Neural correlates of verbal fluency revealed by longitudinal T1, T2 and FLAIR imaging in stroke

Abstract

Diffusion-weighted imaging has been widely used in the research on post-stroke verbal fluency but acquiring diffusion data is not always clinically feasible. Achieving comparable reliability for detecting brain variables associated with verbal fluency impairments, based on more readily available anatomical, non-diffusion images (T1, T2 and FLAIR), enables clinical practitioners to have complementary neurophysiological information at hand to facilitate diagnosis and treatment of language impairment. Meanwhile, although the predominant focus in the stroke recovery literature has been on cortical contributions to verbal fluency, it remains unclear how subcortical regions and white matter disconnection are related to verbal fluency. Our study thus utilized anatomical scans of ischaemic stroke survivors (n = 121) to identify longitudinal relationships between subcortical volume, white matter tract disconnection, and verbal fluency performance at 3- and 12-months post-stroke. Subcortical grey matter volume was derived from FreeSurfer. We used an indirect probabilistic approach to quantify white matter disconnection in terms of disconnection severity, the proportion of lesioned voxel volume to the total volume of a tract, and disconnection probability, the probability of the overlap between the stroke lesion and a tract. These disconnection variables of each subject were identified based on the disconnectome map of the BCBToolkit. Using a linear mixed multiple regression method with 5-fold cross-validations, we correlated the semantic and phonemic fluency scores with longitudinal measurements of subcortical grey matter volume and 22 bilateral white matter tracts, while controlling for demographic variables (age, sex, handedness and education), total brain volume, lesion volume, and cortical thickness. The results showed that the right subcortical grey matter volume was positively correlated with phonemic fluency averaged over 3 months and 12 months. The finding generalized well on the test data. The disconnection probability of left superior longitudinal fasciculus II and left posterior arcuate fasciculus was negatively associated with semantic fluency only on the training data, but the result aligned with our previous study using diffusion scans in the same clinical population. In sum, our results presented evidence that routinely acquired anatomical scans can serve as a reliable source for deriving neural variables of post-stroke verbal fluency performance. The use of this method might provide an ecologically valid and more readily implementable analysis tool.

Keywords: Anatomical scans; Phonemic fluency; Semantic fluency; Stroke; Subcortical grey matter volume; White matter tract disconnection.

Fig. 1

Lesion prevalence maps. A rendered lesion heat map to the MNI 152 template in the radiological orientation across 121-stroke patients. The maximum value (red) indicates the highest lesion overlap among the participants (N = 18). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 2

Phonemic fluency in relation to time and the right subcortical GM volume. A) The phonemic fluency Z scores at 3 months and 12 months after stroke (blue dots). Individual trajectories of the score changes were represented by the connected green lines between the two time points (data with missing values were not connected). The red line was the averaged trajectory and the grey ribbon represented the standard error. B) The positive linear relationship between the right subcortical GM volume and the phonemic fluency scores. The light red and green lines were the fitted linear regression lines with the correspondent color ribbons representing the 95 % confidence interval. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 3

Semantic fluency in relation to time and the disconnection probability of the left SLF II. A) The semantic fluency Z scores at 3 months and 12 months after stroke (black dots). Individual trajectories of the score changes were represented by the connected blue lines between the two time points (data with missing values were not connected). The red line indicated the averaged trajectory and the grey ribbon represented the standard error. B) The negative linear relationship between the disconnection probability (at the 50 % threshold) of the left SLF II and the semantic fluency scores. The violet and blue lines were the fitted linear regression lines with the correspondent color ribbon representing the 95 % confidence interval. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 4

A) The sagittal, coronal and axial views of the overlap between the left SLF II and the lesions of the subjects above the 50 % disconnection probability at the 3 months post-stroke. The left SLF tract was based on the MNI space template reconstructed from 1065 subjects proposed by Yeh and Tseng (2011) using a multishell diffusion scheme. Both the lesion mask and the SLF tract were superimposed upon the MNI152_T1_0.5 mm.nii anatomical image in Dsi_studio ((https://dsi-studio.labsolver.org). There were 1883 fiber tracts of the left SLF II overlapping with the lesion mask. B) The left SLF related to semantic fluency as reported by Egorova-Brumley et al. (2022).