Individualized Prediction of Post-Traumatic Epilepsy Risk and Associated Cognitive Deficits Using Connectome Analysis and Machine Learning
Abstract
Traumatic brain injury (TBI) is a significant health hazard for Service Members and Veterans during times of both peace and war. TBI is one of the major causes of epilepsy, yet the link between TBI and post-traumatic epilepsy (PTE) is not well understood. Prediction and, if possible, prevention of the development of PTE is a major unmet challenge. The goal of the proposed project is to use magnetic resonance imaging (MRI) data collected immediately after the injury to predict the onset of PTE on an individual basis. This critical insight will allow early intervention and targeted therapies for individuals identified as being at high risk of developing PTE. This 3-year project on individualized prediction of PTE leverages and extends the team’s previous work, funded by the 2018-2022 Epilepsy Research Program (ERP) award on finding imaging-based biomarkers for post-traumatic epilepsy using MRI data. The project also directly addresses one of the ERP focus areas: Markers and Mechanisms, and subarea Predictive biomarkers of epileptogenesis (acute and chronic). A large pre-collected longitudinal dataset of TBI patients available in the FITBIR database, with a subpopulation of PTE, offers a unique opportunity for identifying brain regions and networks involved in epileptogenesis as well as for designing and training algorithms to predict PTE. The overall aim and focus of this project are to identify imaging and connectome features that are associated with PTE, and then to use these features to predict PTE onset and cognitive deficits over time, on an individual subject basis. The preliminary results obtained by our team on the MagNeTS dataset, presented in the proposal and our publications in this area, suggest that high predictivity is achievable for PTE onset and cognitive trajectory. To achieve our goals, we will analyze data from three large studies involving military as well as non-military personnel with TBI. The TBI data on military personnel is collected at the National Intrepid Center of Excellence at the Walter Reed National Military Medical Center. The non-military personnel data will be from the TRACK-TBI consortium and Maryland MagNeTS study. All the data is available to us from the FITBIR database. The use of multiple datasets will ensure that the results of the study are not dataset specific. We will apply novel methods for lesion mapping, anomaly detection, and mapping brain connectivity changes for quantifying local structural and network changes in the brain based on MRI data. Through classical statistical analysis, we will then identify the subset of these imaging markers, and their location in the brain, that are most strongly correlated with the onset of PTE. We will also use the markers as the basis for developing and training a deep neural-net for individualized prediction of PTE. Additionally, we will establish bounds on the uncertainty in our predictions. The connectome analysis conducted as part of this work may help in identifying the future origins of epileptic seizures in the brain, and potentially even the severity/recurrence rate for seizures as they develop. This information in turn may help to select appropriate therapies to reduce the effects of disrupted brain networks. Leveraging the team’s expertise in machine learning and connectomics, the project is aimed at modeling the complex interactions and relationships between risk factors, as reflected in imaging data collected shortly after injury, and epilepsy. The multivariate statistical analysis we will perform to determine these relationships may also provide insights into the pathogenesis of epilepsy. In summary, this project, if successful, would allow improved targeting of preventive care for TBI-affected Service Members and Veterans.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jan 04, 2024
- Source ID
- HT94252310149
Entities
People
- Anand A Joshi
Organizations
- United States Army
- University of Southern California