Project Summary:
Epilepsy is a prevalent neurological disorder affecting over 45.9 million people globally. Epilepsy is one of the top-ten neurological conditions with significant disability-adjusted life years in North America. The CDC's report in 2015 estimates that 59,600 people in Wisconsin have active epilepsy, including 7,900 children and 51,700 adults. In 2019, the per-patient annual cost of epilepsy in the U.S. was significantly higher than for non-epilepsy patients, with direct costs of $9,984.04 and indirect costs of $5,924.50. About 20-30% of patients with epilepsy become refractory to antiseizure medications, often known as intractable epilepsy. Predicting treatment outcomes and choosing effective medications for epilepsy is difficult due to the lack of reliable biomarkers. In the U.S., studies show that uncontrolled seizures lead to complications, sudden death, and a significant economic burden that is two-to-eight times that of well-controlled epilepsy. In addition, refractory epilepsy also has a significant negative effect on development.
Recent neuroimaging focuses on network-level brain mechanisms, using graph theory and diffusion-weighted MRI to map white matter pathways and create structural connectomes. These advances offer deeper insights into epilepsy by enhancing understanding of brain connectivity and related network disruptions. Although MRI is widely used to evaluate epilepsy, its full potential is often overlooked, especially with the analysis of advanced sequences like diffusion-weighted imaging for tractography and structural connectivity analysis. This underutilization misses a crucial opportunity to leverage clinically acquired imaging data for more precise clinical decision-making. Further, recent advancements in machine learning have significantly improved neuroimaging analysis, yet most applications in epilepsy remain limited and focused on adult patients.
Pediatric patients, with their developing brains and a wide variety of epilepsy syndromes, present neuroradiology challenges that require unique structural connectivity analysis approaches for accurate diagnosis, treatment prediction, and understanding of cognitive comorbidities.
This project aims to improve epilepsy care by integrating advanced neuroimaging with machine learning to provide more precise diagnostic and treatment strategies, particularly for pediatric patients. By examining a large MRI scan dataset of 800 pediatric epilepsy patients across different age groups treated at Children’s Wisconsin (2010-2023) and 600 normal scans from the Human Connectome Project, the study seeks to identify distinct structural connectivity patterns that predict generalized vs. focal epilepsy and intractable vs. non-intractable cases.
Findings from this study will help radiologists and neurologists to make better-informed decisions by improving imaging interpretation. While benefiting patients in Wisconsin, it also offers global applicability, providing a model for enhancing epilepsy management worldwide, particularly by addressing the unique needs of pediatric patients and advancing personalized epilepsy care.
This approach could be expanded to other neurological conditions, such as genetic disorders and autism, enhancing the research's impact and generating pilot data for future federally funded studies.