NIH-funded Project on Pediatric Epilepsy Surgery
Title
Novel DWI Methods to Minimize Postoperative Deficits in Pediatric Epilepsy Surgery Narrative
This project investigates a novel tool of benefit-risk analysis for the presurgical evaluation of pediatric drug-resistant focal epilepsy which utilizes the state-of-the-art deep convolutional neural network (DCNN)-based tract classification to provide an optimal resection boundary of the epileptogenic area resulting in the maximized benefit (seizure freedom) and minimized risk (functional deficit including motor, language, hearing, and vision). We will also study if the integration of DCNN and DWI connectome helps decide timely surgery by providing preoperative imaging markers underlying a high likelihood of postoperative neurocognitive improvements and mechanistic insight in structural brain reorganization associated with postoperative verbal IQ improvement. The results of this project will translate deep learning-based diffusion MRI techniques to optimize the surgical margin, predict the postoperative neurocognitive outcome, and determine the specific mechanism of postoperative brain reorganization, which will be validated for optimizing clinical benefit-risk analysis before surgical intervention.
Sponsor Name: National Institute of Neurological Disorders and Stroke - NINDS (2R01NS089659)
Recent findings
Chronic seizures are often associated with impaired language functions in children, but these issues may improve shortly after successful epilepsy surgery. This study aims to develop an innovative deep convolutional neural network (DCNN)-based tract classification to enhance the prediction of short-term postoperative language improvement using axonal connectivity markers derived from specific language modular networks (LMNs) within the preoperative whole-brain diffusion-weighted imaging connectome (wDWIC). To achieve this, we employed a three-step approach. First, our previous DCNN-based tract classification to detect true-positive eloquent tracts was extended using an open-source database of high-quality wDWIC to facilitate the accurate classification of true-positive tracts within the preoperative backbone wDWIC of individual patients. Next, we applied psychometry-driven DWIC analysis to the resulting DCNN-based backbone wDWIC in order to create core, expressive, and receptive LMNs. Finally, graph and circuit theory-based connectivity markers were assessed within the three LMNs and compared using a series of machine learning algorithms to predict the presence of postoperative language improvement from a given LMN. The results showed that the extended DCNN tract classification significantly improved the reproducibility of connectivity markers by up to 35.5% of F-statistics across different LMNs. The prediction accuracy increased by up to 40% across different machine learning algorithms. Notably, the best algorithm achieved the accuracy of 94%/94%/96% to predict the presence of language improvement about two months after surgery in core/expressive/receptive domain of an independent validation cohort. These findings hold great potential to assist physicians in identifying candidates whose language skills stand to benefit from early surgery.
A. Results of feature importance analysis. Radius of blue-green sphere denotes the importance weight of the current flow betweeness (CFB) marker for correct prediction (e.g., the larger the more important). B. Z-scores of the CFB marker in the postoperative language improvement group. C. Z-scores of the CFB marker in the no postoperative language improvement group. Radius of red-yellow sphere denotes the absolute Z-score of the CFB marker averaged in each group. (e.g., the larger the more deviation of the CFB marker from that of age-matched healthy control group).
The low Z-scores in the postoperative language improvement group lend strong support to our primary hypothesis: "a child who demonstrates short-term postoperative language improvement exhibits a preoperative imaging marker that signifies the preservation of axonal connectivity within the preoperative language network, even though its functioning is restricted by concurrent seizures.", leading us to conclude that a child who exhibits short-term postoperative language improvement possesses a preoperative imaging marker indicative of preserved axonal connectivity within the preoperative language network. As a result, the CFB values associated with these hub nodes can be employed as valuable markers to reliably predict short-term language improvement following surgery.