Improved accuracy of lesion to symptom mapping with multivariate sparse canonical correlations

Abstract

Lesion to symptom mapping (LSM) is a crucial tool for understanding the causality of brain-behavior relationships. The analyses are typically performed by applying statistical methods on individual brain voxels (VLSM), a method called the mass-univariate approach.Several authors have shown that VLSM suffers from limitations that may decrease the accuracy and reliability of the findings, and have proposed the use of multivariate methods to overcome these limitations. In this study, we propose a multivariate optimization technique known as sparse canonical correlation analysis for neuroimaging (SCCAN) for lesion to symptom mapping. To validate the method and compare it with mass-univariate results, we used data from 131 patients with chronic stroke lesions in the territory of the middle cerebral artery, and created synthetic behavioral scores based on the lesion load of 93 brain regions (putative functional units). LSM analyses were performed with univariate VLSM or SCCAN, and the accuracy of the two methods was compared in terms of both overlap and and displacement from the simulated functional areas. Overall, SCCAN produced more accurate results -higher dice overlap and smaller average displacement -compared to VLSM. This advantage persisted at different sample sizes (N=20-131) and different multiple comparison corrections (false discovery rate, FDR; Bonferroni; permutation-based family wise error rate, FWER). These findings were replicated with a fully automated SCCAN routine that relied on cross-validated predictive accuracy to find the optimal sparseness value. Simulations of one, two, and three brain regions 5, 2017; of aphasia severity (aphasia quotient and picture naming) showed that SCCAN could accurately identify known language-critical areas, while VLSM either produced diffuse maps (FDR correction) or few scattered voxels (FWER correction). Overall, this study shows that a multivariate method, such as, SCCAN, outperforms VLSM in a number of scenarios, including functional dependency on single or multiple areas, different sample sizes, different multi-area combinations, and different thresholding mechanisms (FWER, Bonferroni, FDR). These results support previous claims that multivariate methods are in general more accurate than mass-univariate approaches, and should be preferred over traditional VLSM approaches. All the methods described in this study are available in the newly developed LESYMAP package for R.peer-reviewed) is the author/funder. It is made available under a The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/149260 doi: bioRxiv preprint first posted online Jul.

Enhanced estimations of post‐stroke aphasia severity using stacked multimodal predictions

“…These 12 variables were submitted to a final RF model to create a single final stacked multimodal prediction (STAMP, or “Final_All”). Age and post‐stroke duration were not included as predictors because earlier work on the same data showed no appreciable benefit [Pustina et al, ]. Beside the “Final_All” prediction, another prediction was obtained by running RFE on the 12 predictors and selecting the most useful ones (“Final_RFE”).…”

Thalamic strokes that severely impair arousal extend into the brainstem

“…This approach has several advantages compared to mass univariate (ie, "voxelwise") lesionsymptom mapping. 18 Statistical significance of the overall model was determined, and each voxel received a weight between 0 and 1, with higher values associated with greater impairment in arousal. Resulting voxel weights were color-coded and overlaid onto the template brain for display.…”

“…Statistical analysis was limited to voxels in which 3 or more patients had damage, excluding areas with sparse coverage with less than 10% of the cohort, as performed previously. 18…”

Dissociation between Frontal and Temporal-Parietal Contributions to Connected Speech in Acute Stroke

“…Here, we increased power to assess whether damage to a region significantly affected behavior in comparison to no damage to that region, by providing sufficient behavioral variability and heterogeneity for both between-and within-region lesion distributions (cf. Kimberg et al, 2007;Sperber and Karnath, 2017;Lorca-Puls et al, 2018;Pustina et al, 2018;Sperber et al, 2019).…”

“…Lastly, we applied support vector regression (SVR) based multivariate LBM which considers the pattern of all voxels as a single model to predict a behavioral outcome. In comparison to univariate approaches, multivariate LBM ameliorates limitations from differential lesion distribution across voxels, Type II error from applying statistical corrections across voxels and allows for interactions between different damaged areas to account for behavior (Mah et al, 2014;Zhang et al, 2014;Yourganov et al, 2016;Pustina et al, 2018;Sperber et al, 2019).…”