Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Simulated Attack: A Network Neuroscience Approach to Understanding Stroke and Language Recovery
Poster Session D, Saturday, September 13, 5:00 - 6:30 pm, Field House
This poster is part of the Sandbox Series.
Apoorva Kelkar1, Harrison Stoll1, Cate Scanlon1, Olu Faseyitan2, Denise Y. Harvey2, Roy H. Hamilton2, H. Branch Coslett2, John D. Medaglia1,2; 1Drexel University, Philadelphia, Pennsylvania, USA, 2University of Pennsylvania, Philadelphia, Pennsylvania, USA
Introduction: Stroke is the leading cause of aphasia affecting around 2 million people in the United States. A key challenge in stroke research is the lack of pre-stroke data. This limits our ability to understand the changes resulting from the stroke. Stroke patients often have a different community organization (i.e., modularity) relative to healthy controls. Critically, modularity is related to stroke outcomes, including language deficits. However, the nature of these modularity changes and their relationship to clinical outcomes is confounded by the lack of premorbid data and difficulty in isolating the contribution of individual regions in natural lesions. To fully comprehend how a stroke changes modularity, we must address two questions:1) how does damage to specific brain regions relate to global shifts in modularity? 2) Do these shifts relate to behavioral deficits? This will allow us to identify what brain regions are responsible for the larger global shifts we see in stroke patients and determine if behavioral deficits are related to focal damage or a by-product of the stroke in its totality. To address these questions, we will simulate focal lesions in a healthy brain and examine how the removal of each brain region relates to modularity changes. We will then use an existing stroke data set to predict behavioral deficit based on our simulated models. Study Design, Planned Analyses, and Preliminary Results: We independently simulated the removal of all regions of the brain from the structural brain networks in 41 healthy controls (mean age: 58.57 ± 10.8 years). Specifically, we represented each subject’s brain as a graph, where brain regions are nodes and the pathways connecting them are edges. We performed this procedure in structural networks with nodes defined by a multiscale anatomical atlas (Lausanne) and a multiscale functional atlas (Schaefer). This allowed us to examine if parcellation choice impacts modularity shifts, and if a novel atlas consensus procedure was a better fit than a single atlas. To date, we have removed each node in a given parcellation to address our first question and calculate the modularity difference between the healthy and lesioned brain in each atlas. Preliminary results suggest that modularity increases in frontal, parietal and temporal lobes and decreases in occipital lobe. In planned Aim 2 we will employ a simulated attack applying naturalistic lesions from real anterior and posterior strokes, commonly associated with aphasia, to healthy controls to estimate modularity shifts and compare the resulting shifts to the simulated results in Aim 1. Planned Aim 3 is to use the simulated modularity change scores from Aim 1 and Aim 2 to predict WAB-AQ sub-tests and naming (via Philadelphia Naming Test) in both anterior and posterior strokes depending on the sub-test, and conceptual semantic knowledge (via Pyramids and Palm Trees) in posterior strokes using support vector regression. Sandbox Participation Benefit: This analysis is in the pilot stages. Sandbox feedback will be instrumental in evaluating our methods for the simulated attack approach and refining our network methods to articulate important questions in clinical aphasiology.
Topic Areas: Computational Approaches,