Event Abstract

Back to Event

Cluster correction and graphical user interface for support vector regression lesion-symptom mapping

Andrew T. DeMarco1* and Peter E. Turkeltaub1
  • 1 Georgetown University, Department of Neurology, United States

Voxel-based lesion-symptom mapping (VLSM) has become a commonly-used method for localizing cognitive function in the brain (Bates et al., 2003). The method relates performance on a behavioral measure to the status of each voxel in the brain as lesioned or spared across a group of individuals. Voxels where lesioned status covaries with performance are inferred to relate to the cognitive functions relied upon by the behavioral measure under investigation. Traditional VLSM operates on a mass-univariate basis, assuming statistical independence across voxels. However, this assumption is unwarranted given both the spatial autocorrelation across voxels inherent to neuroimaging data and the non-random distribution of lesions (Mah et al., 2014; Husain & Nachev, 2007). Indeed, large-scale simulations have demonstrated that this assumption can lead traditional mass-univariate VLSM analyses to produce biased lesion-symptom maps. The same work has highlighted the potential for multivariate inference methods to reduce or remove this source of error (Mah et al., 2014). By considering the contribution of many voxels simultaneously, multivariate methods circumvent the assumption of statistical independence across voxels. Recently, one such multivariate method, support vector regression, has been used to outperform traditional VLSM for detecting both synthetic and real lesion-symptom relationships (SVR-LSM; Zhang et al., 2014). Despite the potential to conduct lesion-symptom analyses with increased sensitivity and specificity, multivariate methods have yet to be widely adopted. Naturally, one practical barrier limiting the adoption of multivariate methods is the easy of use of software utilizing the method. Although the authors of SVR-LSM make their algorithm available as a collection of MATLAB scripts, limitations in usability mirror the novelty of the approach. There are three limitations of the current implementation of SVR-LSM which may hinder adoption of the technique more broadly. First, the software requires that users have some minimal familiarity with MATLAB scripting. Analyses are configured by manually editing a MATLAB script file such that relevant variable values reflect a user’s analysis choices. Second, users are required to compile a third-party package for SVR functionality (libSVM; Chang & Lin, 2011). Although this process is easy in some computing environments, it can be burdensome in others due to variable support of compilers across operating systems and MATLAB versions. Third, the current implementation of SVR-LSM does not offer family-wise error correction at the cluster-level, a topic which has recently garnered much critical attention in functional neuroimaging analysis (Eklund et al., 2016). Although SVR-LSM supports a minimum cluster size threshold, this value is arbitrarily provided by the user. We attempt to facilitate the adoption of SVR-LSM as a multivariate alternative to traditional VLSM by addressing these limitations in the existing software implementation. Specifically, we implement cluster-level family-wise error correction using permutation testing, provide a graphic interface wrapper, and add the option of using support vector regression functionality from MATLAB’s Statistics Toolbox. These modifications allow someone with no programming knowledge to configure and conduct fully-corrected analyses out-of-the box. We hope that these improvements will allow the technique to be utilized more broadly. We demonstrate these capabilities in a cohort of individuals with post-stroke aphasia.

Figure 1
Image

References

Bates, E., Wilson, S.M., Saygin, A., Dick, F., Sereno, M.I., Knight, R.T., & Dronkers, N.F. (2003).
Voxel-based lesion-symptom mapping. Nature Neuroscience, 6, 448-450. doi:
10.1038/nn1050

Chang C-C., & Lin, C-J. (2011). LIBSVM: a library for support vector machines. ACM
Transactions on Intelligent Systems and Technology, 2(3). doi:
10.1145/1961189.1961199

Eklund, A., Nichols, T.E., & Knutsson, H. (2016). Cluster failure: Why fMRI inferences for spatial
extent have inflated false-positive rates. Proceedings of the National Academy of
Sciences of the United States of America, 113(28), 7900-7905. doi:
10.1073/pnas.1602413113

Husain M, & Nachev P. (2007). Space and the parietal cortex. Trends in Cognitive Science,
11(1), 30-6. doi: 10.1016/j.tics.2006.10.011

Mah, Y-H., Husain, M., Rees, G., & Nachev, P. (2014). Human brain lesion-deficit inference
remapped. Brain, 137(Pt 9), 2522-31. doi: 10.1093/brain/awu164.

Zhang,1 Y., Kimberg, D., Coslett, B., Schwartz, M.F., & Wang, Z. (2014). Multivariate
lesion-symptom mapping using support vector regression. Human Brain Mapping,
35(12), 5861-76. doi: 10.1002/hbm.22590

Keywords: lesion-symptom mapping, multivariate methods, VLSM, lesion mapping, Aphasia, cluster correction

Conference: Academy of Aphasia 55th Annual Meeting , Baltimore, United States, 5 Nov - 7 Nov, 2017.

Presentation Type: poster or oral

Topic: Consider for student award

Citation: DeMarco AT and Turkeltaub PE (2019). Cluster correction and graphical user interface for support vector regression lesion-symptom mapping. Conference Abstract: Academy of Aphasia 55th Annual Meeting . doi: 10.3389/conf.fnhum.2017.223.00069

Copyright: The abstracts in this collection have not been subject to any Frontiers peer review or checks, and are not endorsed by Frontiers. They are made available through the Frontiers publishing platform as a service to conference organizers and presenters.

The copyright in the individual abstracts is owned by the author of each abstract or his/her employer unless otherwise stated.

Each abstract, as well as the collection of abstracts, are published under a Creative Commons CC-BY 4.0 (attribution) licence (https://creativecommons.org/licenses/by/4.0/) and may thus be reproduced, translated, adapted and be the subject of derivative works provided the authors and Frontiers are attributed.

For Frontiers’ terms and conditions please see https://www.frontiersin.org/legal/terms-and-conditions.

Received: 21 Apr 2017; Published Online: 25 Jan 2019.

* Correspondence: Dr. Andrew T DeMarco, Georgetown University, Department of Neurology, Washington, District of Columbia, 20007, United States, demarco@email.arizona.edu