Predicting Rehabilitation Responses in Persons with Aphasia Undergoing Language Therapy Using Baseline Whole-Brain Task fMRI in a Multivariate Method 公开

Abstract

Background: With the diversity in aphasia profiles as well as a small window of plasticity, it is imperative to deliver the most effective rehabilitation treatment to patients soon after their incidence of stroke. Voxel-based lesion-symptom mapping (VLSM) is a widely used approach for identifying predictors of treatment outcomes, but the lack of network-level information in VLSM may omit the holistic details needed to create valid language-network prediction models. As task-fMRI offers neurobiological inspection of lesion impact on the whole language network, this study aims to develop a multivariate analysis inclusive of whole-brain task-fMRI data to more holistically predict treatment outcomes in persons with aphasia (PWA) undergoing language therapy.

 

Methods: This thesis will be a correlational study focused on analyzing the relationship between patterns of pretreatment brain activity across multiple PWA and treatment-induced language improvements over time to predict post-treatment outcomes. Initially, multivariate methods will be optimized to handle whole-brain task fMRI data by inserting a customized mask representative of the whole-brain as well as testing a series of k-fold cross-validations in order to find the optimal sparseness value. This will then be tested systematically for reliability, extended utility, and overall sensitivity.

 

Results: Multivariate optimizations included the modification of k-fold CVs, brain masks, and lesion size correction. With these changes, the multivariate analysis was able to feasibly predict post-treatment outcomes at both whole-level and group-level analyses. Additionally, predictive biomarkers such as the right extrastriate area and right fusiform gyrus were discovered during brain-behavior analysis of semantic fluency. When analyzing the extended utility of multivariate methods, the left medial frontal gyrus was found to predict post-treatment responses to more sentence-level language functions such as grammar.

 

Conclusion: Ultimately, the goal of this study is to create an optimized prediction model that is better able to capture network-level information for functions such as language. The use of such information would help clinicians predict rehabilitation responses in PWA to develop a more effective rehabilitation regimen for aphasia patients. 

Table of Contents

Introduction……………………………………………………………………...........................................………………….1

      Aphasia………………………………………………………………………………................................................…....1

      Diversity of Language Treatments Available………………………………...…..........................................……...3

      Imaging-Based Prediction Models for Aphasia………………………...................................…....………………...4

Hypothesis……………………………………………………………………………………………………………..…………………7

Methods…………………………………………………………………………………………………………………..………………8

      Participants……………………………………………………………………………………………………....………………...8

      fMRI Acquisition…………………………………………………………………..……...……………....………………………9

      Task Design………………………………………………………………………………………………....……………..……...10

      Treatment…………………………………………………………………………………………………....………….………...10

      Behavioral Assessment.……………………………………………………………………………....……..………………….12

      Image Processing………………………………………………………………………………....……………………………...14

      Multivariate Analysis……………………………………………………………………………....……………………………16

      Mass Univariate Analysis…………………………………………..……………………....…………………………………..25

      Correcting for Lesion Size…………………………………………..…………………....………………………….…………26

Results……………………………………………………………………………………………………………...……………………28

      Aim 1: To develop and optimize multivariate based predictors in order to analyze significant

      relationships between whole-brain task fMRI and post-treatment behavioral data.....……….....……………...28

      Aim 2: To assess the reliability of multivariate analysis by analyzing how well it converges with

      conventionally used predictors such as mass univariate analysis…………………………………....…….....….....29

      Aim 3: To validate both multivariate and mass univariate analyses using another behavioral measure........31

      Aim 4: To sensitize the multivariate method to predict treatment-specific language changes.....................34

      Correcting for lesion size is feasible for both multivariate and mass univariate analysis.............................35

Discussion……………………………………………………………………………………………………..………………………..38

      Multivariate analysis can predict post-treatment responses to CEG from whole-brain task-fMRI data.........39

      Multivariate analysis can predict network-level language rehabilitation in response to treatment…….....….41

      Multivariate analysis is sensitive in detecting treatment-induced changes………………………....….....……...41

      Limitations and Future Works………………………………….....………………………………………………….....…...44

      Conclusion…………………………………………………………….…………………………………………….....………….45

Tables and Figures

      Table 1. Participant Demographics…………………………………………………………………....……………………….9

      Figure 1. Intention Treatment Visual………………………………………………………………....…………..…………12

      Figure 2. Treatment and Behavioral Assessment Timeline………………………………………....……………..……14

      Figure 3. Local vs. Distributed Sparseness……………………………………………………………....…….……………17

      Figure 4. LESYMAP Optimal Sparseness Computation…………………………………………....…………….……...18

      Figure 5. LESYMAP Statistical Image Computation………………………………………………....……………………19

      Figure 6. K-fold Settings……………………………………………………………………..………....………………………20

      Figure 7. 14-fold CV Results for CEG.………………………………………………………………....…………………20-21

      Figure 8. Brain Mask Optimizations……………………………………………………………....……………..…………..23

      Figure 9. LESYMAP Inputs Visual…………………………………………..…………………....……………….………….24

      Figure 10. Different Brain Mask Results………..…………………………….……………....…………………….………29

      Figure 11. Mass Univariate and Multivariate Spatial Overlap for CEG predictors….…....…………………...…..30

      Figure 12. CEG ROIs for brain-behavior relationships……………………………………....……………….………….31

      Figure 13. Mass Univariate and Multivariate Spatial Overlap for GRAM predictors…....………………….........32

      Figure 14. GRAM ROI for brain-behavior relationships………………………………………………....……………....32

      Figure 15. 14-fold CV results for GRAM………………………………………………………………………....…………..33

      Figure 16. Group-level analysis results…………………………………………………………………....…………....34-35

      Figure 17. Lesion Size Correction results…………………………………………………………………....……..…..36-37

References……………………………………………………………………………………………………………………..………..46

About this Honors Thesis

Rights statement

• Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.

School	Emory College
Department	Neuroscience and Behavioral Biology
Degree	B.S.
Submission	Honors Thesis
Language	English
Research Field	Health Sciences, Speech Pathology Health Sciences, Rehabilitation and Therapy Biology, Neuroscience
关键词	Predictive Biomarkers Aphasia Neuroimaging
Committee Chair / Thesis Advisor	Krishnamurthy, Venkatagiri, Emory University
Committee Members	Murugan, Malavika, Emory University Rodriguez, Amy, Emory University

最新修改

Primary PDF

Thumbnail	Title	Date Uploaded	Actions
	Predicting Rehabilitation Responses in Persons with Aphasia Undergoing Language Therapy Using Baseline Whole-Brain Task fMRI in a Multivariate Method ()	2022-04-11 19:51:19 -0400	Press to Select an action Download

Supplemental Files

Thumbnail	Title	Date Uploaded	Actions