Using Static and Dynamic connectivity analyses of Eloquent brain tissue to assess surgical risk in patients with intractable seizures

Clinical Center

Investigators

Abstract

We have completed the review of 41 healthy individuals who were recruited as part of the epilepsy seizure protocol and completed the same analyses on a larger cohort on 205 healthy volunteers using the Human Connectome Project's publicly available datasets. We have now completed statistical analyses of the static and dynamic connectivity for both task related and resting state auditory language related networks for 41 healthy control subjects, recruited via NINDS epilepsy protocol, controlled for the level difficulty and who are matched for educational level and age. We have also confirmed our preliminary findings using the static and dynamic connectivity network analyses for 205 healthy adults from the HCP database. We have tested 4 different computational models as to their ability to predict functional connectivity changes related to cognitive mental states. We have found: 1. We also found gender and age interaction effects such that gender effects were more pronounced in younger subjects that become less salient with advancing age. We interpret these results to indicate that age is the driving factor in the performance of language related brain networks over time. These results were successfully presented at the Organization of Human Brain Mapping Conference in June 2022. We are confirming these findings using a larger cohort using publicly available datasets that are IRB exempt. 2. We have completed the preliminary analyses of the dynamic functional connectivity and found differences in the dynamic performance of neural networks in rest vs task related conditions. We conclude that, dynamic FC is not interchangeable even though spatial region of interests may be. We hypothesize that the differences in dynamic FC may be important in determining patient clinical outcomes and language recovery following CNS injury. This work will be presented at the American Speech and Hearing Association in Nov 2023. The manuscript associated with this work is in the advanced stages of internal review and will be submitted for Journal review shortly. 3. We have assessed the performance of 4 different types of computational models in their ability to predict functional connectivity changes that occur during an auditory language task. We identified that a particular type of computational model outperformed the other 3 three by 18% without overgeneralizing when given new subject data. This work was successfully presented at the Organization for Human Brain Mapping in Montreal in July 2023. The manuscript associated with this work has cleared internal review and is being prepared for IC clearance and then submitted for Journal review. Future work: We will be using the computational models to run machine learning based simulations of the predicted performance of 4 patients who have undergone surgery for intractable epilepsy. Significance: We think that machine learning based simulations can be used to better understand functional neuronal recovery following CNS surgery and other injuries and or interventions.