We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Statistics > Applications

Title: Permutation-based cluster size correction for voxel-based lesion-symptom mapping

Abstract: Voxel-based lesion-symptom mapping (VLSM) is a major method for studying brain-behavior relationships that leverages modern neuroimaging analysis techniques to build on the classic approach of examining the relationship between location of brain damage and cognitive deficits. Testing an association between deficit severity and lesion status in each voxel involves very many individual tests and requires statistical correction for multiple comparisons. Several strategies have been adapted from analysis of functional neuroimaging data, though VLSM faces a more difficult trade-off between avoiding false positives and statistical power (missing true effects). One such strategy is using permutation to non-parametrically determine a null distribution of cluster sizes, which is then used to establish a minimum cluster size threshold. This strategy is intuitively appealing because it respects the necessary spatial contiguity of stroke lesions and connects with the typical cluster-based interpretation of VLSM results. We evaluated this strategy for detecting true lesion-symptom relations using simulated deficit scores based on percent damage to defined brain regions (BA 45 and BA 39) in a sample of 124 individuals with left hemisphere stroke. Even under the most conservative settings tested here, the region identified by VLSM with cluster size correction systematically extended well beyond the true region. As a result, this strategy appears to be effective for ruling out situations with no true lesion-symptom relations, but the spatial contiguity of stroke lesions may cause identified lesion-symptom relations to extend beyond their true regions.
Comments: 9 pages, 5 figures
Subjects: Applications (stat.AP); Quantitative Methods (q-bio.QM); Methodology (stat.ME)
Cite as: arXiv:1606.00475 [stat.AP]
  (or arXiv:1606.00475v1 [stat.AP] for this version)

Submission history

From: Daniel Mirman [view email]
[v1] Wed, 1 Jun 2016 21:28:53 GMT (418kb,D)

Link back to: arXiv, form interface, contact.