of Neuroscience and Mental Health|Department of Neurology, Austin Health
Victoria, Australia|Victoria, Australia
Florey Institute of Neuroscience and Mental Health|Department of Neurology, Austin Health
Victoria, Australia|Victoria, Australia
Co-Author(s):
Chris Tailby
Florey Institute of Neuroscience and Mental Health|Department of Clinical Neuropsychology, Austin Health
Victoria, Australia|Victoria, Australia
Florey Institute of Neuroscience and Mental Health|Department of Clinical Neuropsychology, Austin Health
Victoria, Australia|Victoria, Australia
Patrick Carney
Department of Neurology, Austin Health|Eastern Health Clinical School, Monash University
Victoria, Australia|Victoria, Australia
Department of Neurology, Austin Health|Eastern Health Clinical School, Monash University
Victoria, Australia|Victoria, Australia
Graeme Jackson
Florey Institute of Neuroscience and Mental Health|Department of Neurology, Austin Health
Victoria, Australia|Victoria, Australia
Florey Institute of Neuroscience and Mental Health|Department of Neurology, Austin Health
Victoria, Australia|Victoria, Australia
David Vaughan
Florey Institute of Neuroscience and Mental Health|Department of Neurology, Austin Health
Victoria, Australia|Victoria, Australia
Florey Institute of Neuroscience and Mental Health|Department of Neurology, Austin Health
Victoria, Australia|Victoria, Australia
Introduction:
Predicting seizure recurrence after a first unprovoked seizure (FUS) is a major clinical dilemma, with prediction models based on clinical factors alone having poor predictive performance (Bonnett et al., 2022; Ooi et al., 2024). An epileptogenic lesion on MRI-brain (MRI) may satisfy International League Against Epilepsy (ILAE) criteria for the diagnosis of epilepsy after a single seizure (Fisher et al., 2014). However, visual assessment of MRI is often unremarkable, so-called 'MRI-negative' (Ho et al., 2013), making prognostication highly uncertain. We aimed to create a prediction model for 12-month seizure recurrence after FUS, assessing if T1-weighted (T1w) MRI features can improve seizure recurrence prediction beyond clinical variables alone.
Methods:
We assembled a retrospective, multicentre, first seizure case-control cohort (total n=197) from three metropolitan tertiary hospital First Seizure Clinics in Melbourne, Australia. Neither EEG nor MRI showed diagnostic epileptic abnormality. 114 'Controls' had FUS only, and 83 'Cases' had seizure recurrence at 12-month follow up. Clinical T1w 3T MRI was obtained from 7 scanning sites, isotropic 0.9 mm3 resolution. Imaging features were derived from T1w images using FreeSurfer version 7.4.0, and comprised 6 'groups' of variables: cortical thickness, gyral curvature, subcortical volumes, left/right asymmetry, grey-white boundary contrast, and hemispheric/whole-brain metrics. In total, 695 imaging features and 9 clinical variables were considered. Predictive models for seizure recurrence were trained using 4 algorithm classes (logistic regression with L1 penalty, random forest, XGBoost and support vector machine) and 4 filter based feature selection methods (no filter, principal components analysis (PCA), Area Under the Receiver Operating Characteristic Curve (AUC-ROC), feature number), using nested 10-fold cross validation. Results were compared to a baseline logistic regression model derived from clinical factors only. NeuroCombat harmonisation of imaging metrics was performed within each training fold (Fortin et al., 2018). Model performance was assessed by AUC-ROC for predictions on the outer-fold test set. Model interpretation was performed using Shapley values (Molner, 2022). See figure 1 for the analysis pipeline.
Results:
Predictive models trained on the combined T1w image feature set yielded higher AUC-ROC than the clinical-only model (figure 2a). The top performer was a radial-basis-kernel support vector machine, with PCA variance filter fraction of 0.5, yielding AUC-ROC 0.65 (95% confidence interval 0.57-0.73), versus clinical only model AUC-ROC 0.55 (CI 0.47-0.63). Imaging features that were most predictive (largest mean Shapley values) included asymmetry of ventral diencephalon volume (including the hypothalamus), asymmetry of curvature and grey matter volume in orbitofrontal cortex, and asymmetry of parahippocampal gyrus grey matter volume and precuneus. Asymmetry of curvature of the lateral occipital cortex, left supramarginal gyrus and right superior temporal sulci also appeared predictive. See figure 2b and 2c.
Conclusions:
Quantitative imaging features from T1-weighted structural MRI improve predictive models of seizure recurrence after first unprovoked seizure, beyond clinical factors alone. Asymmetry indices of curvature and grey matter volumes, also described in group level studies of patients with established epilepsy (Park et al., 2021), appear the most promising biomarker. Brain regions such as diencephalon, parahippocampal gyrus and precuneus have been described as involved in seizure propagation or the development of epilepsy (Bernasconi et al., 2003; Bian et al., 2024; Galovic et al., 2019). Despite modest overall performance of the best models in this study, the increment obtained by adding imaging features suggests that further addition of other imaging contrasts and feature refinement holds future hope for clinical application of such models.
Disorders of the Nervous System:
Neurodevelopmental/ Early Life (eg. ADHD, autism) 1
Modeling and Analysis Methods:
Classification and Predictive Modeling 2
Keywords:
DISORDERS
Epilepsy
Machine Learning
STRUCTURAL MRI
Other - clinical prediction model
1|2Indicates the priority used for review
Abstract Information
By submitting your proposal, you grant permission for the Organization for Human Brain Mapping (OHBM) to distribute your work in any format, including video, audio print and electronic text through OHBM OnDemand, social media channels, the OHBM website, or other electronic publications and media.
The Open Science Special Interest Group (OSSIG) is introducing a reproducibility challenge for OHBM 2025. This new initiative aims to enhance the reproducibility of scientific results and foster collaborations between labs. Teams will consist of a “source” party and a “reproducing” party, and will be evaluated on the success of their replication, the openness of the source work, and additional deliverables. Click here for more information. Propose your OHBM abstract(s) as source work for future OHBM meetings by selecting one of the following options:
Please indicate below if your study was a "resting state" or "task-activation” study.
Healthy subjects only or patients (note that patient studies may also involve healthy subjects):
Was this research conducted in the United States?
Were any human subjects research approved by the relevant Institutional Review Board or ethics panel? NOTE: Any human subjects studies without IRB approval will be automatically rejected.
Were any animal research approved by the relevant IACUC or other animal research panel? NOTE: Any animal studies without IACUC approval will be automatically rejected.
Please indicate which methods were used in your research:
For human MRI, what field strength scanner do you use?
Which processing packages did you use for your study?
Provide references using APA citation style.
Bian, X., Yang, W., Lin, J., Jiang, B., & Shao, X. (2024). Hypothalamic-Pituitary-Adrenal Axis and Epilepsy. J Clin Neurol, 20(2), 131-139.
Bonnett, L. J., Kim, L., Johnson, A., Sander, J. W., Lawn, N., Beghi, E., Leone, M., & Marson, A. G. (2022). Risk of seizure recurrence in people with single seizures and early epilepsy - Model development and external validation. Seizure, 94, 26-32.
Fisher, R. S., Acevedo, C., Arzimanoglou, A., Bogacz, A., Cross, J. H., Elger, C. E., Engel, J., Jr., Forsgren, L., French, J. A., Glynn, M., Hesdorffer, D. C., Lee, B. I., Mathern, G. W., Moshé, S. L., Perucca, E., Scheffer, I. E., Tomson, T., Watanabe, M., & Wiebe, S. (2014). ILAE official report: a practical clinical definition of epilepsy. Epilepsia, 55(4), 475-482.
Fortin, J.-P., Cullen, N., Sheline, Y. I., Taylor, W. D., Aselcioglu, I., Cook, P. A., Adams, P., Cooper, C., Fava, M., McGrath, P. J., McInnis, M., Phillips, M. L., Trivedi, M. H., Weissman, M. M., & Shinohara, R. T. (2018). Harmonization of cortical thickness measurements across scanners and sites. Neuroimage, 167, 104-120.
Galovic, M., van Dooren, V. Q. H., Postma, T. S., Vos, S. B., Caciagli, L., Borzì, G., Cueva Rosillo, J., Vuong, K. A., de Tisi, J., Nachev, P., Duncan, J. S., & Koepp, M. J. (2019). Progressive Cortical Thinning in Patients With Focal Epilepsy. JAMA Neurology, 76(10), 1230-1239.
Ho, K., Lawn, N., Bynevelt, M., Lee, J., & Dunne, J. (2013). Neuroimaging of first-ever seizure: Contribution of MRI if CT is normal. Neurol Clin Pract, 3(5), 398-403.
Molner, C. (2022). Interpretable Machine Learning - A Guide for Making Black Box Models Explainable (2 ed.).
Ooi, S., Tailby, C., Nagino, N., Carney, P. W., Jackson, G. D., & Vaughan, D. N. (2024). Prediction begins with diagnosis: Estimating seizure recurrence risk in the First Seizure Clinic. Seizure, 122, 87-95.