Prediction of response to antiseizure medication based on MRI
Poster No:
1094
Submission Type:
Abstract Submission
Authors:
Mohammad-Reza Nazem-Zadeh1, Richard Chang1, Hadi Kamkar2, Debabrata Mishra1, Duong Nhu1, Deval Metha1, Daniel Thom1, Zhibin Chen1, Zongyuan Ge1, Ben Sinclair1, Jacqueline French3, Terence O’Brien1, Meng Law1, Patrick Kwan1
Institutions:
1Monash University, Melbourne, Victoria, 2Tarbiat Modares University, Tehran, Tehran, 3NYU Comprehensive Epilepsy Center, New York, NY
First Author:
Co-Author(s):
Introduction:
Anti-seizure medications (ASMs) are the first-line treatment for epilepsy, yet they are effective in controlling seizures in only about 60% of patients. Individual response to ASM treatment is unpredictable (Lancet, 2012; Zhang et al., 2018). This study aimed to develop and validate artificial intelligence (AI) models using clinical and magnetic resonance imaging (MRI) information to predict ASM response in people with newly diagnosed epilepsy (Bonacchi, Filippi, & Rocca, 2022; Pesapane, Codari, & Sardanelli, 2018).
Methods:
Patients with newly diagnosed epilepsy treated with ASM monotherapy at the Alfred Hospital, Melbourne, Australia formed the development cohort. We developed machine-learning (ML) models (Duda & Hart, 1973; Hoerl & Kennard, 1970; Vapnik, 1995; Breiman, 2017; Cox, 1958) employing various combinations of clinical features, ASMs and brain MRI features (by FreeSurfer software; https://surfer.nmr.mgh.harvard.edu; Fischl, 2012) to predict the probability of seizure freedom (SF) for 12 months while taking the first or second ASM monotherapy. Similarly, we developed deep learning (DL) algorithms to predict SF based on various combinations of brain MRI modalities. We externally validated the developed models in individuals recruited to the Human Epilepsy Project (validation cohort; Bank et al., 2022).
·Fig. 1. The DL models predict drug outcomes based on a combination of sequential data (drug sequences), brain MRI, and tabular data (demographic and clinical features).
Results:
Among 154 patients (36% female, median age 39 years) included in the development cohort, 117 (76%) were seizure-free while taking the first or second ASM monotherapy. Among ML models, the model integrating MRI, ASMs and clinical features with a Decision-Tree as the classifier and Genetic Algorithm as the feature selection method reached the best performance in prediction of SF with an F1-Score of 71%±1% (95% confidence interval), which was significantly higher than the best performance of other combinations (p < 0.001). We developed a fusion DL model comprising an 18-layer 3D videoResNet (He, Zhang, Ren, & Sun, 2016), a transformer encoder, and a dual linear neural network to integrate multimodal MRI data, ASM regimens, and clinical characteristics, respectively. It had an internal cross validation resulted in F1-Score of 75% ±5% in prediction of SF, outperforming other DL models with less complex architecture or integration of fewer imaging modalities. It also outperformed conventional ML models with a significantly higher F-Score (p < 0.001). External validation demonstrates the superiority of DL models compared to conventional ML models (the highest F1-score of 66%±4% for DL vs. 59%±3% for ML models, p < 0.001).
·Figure. 2. The F1-scores of the predictive DL models of seizure freedom (SF).
Conclusions:
AI-driven models incorporating neuroimaging and clinical data offer a promising approach to predicting epilepsy treatment outcomes. The fusion model developed in this study has the potential to guide personalized treatment strategies by improving clinical decision-making and patient outcomes.
Modeling and Analysis Methods:
Classification and Predictive Modeling 1
Neuroanatomy, Physiology, Metabolism and Neurotransmission:
Cortical Anatomy and Brain Mapping 2
Subcortical Structures
Neuroinformatics and Data Sharing:
Informatics Other
Keywords:
Cortex
Epilepsy
Machine Learning
Structures
Sub-Cortical
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.
Bonacchi, R., Filippi, M., & Rocca, M. (2022). Role of artificial intelligence in MS clinical practice. NeuroImage: Clin 35: 103065. In.
Breiman, L. (2017). Classification and regression trees: Routledge.
Cox, D. R. (1958). The regression analysis of binary sequences. Journal of the Royal Statistical Society Series B: Statistical Methodology, 20(2), 215-232.
Duda, R. O., & Hart, P. E. (1973). Pattern classification and scene analysis (Vol. 3): Wiley New York.
Fischl, B. (2012). FreeSurfer. Neuroimage, 62(2), 774-781.
He, K., Zhang, X., Ren, S., & Sun, J. (2016). Deep residual learning for image recognition. Paper presented at the Proceedings of the IEEE conference on computer vision and pattern recognition.
Hoerl, A. E., & Kennard, R. W. (1970). Ridge regression: Biased estimation for nonorthogonal problems. Technometrics, 12(1), 55-67.
Lancet, T. (2012). Wanted: a global campaign against epilepsy. In (Vol. 380, pp. 1121): Elsevier.
Pesapane, F., Codari, M., & Sardanelli, F. (2018). Artificial intelligence in medical imaging: threat or opportunity? Radiologists again at the forefront of innovation in medicine. European radiology experimental, 2, 1-10.
Vapnik, V. (1995). Support-vector networks. Machine learning, 20, 273-297.
Zhang, J. h., Han, X., Zhao, H. w., Zhao, D., Wang, N., Zhao, T., . . . Han, J. y. (2018). Personalized prediction model for seizure‐free epilepsy with levetiracetam therapy: a retrospective data analysis using support vector machine. British Journal of Clinical Pharmacology, 84(11), 2615-2624.