Welcome to SNL 2021
Tuesday, October 5, 3:30 - 5:00 pm PDTLog In to set timezone
Natalie Hetherington1, Argye, E. Hillis2, Lisa Bunker2, Roger Newman-Norlund1, Chris Rorden1, Leo Bonilha3, Erin Meier2,4, Emily Goldberg2,5, Gregory S. Hickok6, Gregori Yourganov7, Julius Fridriksson1; 1University of South Carolina, 2Johns Hopkins University, 3Medical University of South Carolina, 4Northeastern University, 5University of Pittsburgh, 6University of California Irvine, 7Clemson University
It is widely accepted that the lesion is among the strongest predictors of stroke aphasia severity. This appears to hold true in both acute and chronic recovery. However, there are arguments for using data from acute patients over chronic, and vice versa. Aphasia severity is often thought to be more stable in the chronic phase, and therefore may be the optimal time to investigate these relationships. Conversely, others argue that therapy and recovery-related morphological changes influence chronic behavior, therefore the acute stage is more appropriate. Despite this, to our knowledge there are no comprehensive studies that provide strong evidence in favor of one or the other. To that end, we investigated two aims; (1) whether the relationship between lesion load and behavioral impairment is similar in acute and chronic patients, and (2) if model based on acute data accurately estimates chronic data, and vice versa. Lesions and WAB-AQ scores from acute (N=63) and chronic (N=109) left hemisphere stroke survivors with aphasia were entered into separate univariate voxel-based lesion-symptom mapping (VLSM) analyses using the NiiStat toolbox for Matlab (https://www.nitrc.org/projects/niistat/). A support vector regression (SVR) model was trained on lesion data (proportion of ROI damage) from either the acute or chronic dataset; the coefficients were calculated for each region, as well as a constant term so that a sum of weighted percent damage for all regions plus the offset, gives an estimate of WAB-AQ. Predictions of WAB-AQ score were computed using linear-kernel SVR. Four analyses were conducted: 1) Trained on acute data, leave-one-out prediction of behavior 2) Trained on chronic data, leave-one-out prediction of behavior 3) Trained on acute data, predicting chronic behavior 4) Trained on chronic data, predicting acute behavior VLSM analyses revealed that similar, but not identical, regions in the acute and chronic samples were predictive of WAB-AQ. In both datasets, lesions in the superior temporal gyrus (STG), posterior STG, STG pole, middle temporal gyrus (MTG), posterior MTG, inferior temporal gyrus, inferior and middle occipital gyri, angular gyrus, superior longitudinal fasciculus, supramarginal gyrus, pre- and postcentral gyri, and the corona radiata were significantly associated with reduced WAB-AQ scores. A further 19 regions were predictors in the chronic dataset. Positive correlations were found between actual and predicted WAB-AQ scores: predicting acute using leave-one-out (r = .7221, p < 0.001), chronic using leave-one-out (r = .5205, p < 0.001), training on chronic, predicting acute (r = .5568, p < 0.001), and training on acute, predicting chronic (r = .7220, p < 0.001). Our findings suggest that brain-behavior relationships in both stages of stroke are comparable, and that data acquired at one timepoint may be used to estimate the other. This has important implications for clinicians, as it suggests that models based on acute data may also give us good insight into behavior in the chronic stage. Differences between critical brain regions in acute and chronic patients may give us further understanding into recovery patterns and/or regions likely to have further degeneration, however this should be tested more formally with longitudinal studies.
Maria Ivanova1, Francois Rheault2, Nina Dronkers1,3; 1University of California, Berkeley, California, USA, 2Vanderbilt University, Nashville, USA, 3University of California, Davis, California, USA
Significant discrepancies exist across diffusion MRI studies regarding the contributions made by major fiber pathways, particularly as they relate to language pathology (Ivanova et al., in press; Vaidya et al., 2019). Here we investigate the method-specific impact of different tractography algorithms on observed patterns of tract-behavior associations and present a new approach for evaluating the role of different tracts in language processing. By simultaneously investigating four different iterative tractography algorithms and analytically combining their results, we provide a more definitive answer on the differential role that white matter pathways play in language processing. We investigated language abilities in 33 individuals with post-stroke aphasia using the Western Aphasia Battery (Kertezs, 2007). Structural and diffusion-weighted MRIs (64 directions, 2mm isovoxel, b=2000s/mm2) were performed on a Siemens Verio 3T scanner. The diffusion MRI data processing was done with the tractoflow pipeline, local models, i.e., tensor and fiber orientation distribution function (fODF), the tractography as well as the automatic bundle segmentation using Recobundles with multiple execution followed by labels fusion was performed using the Dipy library (Garyfallidis et al., 2014). We explored the following tractography algorithms: (1) single-peak tractography based on the tensor; (2) multi-peaks eudx tractography based on the fODF; (3) deterministic, and (4) probabilistic tractography based on the fODF. For each tracking algorithm main associative fibers (ILF, IFOF, UF, FAT and 3 segments of the AF) were automatically segmented from the whole brain tractogram in both hemispheres. As expected, discrepancies between the four algorithms were readily observed on tract reconstructions and occurrence heatmaps. Tensor-based tractography provided the smallest sized tracts and probabilistic tractography led to the largest reconstructions, with tracts generated with eudx and deterministic tractography falling in between. Then the volume of each bundle normalized by the hemisphere volume and mean fractional anisotropy (FA) values obtained for each tracking algorithm were correlated with language measures while accounting for lesion volume. Some tract-language associations were highly method-specific (observed only for 1 -2 algorithms), while a select number of patterns were all significant, irrespective of the tracking method employed. Next, these significant correlations were combined to determine which were observed systematically across the different algorithms, with correlations appearing in at least 3 out of 4 methods deemed reliable and amenable to further interpretation. Accordingly, the AF posterior branch was critical for naming and repetition, and the ILF for naming. FA analyses confirmed and expanded these findings, additionally implicating the IFOF and the long branch of the AF stemming from the middle temporal gyrus in lexical-semantic processing, and the anterior branch of the AF in spontaneous speech. The current study highlights method-specific variability in the establishment of tract-behavior relationships. The proposed multi-method approach helps to robustly outline the crucial language connections that need to be represented in future models of language processing and further explored as predictors of clinical outcomes. The proposed method of analysis is highly replicable and reliable. It should be used in future studies to minimize the impact of algorithm-specific artifacts and help to solidify and verify detected brain-language relationships.
Alexandra C. Brito1, Deborah F. Levy1, Sarah M. Schneck1, Jillian L. Entrup1, Caitlin Onuscheck1, Marianne Casilio1, Larry Davis1, Stephen M. Wilson1; 1Vanderbilt University Medical Center
Individuals with post-stroke aphasia often recover language function over time. This recovery is thought to be driven by neural plasticity, which may be related to the health of surviving brain regions. Brain health is associated with observable changes on neuroimaging, including atrophy and leukoaraiosis (small-vessel cerebrovascular disease that appears as white matter hyperintensities on MRI) . Prior work has suggested that leukoaraiosis severity may be a clinically relevant predictor of recovery from aphasia after stroke [2-4], although findings have been inconsistent . This study investigates the relationship between leukoaraiosis and language recovery as measured by repeated aphasia assessments during the first year after stroke. In the context of a larger project on the neural correlates of recovery from aphasia in the first year after stroke, we recruited 269 patients with left-hemispheric stroke and coincident MRI (237 ischemic, 32 hemorrhagic; age: 63.0 ± 14.2 years). Leukoaraiosis was evaluated based on Fluid-Attenuated Inversion Recovery (FLAIR) images using the Fazekas visual rating scale , which ranges from 0 (absent) to 6 (most severe). Patients were evaluated for aphasia within 2–5 days of stroke using the Quick Aphasia Battery (QAB) . For patients with acute aphasia (173 patients), follow-up aphasia assessments were performed at 1 month, 3 months, and 1 year following stroke, whenever possible. Multivariable regression models were fitted to the data to determine the relationships between leukoaraiosis and QAB scores at each time point, as well as changes in QAB scores over time. Covariates of age, sex, handedness, education, and stroke lesion extent were included in the models. We found that 90% of patients had some degree of leukoaraiosis. The mean score on the Fazekas scale was 2.7 ± 1.8 points. Leukoaraiosis severity was positively associated with age, female sex, and cerebrovascular risk factors such as hypertension and tobacco use. Multivariable regression analyses showed that leukoaraiosis did not predict initial QAB scores (n = 269, t = –0.92, p = 0.36), follow-up QAB scores at 1 month (n = 80, t = 0.50, p = 0.62), 3 months (n = 77, t = 0.36, p = 0.72), or 1 year (n = 54, t = –0.35, p = 0.73), or changes in QAB scores between any pairs of time points (all p > 0.50). In sum, we did not observe any relationships between leukoaraiosis severity and aphasia outcomes or change in language function over time. This negative finding contrasts with other domains such as cognition and motor control in which leukoaraiosis is predictive of deficits . The lack of association between leukoaraiosis and language function may reflect the anatomical distribution of small-vessel cerebrovascular disease, which is largely medial and dorsal to critical language tracts.  Smith. Stroke 2010;41:S139–43.  Wright et al. Neurology 2018;91:e526–32.  Basilakos et al. Neurorehabil Neural Repair 2019;33:718–29.  Varkanitsa et al. Neurorehabil Neural Repair 2020;34:945–53.  Hatier et al. Ann Phys Rehabil Med 2019;61:e48.  Fazekas et al. AJR Am J Roentgenol 1987;149:351–6.  Wilson et al. PLoS One 2018;13:e0192773.  Longstreth et al. Stroke 1996;27:1274–82.
Junhua Ding1, Tatiana Schnur1; 1Baylor College of Medicine
Introduction: Left hemisphere (LH) stroke impacts connected speech to different degrees but the neural predictors of differential recovery from the acute stage of stroke are unknown. Damage to disparate brain regions and their connections impairs connected speech (Alyahya et al., 2020) and white-matter tracts’ acute damage has been found associated with the recovery of naming ability (Hillis et al., 2018). To examine the role of white-matter tract integrity in connected speech recovery, we assessed changes in lexical-syntactic aspects of connected speech during the first year after acute LH stroke while measuring the extent of acute lesions on white-matter tracts. Methods: We tested 41 individuals with LH stroke from the acute (within an average 4-days post-stroke) to subacute (2-mo, n=32) and chronic stages of stroke (6-mo, n=31; 12-mo, n=25). We elicited connected speech using story-telling and quantified speech using quantitative production analysis (Rochon et al., 2000). Following Ding et al. (2020), we extracted four connected speech components, including fluency, syntax, lexical selection and structural complexity. At the group level, acute and follow-up component scores were compared using paired t-tests. At the individual level, impairment was defined as below -1.67 S.D. of 13 age-and education-matched controls. Acute white-matter tract damage was calculated by the intersection between patients’ lesion masks and the Rojkova et al. (2015) white-matter atlas. Dorsal tracts included anterior, long and posterior segments of the arcuate fasciculus (AF) and the frontal aslant tract (FAT). Ventral tracts included inferior frontal-occipital, inferior longitudinal and uncinate fasciculi (UF). We conducted a lesion-symptom mapping analysis to determine the difference in recovery (follow-up – acute) between tract-damaged and -preserved groups (lesion > 100mm3) while controlling for lesion volume and acute performance. Results: Behavior. At the group level, fluency improved significantly by the subacute stage (p<0.02). Syntactic ability marginally improved by 6-mo (p=0.08). Lexical selection significantly improved after 12-mo (p=0.01). No significant improvement occurred for structural complexity (p’s>0.46). At the individual level across time points, the ratio of patients who recovered to within control performance from impaired acute performance for syntax was above 80%, but less so for fluency (50%-61%). Ratios for lexical selection and structural complexity were 67% subacutely and reached above 80% post 12-mo. Acute white matter tract damage. Subacutely, acute FAT damage related with diminished recovery of fluency and syntax (p’s<0.03). Damage to UF marginally related with diminished recovery of lexical selection (p=0.07). Chronically, acute FAT damage related with diminished recovery of structural complexity (p’s<0.006). Damage to long and posterior segments of AF related with diminished recovery of syntax at 6- and 12-mo (p’s<0.01), respectively. Summary: Connected speech showed different degrees of recovery across follow-up stages where fluency recovered first but only half of patients recovered to within control performance, while syntax and lexical selection recovered later, but most participants recovered to control performance levels by 12-mo. Acute white-matter damage predicted connected speech recovery after stroke where dorsal pathway damage related to reduced recovery of fluency, syntax and structural complexity, while damage to ventral pathways reduced lexical selection recovery.