Thalamic Nuclei and Thalamocortical Pathways After Left Hemispheric Stroke and Their Association with Picture Naming

Abstract

Background: Previous studies utilized lesion-centric approaches to study the role of the thalamus in language. In this study, we tested the hypotheses that non-lesioned dorsomedial and ventral anterior nuclei (DMVAC) and pulvinar lateral posterior nuclei complexes (PLC) of the thalamus and their projections to the left hemisphere show secondary effects of the strokes, and that their microstructural integrity is closely related to language-related functions.

Methods: Subjects with language impairments after a left-hemispheric cortical and/or subcortical, early stroke (n = 31, ≤6 months) or late stroke (n = 30, ≥12 months) sparing thalamus underwent the Boston Naming Test (BNT) and diffusion tensor imaging (DTI). The tissue integrity of DMVAC, PLC, and their cortical projections was quantified with DTI. The right-left asymmetry profiles of these structures were evaluated in relation to the time since stroke. The association between microstructural integrity and BNT score was investigated in relation to stroke chronicity with partial correlation analyses adjusted for confounds.

Results: In both early stroke and late stroke groups, left-sided tracts showed significantly higher mean diffusivities (MDs), which were likely due to Wallerian degeneration. Higher MD values of the cortical projections from the left PLC (r = −0.5, p = 0.005) and DMVAC (r = −0.53, p = 0.002) were correlated with lower BNT score in the late stroke but not early stroke group.

Conclusion: Nonlesioned thalamic nuclei and thalamocortical pathways show rightward lateralization of the microstructural integrity after a left hemispheric stroke, and this pattern is associated with poorer naming.

Impact statement

To the best of our knowledge, our study is the first diffusion tensor imaging study suggesting that the thalamic nuclei and pathways of the left hemisphere spared by direct ischemic insult undergo secondary degeneration over time that is associated with poorer picture naming. Our study may pave the way for targeted interventions such as invasive or noninvasive brain stimulation techniques that engage these spared pathways to prevent secondary degeneration and lead to better outcomes in poststroke aphasia.

Introduction

The thalamus is a fundamental subcortical processing structure, consisting of numerous nuclei communicating with distinct networks, including motor, sensory, limbic, language, and cognitive systems (Blumenfeld, 2010). The dorsomedial and ventral anterior nuclei complex (DMVAC) and the pulvinar lateral posterior nuclei complex (PLC) are believed to be engaged in language-related processes (Crosson, 2019; Nadeau and Crosson, 1997; Stemmer and Whitaker, 2008). The DMVAC projects into frontal cortices, including the prefrontal association cortex and speech motor areas; the PLC projects into posterior parietal–temporal–occipital association cortices as well as the prefrontal cortex (Blumenfeld, 2010; Bohsali et al., 2015; Keser et al., 2018). Although some studies reported that other thalamic nuclei complexes such as ventrolateral complex may potentially be involved in language processing (Gourovitch et al., 2000), our study focused on the more commonly reported thalamic nuclei structures (Crosson, 2019; Nadeau and Crosson, 1997; Stemmer and Whitaker, 2008).

That thalamic lesions can lead to language impairments has been known for a long time (Crosson, 2013; Hillis et al., 2002; Maeshima and Osawa, 2018; Raymer et al., 1997; Schmahmann, 2003; Sebastian et al., 2014; Tuszynski and Petito, 1988). Previous studies suggested that the thalamus plays a role in language processing, and these functions are lateralized to the left thalamus (De Witte et al., 2011; Fritsch et al., 2020). The frequency of aphasia after thalamic stroke in a small study series is reported to be around 10% (Fritsch et al., 2020), with a favorable recovery profile (Crosson, 2013). Functional imaging studies show that the left thalamus appears to be involved in the manipulation of lexical information and is modulated by task demands (Llano, 2013). However, most of the previous stroke literature utilized lesion-centric approaches (e.g., lesion-symptom mapping) to study the role of the thalamus in the aphasia (De Witte et al., 2011; Fritsch et al., 2020; Schmahmann, 2003). More recent studies using advanced imaging techniques reveal that neural dysfunction can be distant from the lesion (diaschisis), and lesion-centric approaches miss the downstream effects of the lesions (Griffis et al., 2019; Price et al., 2017). Diffusion tensor imaging (DTI) is an advanced imaging technique that can reliably quantify the microstructural integrity of brain tissue across different pathologies (Keser et al., 2020; Younes et al., 2019; Yozbatiran et al., 2017).

In this study, we investigated the temporal DTI profiles of the relevant thalamic nuclei complexes, DMVAC and PLC, and their cortical projections in a cohort with language impairments after left hemispheric stroke sparing thalamus. The cohort consisted of two groups of individuals who differed in stroke acuity. We also tested the relation of these thalamic nuclei complexes and thalamocortical connections to naming abilities, which is the most common residual deficit of poststroke aphasia (Goodglass and Wingfield, 1997). We hypothesized that even nonlesioned DMVAC and PLC and their projections to the left-sided cortical areas would show secondary effects of the strokes, possibly more so in the late stroke stage, and that their microstructural integrity would be closely related to picture naming.

Methods

Sixty-one participants with left hemispheric cortical and/or subcortical ischemic infarcts sparing thalamus were enrolled in the study (Table 1). This study consisted of a retrospective analysis of prospectively collected data on subjects with left hemispheric stroke leading to language impairments acutely (determined by either treating neurologist or detailed speech and language assessment) who had both DTI and Boston Naming Test (BNT) data available. Inclusion criteria included: premorbid proficiency in English and 18 years or older. Exclusion criteria were pregnancy, severe claustrophobia, cardiac pacemaker or ferromagnetic implants, prior history of neurological disease affecting the brain other than stroke, known hearing loss, and uncorrected vision loss.

Table 1.

Summary of the Demographics

sID	Sex	Handedness	Age (in years)	Lesion load	Education	BNT	Time since stroke (weeks)	Group	Langsumm_zscores	Lesion location
1	F	R	72	0.82	16	0	1	Early stroke	0.45	AG, IPG
2	M	R	79	1.10	20	3	2	Early stroke	0.58	IFG, AG
3	M	R	84	2.56	16	38	2	Early stroke	0.52	STG
4	M	R	57	2.54	12	10	3	Early stroke	NA	IFG, AG
5	F	R	69	4.57	16	50	3	Early stroke	0.58	AG, IPG
6	M	R	48	0.22	16	36	5	Early stroke	0.66	Cuneus
7	M	R	79	3.60	12	17	6	Early stroke	0.67	IFG
8	M	R	61	6.94	12	53	6	Early stroke	0.66	STG, IPG, AG, OTG
9	F	R	67	1.85	12	1	6	Early stroke	0.82	AG, IPG
10	F	R	65	5.77	20	47	7	Early stroke	−0.26	STG, MTG
11	F	R	67	2.80	16	1	7	Early stroke	0.73	IFG
12	F	R	85	13.03	12	48	8	Early stroke	0.84	Lateral precentral gyrus
13	F	R	83	4.80	13	9	10	Early stroke	0.66	AG
14	M	R	29	0.98	12	28	10	Early stroke	0.66	IFG, STG
15	F	R	70	1.28	12	25	10	Early stroke	0.65	IFG, MFG
16	M	R	48	0.25	16	46	11	Early stroke	0.73	IFG, MFG, SFG
17	M	R	67	2.03	14	28	16	Early stroke	NA	STG
18	M	R	78	2.42	16	50	17	Early stroke	0.73	IFG
19	F	R	54	0.16	12	56	18	Early stroke	0.77	Cuneus
20	F	R	55	0.25	12	58	20	Early stroke	−1.46	ALIC
21	F	R	36	1.77	12	36	22	Early stroke	0.73	STG and MTG
22	M	R	55	0.55	20	58	24	Early stroke	0.66	MeFG and SFG
23	M	R	60	0.67	14	22	24	Early stroke	0.42	STG and IPG
24	M	R	54	8.18	16	32	24	Early stroke	−0.12	Lateral PFC and IFG
25	F	R	64	0.21	18	56	24	Early stroke	−0.80	IFG
26	F	R	41	0.64	12	46	24	Early stroke	−1.91	ALIC and extreme capsule
27	M	R	68	1.71	18	54	24	Early stroke	0.11	Caudate nucleus
28	M	R	55	8.63	18	1	24	Early stroke	0.71	STG and IPG
29	F	R	47	0.03	15	56	24	Early stroke	−0.70	Superior CR
30	F	R	60	0.67	14	46	24	Early stroke	0.61	IFG
31	M	R	66	2.70	14	0	24	Early stroke	−1.96	IFG, MFG
32	M	R	74	4.13	16	2	50	Late stroke	NA	STG, L IPG
33	F	R	36	0.08	14	60	50	Late stroke	−0.49	Anterior CR and putamen
34	F	L	65	0.14	14	58	50	Late stroke	0.77	MeFG
35	F	R	78	7.13	16	0	50	Late stroke	−0.01	STG, IPG, Cuneus
36	M	R	42	0.84	12	60	50	Late stroke	0.33	MFG
37	F	R	33	2.75	12	50	51	Late stroke	−0.54	IFG
38	F	R	41	0.77	18	58	51	Late stroke	−1.67	Anterior CR
39	M	R	56	0.24	16	52	52	Late stroke	0.73	STG
40	M	R	56	3.67	16	56	52	Late stroke	0.52	Cuneus
41	M	R	69	0.62	18	58	52	Late stroke	−0.83	IFG
42	F	R	50	0.10	12	40	52	Late stroke	−0.44	Anterior CR
43	M	R	37	6.27	12	0	52	Late stroke	0.73	IFG, MFG, STG, IPG
44	M	R	65	0.91	14	58	52	Late stroke	−1.41	Posterior CR
45	M	R	53	0.51	18	12	52	Late stroke	0.66	STG
46	M	R	64	1.16	16	13	52	Late stroke	−2.16	Insular
47	M	R	68	3.87	10	46	56	Late stroke	−1.30	STG
48	M	R	60	0.12	12	44	59	Late stroke	0.40	Caudate nucleus
49	F	R	69	0.04	18	58	60	Late stroke	0.83	MFG
50	F	R	66	0.50	9	32	63	Late stroke	0.07	MeFG, CG, STG
51	M	R	79	1.61	18	0	72	Late stroke	0.69	CR
52	M	R	58	4.18	15	10	83	Late stroke	0.45	STG, IPG,IFG
53	M	L	81	1.09	16	52	88	Late stroke	0.13	Hippocampus
54	F	R	69	2.80	12	6	104	Late stroke	0.79	IFG
55	M	R	65	2.84	18	7	108	Late stroke	−0.45	IFG, STG, CR, temporal pole
56	M	R	44	0.35	16	16	112	Late stroke	−0.02	IPG
57	F	R	52	5.82	14	0	116	Late stroke	0.24	STG, IFG, and MFG
58	M	R	67	4.14	10	0	116	Late stroke	−0.71	IFG, MFG, IPG
59	M	R	58	15.20	16	5	216	Late stroke	−0.51	IFG, MFG, STG, IPG
60	F	R	64	2.81	12	2	228	Late stroke	−1.14	Insula, CR
61	M	R	59	8.33	13	10	472	Late stroke	−0.64	IFG,MFG, IPG

ALIC, anterior limb of internal capsule; BNT, Boston Naming Test (0–60); CR, corona radiata; Edu, education in years; FIFG, inferior frontal gyrus; IPG, inferior parietal gyrus; L, left; LangSumm_zcores, zscores from language summary from Boston Diagnostic or Western Aphasia Battery; Lesion load = stroke lesion + white matter lesions percentage to intracranial volume; MeFG, medial frontal gyrus; MFG, middle frontal gyrus; MTG, middle temporal gyrus; NA, not available; R, right; STG, superior temporal gyrus; sID, subject number.

Participants were grouped into early stroke (up to 6 months) (n = 31) and late stroke groups (12 months and beyond) (n = 30). Ethics approval was obtained from the local institutional review board of Johns Hopkins University School of Medicine, and written informed consent was obtained from all the patients. All the subjects underwent behavioral assessment, including the BNT and magnetic resonance imaging (MRI).

Behavioral assessment

The BNT is a commonly used visual confrontation naming test (Kaplan et al., 2001; Mack et al., 1992). The subjects were asked to name line drawing of the common objects (30 or 60 items for subjects who underwent short version, their score was multiplied by 2). The subjects underwent an MRI scan within 2 weeks of their behavioral assessment. In addition to the BNT, we administered an assessment of global language skills to all participants. Due to the partially retrospective nature of the analysis, some participants received the language assessments of the Boston Diagnostic Aphasia Examination, 3rd edition (BDAE-3; Goodglass et al., 2001) (n = 33), whereas others completed Part 1 of the Western Aphasia Battery-Revised (WAB-R; Kertesz, 2007) (n = 25). Similar to our previous work (Meier et al., 2020), we selected equivalent auditory comprehension and verbal expression subtests from each assessment (Table 2) and generated a within-subtest z-score for each individual by using the standard formula: z = (x − μ)/σ, where x is the individual's raw subtest score, μ is the sample's subtest mean, and σ is the sample's subtest standard deviation (SD). We averaged the z-scores to obtain a single measure of global language impairment for each participant. Language deficits in our sample ranged from minimal impairments for some individuals to severe aphasia for others; thus, the language summary z-scores are able to capture the breadth of language impairment that likely occurs after left hemisphere stroke. The z-scores also provide a suitable substitution for other global language measures (e.g., aphasia classification) that were not available for all participants in this study.

Table 2.

Summary of Z Scores from Boston Diagnostic Aphasia Examination-3 and Western Aphasia Battery-Revised Presented as Mean ± Standard Deviation

BDAE-3 subtest	Early stroke z-scores, M ± SD	Late stroke z-scores, M ± SD	WAB-R subtest	Early stroke z-scores, M ± SD	Late stroke z-scores, M ± SD
Word Comprehension	0.12 ± 0.92	2.62 ± 1	Auditory Word Recognition	0.33 ± 1	0.51 ± 0.16
Complex Ideational Material	0.004 ± 1.11	0 ± 1	Auditory Verbal Comprehension	0.26 ± 1	0.89 ± 0.11
Commands	0.20 ± 1.01	0 ± 1	Sequential Commands	−0.13 ± 1	1.17 ± 0.14
Repetition (words and sentences)	−0.01 ± 1.13	−0.89 ± 1	Repetition	−0.12 ± 1	0.98 ± 0.05
Automatized Sequences	0.27 ± 0.94	0.58 ± 1	Sentence Completion	−0.08 ± 1	0.91 ± 0.01
Responsive Speech	−0.03 ± 1.12	−1.41 ± 1	Responsive Speech	−0.07 ± 1.03	1.06 ± 0.01

BDAE-3, Boston Diagnostic Aphasia Examination, 3rd edition; M, mean; WAB-R, Western Aphasia Battery-Revised.

MRI acquisition and preprocessing

Whole-brain MRI data including high-resolution T1-weighted, T2-weighted, Fluid Attenuated Inversion Recovery (FLAIR), and diffusion-weighted imaging (DWI) were acquired on a Philips Achieva 3T scanner by using a SENSE receive 32 channel head coil. Three-dimensional (3D) sagittal-acquired T1-weighted magnetization prepared rapid acquisition with gradient echo had a spatial resolution of 1 × 1 × 1 mm, and the field of view (FOV) was 256 × 256, TR/TE: 2000/3 msec, and flip angle: 8°. T2-weighted spin-echo images were acquired axially with a slice thickness of 2 mm with no gaps, FOV = 212 × 212, TR/TE: 4171/12 msec, and flip angle: 90°. FLAIR images were acquired axially with 5 mm slice thickness and FOV: 512/512, TR/TE: 8000/140 msec, and flip angle: 90°. DWI data were acquired axially with no gaps and a total of 32 diffusion orientations, with an additional b0 with TR/TE: 7013/71 msec, FOV = 212 × 212 mm; the b-value/slice thickness/in-plane resolution was 700 sec/mm²/2.2 mm/0.82 mm, flip angle: 90°. The number of slices was 70, and the scan duration was 4 min 34 sec.

Preprocessing

DWI data files were converted into nifti format by using dcm2nii (nitrc.org/plugins/mwiki/index.php/dcm2nii:MainPage). We used DSI Studio for the remainder of the DWI prepreprocessing. We uploaded Nifti files to DSI Studio (dsistudio.labsolver.org) by opening source images function to create “.src” files (step 1). We opened the “.src” file for reconstruction, inspected the source images for significant movement artifacts, and performed eddy motion correction (step 2). Then, diffusion-weighted images were resampled to 1 mm isotropic voxels and the b-table was checked by an automatic quality control routine to ensure its accuracy (Schilling et al., 2019). Next, we executed reconstruction that includes Eigen analysis on the calculated tensor and generation of indices of fractional anisotropy (FA; μ ± σ) and axial, radial, and mean diffusivities (MDs) ( × 10⁻³ mm² sec⁻¹) (Jiang et al., 2006) (step 3). No lesion masking process was performed during preprocessing. Except what is detailed here, we used the default settings in DSI Studio. For illustration purposes, DWI data were transformed to the T1-weighted space in DSI Studio. No T1-weighted atlas-based segmentation was performed.

Lesion load segmentation

Lesion load included white matter (WM) hyperintensities and infarct volumes. Lesion tracings were quantified manually in MRIcron software (nitrc.org/projects/mricron) by a neurologist experienced in lesion segmentation using high-resolution T2-weighted, FLAIR and T1-weighted images (Keser et al., 2020). Lesions on T1-weighted images were subsequently converted to a common space by using SPM Clinical Toolbox (Rorden et al., 2012). A lesion overlay heat map was calculated from all lesions and superimposed on a template brain by using MRIcron (Rorden et al., 2007) (Fig. 1). For the normalization of the volumes of WM hyperintensities and infarcts, the total intracranial volume (ICV) was estimated by summing all segmented white matter, gray matter, and cerebrospinal fluid (CSF) from each subject by using T1-weighted maps. The lesion load was calculated as the percentage of total lesion volume to ICV (lesion load = lesion volume/ICV × 100).

FIG. 1.

Lesion overlay map from entire cohort (early group first two rows and late group last two rows) is illustrated on T1-weighted maps.

Atlas-based thalamic segmentation

Thalamic nuclei segmentation obtained from the ICBM (International Consortium for Brain Mapping) Deep Nuclei Probabilistic Atlas map was warped by performing a nonlinear registration into single subject space through DSI Studio (Keser et al., 2018; Ying et al., 2017). Dorsomedial and ventral anterior nuclei were merged into a DMVAC region of interest (ROI), and pulvinar and lateral posterior nuclei were merged into a PLC ROI (Crosson, 2019; Nadeau and Crosson, 1997; Stemmer and Whitaker, 2008) (Fig. 2A–C).

FIG. 2.

DMVAC in yellow and PLC in red are illustrated (A–C). DMVAC, dorsomedial and ventral anterior nuclei complex; PLC, pulvinar lateral posterior nuclei complex.

Cerebrospinal fluid contamination was removed from the thalamic nuclei by using the upper threshold radial diffusivities (RDs) of 1.5 × 10⁻³ mm² sec⁻¹ (Metzler-Baddeley et al., 2012). The regions were then edited with neuroanatomical guidance (i.e., moving the volume of interest in 3D space, removing neighboring tissues with smoothing and shaving functions). This semi-automated region segmentation procedure eliminated issues related to full automation.

Tractography

We utilized a brute force and multiple ROIs tractography method and the fiber assignment with continuous tractography algorithm (Wakana et al., 2007).

For deterministic fiber tractography of thalamocortical connections arising from these nucleus complexes, DMVAC cortical projections (Fig. 3A) and PLC cortical projections (Fig. 3B), the initial seed was placed in DMVAC and PLC separately on both sides. Fiber pathways projecting into noncortical regions such as the brainstem and cerebellum were removed with the region of avoidance function to isolate cortical projections.

FIG. 3.

Fiber pathways between PLC and cortex (more diffuse) (A) and DMVAC and cortex (mostly frontal cortex) are illustrated (B).

The FA threshold was 0.15, and the angular threshold was 70° (Keser et al., 2015, 2018). Based on our visual inspection during the analysis, FA thresholding was sufficient for removing spurious fibers that intersect with lesion boundaries from the tracts (Fig. 4A, B). Tracts were visually inspected after being generated on a slice-by-slice basis to also ensure that spurious fibers that include CSF contamination were removed. The step size was randomly selected from 0.5 to 1.5 voxels. The fiber trajectories were smoothed by averaging the propagation direction with 40% of the previous direction. Fiber tracks with a length of 0–140 mm were included, and a total of 50,000 seeds were placed [automated default DSI function to place random seeds on the assigned ROIs (Cheng et al., 2012)]. Once the tracts or nuclei were generated, DSI Studio provided the number of voxels (1 mm³/voxel) of each structure. The total number of voxels was then divided by 1000 to obtain the volume in milliliter. Nuclei and tract volumes were scaled by the ICV (tract or nuclei volumes = tract or nuclei volumes/ICV × 100).

FIG. 4.

Cortical projections arising from PLC are shown bilaterally in a single subject with a large middle cerebral artery infarct (represented as a 3D volume in red) to illustrate the delineated fiber pathways were spared by the stroke lesions. 3D, three-dimensional.

Statistical analyses

The normal distribution of all the variables (including covariates) was assured by visually inspecting the histograms. Baseline demographics (age, sex, education), and stroke and language metrics (lesion load, z-scores from BDAE/WAB) of the patients in the early and late were compared by two-tailed unpaired t-test or Fisher's exact tests for group matching. Two-tailed paired (within-subject) and unpaired (between groups) Student's t-tests were conducted to compare the means of the right and left-sided tract volumes, FA and RD values, as well as these tract values from early stroke and late stroke groups. The laterality index (LI) for diffusion metrics was obtained for each subject separately through a formula of LI: (Left_tract_[scalar] − Right_tract_[scalar])/(Left_tract_[scalar] + Right_tract_[scalar])

For the partial Pearson correlation analyses adjusted for age, education (in years), and lesion load (lesion volume as a percentage of ICV), we obtained standardized residual values from linear regression models. In these models, independent variables were age, education, and lesion load, and the dependent variable was a tract value of interest or BNT score. Standardized residuals were then used in Pearson correlation analyses to obtain p and rho values. We controlled for multiple comparisons by using false discovery rate (FDR) correction (q = 0.05). R software was used for the statistical analysis.

Data availability

The authors certify that they have documented all data, methods, and materials used to conduct the research presented. Anonymized data pertaining to the research presented will be made available to qualified investigators by request via email to the corresponding author.

Results

Two groups did not significantly differ in baseline characteristics, including age (in years) (62.45 ± 14.07 in early stroke; 58.80 ± 12.61 in late stroke; t = 1.06, p = 0.29), sex (14 females in early stroke; 12 females in late stroke, p = 0.79), education (in years) (14.84 ± 2.68 in early stroke, 14.37 ± 2.67 in late stroke; t = 0.68, p = 0.49), lesion load (in percentage to total ICV) (2.81 ± 3.03 in early stroke, 2.65 ± 3.30 in late stroke; t = 0.19, p = 0.85), and z-scores (−0.08 ± 0.84 in early stroke, 0.08 ± 0.87 in late stroke group; t = −0.31, p = 0.55).

Hemispheric comparisons

The tract values are presented as mean ± SD, and the results from each paired t-test include the p-value, corrected p-value (q) if native p-value <0.05, t-statistic, 95% confidence interval (in double bracket). The results are summarized in Table 3.

Table 3.

Diffusion Tensor Imaging Metrics of Right- and Left-Sided Thalamic Nuclei and Pathways Are Presented Along with 95% Confidence Intervals and p-Values Derived from Paired t-Test

Structure and DTI metrics	Early stroke				Late stroke				Early versus late stroke (for left hemispheric structure)
	Left	Right	CI	p_adj	Left	Right	CI	p_adj	p_adj
Dorsomedial and ventral anterior nuclei complex_fractional anistropy	0.37 ± 0.04	0.37 ± 0.04	[−0.01 to 0.003]	0.98	0.37 ± 0.04	0.36 ± 0.04	[−0.008 to 0.02]	0.28	0.89
Dorsomedial and ventral anterior nuclei complex_mean diffusivity	0.91 ± 0.09	0.86 ± 0.07	[0.02 to 0.07]	0.004	0.94 ± 0.09	0.85 ± 0.06	[0.05 to 0.11]	0.001	0.09
Dorsomedial and ventral anterior nuclei complex _adjusted volume	0.59 ± 0.28	0.71 ± 0.29	[−0.18 to −0.04]	0.01	0.61 ± 0.32	0.78 ± 0.28	[−0.27 to −0.06]	0.01	0.22
Pulvinar and lateral posterior nuclei complex_fractional anistropy	0.35 ± 0.04	0.36 ± 0.04	[−0.03 to 0.003]	0.28	0.35 ± 0.05	0.35 ± 0.04	[−0.02 to 0.02]	0.92	0.81
Pulvinar and lateral posterior nuclei complex_mean diffusivity	0.87 ± 0.09	0.81 ± 0.06	[0.02 to 0.09]	0.01	0.91 ± 0.12	0.80 ± 0.08	[0.06 to 0.16]	0.001	0.09
Pulvinar and lateral posterior nuclei complex_adjusted volume	1.44 ± 0.45	1.50 ± 0.48	[−0.14 to 0.03]	0.58	1.43 ± 0.62	1.53 ± 0.45	[−0.24 to 0.04]	0.33	0.75
Dorsomedial and ventral anterior nuclei complex _cortex_fractional anistropy	0.41 ± 0.04	0.43 ± 0.04	[−0.03 to −0.003]	0.04	0.40 ± 0.04	0.42 ± 0.02	[−0.03 to −0.008]	0.04	0.14
Dorsomedial and ventral anterior nuclei complex _cortex_mean diffusivity	0.98 ± 0.10	0.94 ± 0.10	[−0.0006 to 0.008]	0.14	1.05 ± 0.17	0.93 ± 0.08	[0.05 to 0.18]	0.008	0.08
Dorsomedial and ventral anterior nuclei complex _cortex_adjusted volume	6.58 ± 3.45	7.98 ± 3.61	[−2.89 to 0.08]	0.18	5.77 ± 4.44	8.23 ± 3.49	[−4.27 to −0.24]	0.05	0.24
Pulvinar and lateral posterior nuclei complex_cortex_fractional anistropy	0.42 ± 0.03	0.44 ± 0.03	[−0.03 to −0.01]	0.001	0.41 ± 0.03	0.44 ± 0.03	[−0.04 to −0.02]	0.001	0.11
Pulvinar and lateral posterior nuclei complex_cortex_mean diffusivity	0.95 ± 0.07	0.89 ± 0.08	[0.03 to 0.09]	0.003	1.03 ± 0.14	0.89 ± 0.08	[0.09 to 0.20]	0.001	0.02
Pulvinar and lateral posterior nuclei complex _cortex_adjusted volume	9.46 ± 4.01	11.14 ± 4.49	[−3.16 to −0.19]	0.10	9.00 ± 6.08	11.04 ± 3.88	[−4.14 to 0.07]	0.18	0.28

Differences that remained significant after the correction for multiple comparisons are highlighted in bold.

CI, confidence interval; DTI, diffusion tensor imaging; FDR, false discovery rate; p_adj, p-values obtained after FDR correction.

Dorsomedial and ventral anterior nuclei complex

Early stroke group

The FA of the left DMVAC compared with right DMVAC was very similar (0.37 ± 0.04 vs. 0.37 ± 0.04, p = 0.97, t = −0.03, [−0.01 to 0.003]), whereas MD of left DMVAC was significantly higher than right DMVAC (0.91 ± 0.09 vs. 0.86 ± 0.07, p = 0.0003, q = 0.004, t = 4.05, [0.02–0.07]) and volume of the nuclei complex was significantly lower on the left (0.59 ± 0.28 vs. 0.71 ± 0.29, p = 0.002, q = 0.01, t = −3.38, [−0.18 to −0.04]).

Late stroke group

The FA of the left DMVAC compared with the right DMVAC was numerically higher (0.37 ± 0.04 vs. 0.36 ± 0.04), but the difference was not significant (p = 0.43, t = 0.80, [−0.008 to 0.02]), whereas the MD of left DMVAC was significantly higher than right DMVAC (0.94 ± 0.09 vs. 0.85 ± 0.06, p = 2.80 × 10⁻⁵, q = 0.001, t = 4.96, [0.05–0.11]). Likewise, the volume of left DMVAC was significantly lower than the right (0.61 ± 0.32 vs. 0.78 ± 0.28, t = −3.40, p = 0.002, q = 0.01, [−0.27 to −0.06]).

Posterior nuclei complex

Early stroke group

The FA and volume of the left PLC compared with the right PLC was lower (0.35 ± 0.04 vs. 0.36 ± 0.04 and 1.43 ± 0.62 vs. 1.53 ± 0.45), but these differences were not significant (FA; p = 0.13, t = −1.55, [−0.03 to 0.003] and volume; p = 0.20, t = −1.30, [−0.14 to 0.03]). In contrast, the MD of left PLC was significantly higher than right PLC (0.87 ± 0.09 vs. 0.81 ± 0.06, p = 0.002, q = 0.01, t = 3.30, [0.02–0.09]).

Late stroke group

The FA and tract volume of the left PLC were not significantly different from the right PLC (0.35 ± 0.05 vs. 0.35 ± 0.04, p = 0.85, t = −0.19, [−0.02 to 0.02] and 1.43 ± 0.62 vs. 1.53 ± 0.45, [−0.24 to 0.04], p = 0.17, t = −1.40). In contrast, MD of left PLC was significantly higher than right PLC (0.91 ± 0.12 vs. 0.80 ± 0.08, p = 6.76 × 10⁻⁵, q = 0.001, t = 4.64, [0.06–0.16]).

DMVAC cortical projections

Early stroke group

The FA of the left DMVAC cortical projections compared with right DMVAC cortical projections was significantly lower (0.41 ± 0.04 vs. 0.43 ± 0.04, p = 0.01, q = 0.04, t = −2.52, [−0.03 to −0.003]). The MD of left DMVAC cortical projections was insignificantly higher than right DMVAC cortical projections (0.98 ± 0.10 vs. 0.94 ± 0.10) (p = 0.06, t = 2.03, [−5.5 × 10⁻⁵; 0.008]). Although the tract volume was lower on the left, this difference was not significant (p = 0.06).

Late stroke group

The FA of the left DMVAC cortical projections compared with right DMVAC cortical projections was significantly lower (0.40 ± 0.04 vs. 0.42 ± 0.02) (p = 0.004, q = 0.04, t = −3.09, [−0.03 to −0.008]). The MD of left DMVAC cortical projections was significantly higher than right DMVAC cortical projections (1.05 ± 0.17 vs. 0.93 ± 0.08, p = 0.0008, q = 0.008, t = 3.73, [0.05 to 0.18]). The tract volume of left DMVAC was significantly lower than the right, but this difference did not remain significant after FDR correction (5.77 ± 4.44 vs. 8.23 ± 3.49, t = −2.76, [−4.27 to −0.24], p = 0.011, q = 0.05).

PLC cortical projections

Early stroke group

The FA of the left PLC cortical projections was significantly lower compared with right PLC cortical projections (0.42 ± 0.03 vs. 0.44 ± 0.03, p = 0.0001, q = 0.001, t = −4.42, [−0.03 to −0.01]). The MD of left PLC cortical projections was significantly higher than right PLC cortical projections (0.95 ± 0.07 vs. 0.89 ± 0.07, p = 0.0002, q = 0.003, t = 4.25, [0.03 to 0.09]) (Fig. 5A). Although the tract volume was significantly lower on the left side, this difference was not significant after FDR correction. (9.46 ± 4.01 vs. 11.14 ± 4.49, p = 0.03, q = 0.1, t = −2.31, [−3.16 to −0.19]).

FIG. 5.

Boxplot illustration of MDs of the PLCs cortical projections show significantly higher MD values on the left side in both groups (A). Of note, this difference was significantly larger in the late stroke group (p = 0.004). Scatter plots illustrate the correlation between standardized residuals (corrected for age, education, lesion load) of BNT score and the LI of the MD of the left DMVAC (B), LI of DMVAC cortical projections (C), left PLC cortical projections (L_PLC_cortex_MD) (D), left DMVAC adjusted volumes (vol_adj) (E), and, left PLC adjusted volumes (vol_adj) (F). The partial correlations were only significant (after FDR corrections) in the late stroke group. BNT, Boston Naming Test; FDR, false discovery rate; LI, laterality index; MD, mean diffusivity.

Late stroke group

The FA of the left PLC cortical projections was significantly lower compared with right PLC cortical projections (0.41 ± 0.03 vs. 0.44 ± 0.03, p = 6.92e-05, q = 0.001, t = −4.63, [−0.04 to −0.02]). The MD of left PLC cortical projections was significantly higher than right PLC cortical projections (1.03 ± 0.14 vs. 0.89 ± 0.08, p = 1.65e-05, q = 0.001, t = 5.15, [0.09–0.20]) (Fig. 5A). Tract volumes were not significantly different on both sides (9.00 ± 6.08 vs. 11.04 ± 3.88, p = 0.06, t = −1.98, [−4.14 to 0.07]).

Comparisons between early and chronic stroke groups

The FA, MD, and tract values for right-sided nuclei and tracts were very similar between early stroke and late stroke groups, and there were no significant differences for any structures (p > 0.05) (Table 3). The FA and tract values of the left DMVAC and PLC nuclei and tracts were either the same or insignificantly lower (p > 0.05). In contrast, MD values of the left-sided structures were consistently higher in the late stroke group than the early stroke group, and the difference between groups was significant for the MD of the PLC cortical projections (t = 2.96, p = 0.004, q = 0.02, 1.03 ± 0.14 vs. 0.95 ± 0.07) (Fig. 5A).

Partial correlations between naming scores and DTI metrics

Partial Pearson correlations between BNT naming score and DTI metrics, adjusted for age, education, and lesion load, are summarized with rho (r) and FDR-adjusted p-values in Table 4 and are described next.

Table 4.

Correlations of Boston Naming Test Score Through Partial Correlation Analyses Adjusted for Age, Education, and Lesion Load in Early Stroke and Late Stroke Groups

	Early stroke		Late stroke
Structure and DTI metrics	rho	p_adj	rho	p_adj
Left dorsomedial and ventral anterior nuclei complex_fractional anistropy	0.27	0.30	−0.23	0.39
Right dorsomedial and ventral anterior nuclei complex_fractional anistropy	0.05	0.90	0.02	0.93
Dorsomedial and ventral anterior nuclei complex_fractional anisotropy_laterality index	0.30	0.26	−0.25	0.35
Left dorsomedial and ventral anterior nuclei complex_mean diffusivity	−0.11	0.65	−0.42	0.09
Right dorsomedial and ventral anterior nuclei complex_mean diffusivity	0.04	0.91	0.13	0.61
Dorsomedial and ventral anterior nuclei complex _MD_laterality index	−0.18	0.48	−0.53	0.02
Left dorsomedial and ventral anterior nuclei complex _adjusted volume	0.38	0.11	0.53	0.02
Right dorsomedial and ventral anterior nuclei complex _adjusted volume	0.12	0.65	0.13	0.62
Dorsomedial and ventral anterior nuclei complex _adjusted volume_laterality index	0.39	0.14	0.51	0.03
Left pulvinar and lateral posterior nuclei complex_fractional anistropy	−0.29	0.28	−0.35	0.21
Right pulvinar and lateral posterior nuclei complex_fractional anistropy	0.01	0.98	−0.24	0.36
Pulvinar and lateral posterior nuclei complex_fractional anistropy_laterality index	−0.27	0.30	−0.15	0.58
Left pulvinar and lateral posterior nuclei complex_mean diffusivity	0.13	0.62	−0.25	0.34
Right pulvinar and lateral posterior nuclei complex_mean diffusivity	0.04	0.91	0.003	0.98
Pulvinar and lateral posterior nuclei complex_mean diffusivity_laterality index	0.07	0.79	−0.28	0.29
Left pulvinar and lateral posterior nuclei complex_adjusted volume	0.34	0.18	0.63	0.001
Right pulvinar and lateral posterior nuclei complex_adjusted volume	0.19	0.48	0.42	0.11
Pulvinar and lateral posterior nuclei complex_adjusted volume_laterality index	0.11	0.67	0.49	0.04
Left dorsomedial and ventral anterior nuclei complex cortical projections_fractional anistropy	−0.04	0.91	0.24	0.37
Right dorsomedial and ventral anterior nuclei complex cortical projections_fractional anistropy	−0.33	0.22	−0.10	0.70
Dorsomedial and ventral anterior nuclei complex cortical projections_fractional anistropy_laterality index	0.28	0.28	0.28	0.30
Left dorsomedial and ventral anterior nuclei complex cortical projections_mean diffusivity	0.10	0.70	−0.50	0.04
Right dorsomedial and ventral anterior nuclei complex cortical projections_mean diffusivity	0.20	0.44	0.13	0.61
Dorsomedial and ventral anterior nuclei complex cortical projections_mean diffusivity_laterality index	−0.14	0.59	−0.55	0.01
Left dorsomedial and ventral anterior nuclei complex cortical projections_adjusted volume	0.13	0.63	0.36	0.19
Right dorsomedial and ventral anterior nuclei complex cortical projections_adjusted volume	0.007	0.98	0.27	0.31
Dorsomedial and ventral anterior nuclei complex cortical projections_adjusted volume_laterality index	0.25	0.33	0.37	0.19
Left pulvinar and lateral posterior nuclei complex cortical projections_fractional anistropy	−0.17	0.52	0.32	0.23
Right pulvinar and lateral posterior nuclei complex cortical projections_fractional anistropy	−0.31	0.23	0.21	0.45
Pulvinar and lateral posterior nuclei complex cortical projections_fractional anistropy_laterality index	0.19	0.48	0.15	0.59
Left pulvinar and lateral posterior nuclei complex cortical projections_mean diffusivity	−0.17	0.52	−0.50	0.03
Right pulvinar and lateral posterior nuclei complex cortical projections_mean diffusivity	0.24	0.35	−0.32	0.23
Pulvinar and lateral posterior nuclei complex cortical projections_mean diffusivity_laterality index	−0.22	0.39	−0.17	0.52
Left pulvinar and lateral posterior nuclei complex cortical projections_adjusted volume	0.02	0.93	0.31	0.26
Right pulvinar and lateral posterior nuclei complex cortical projections_adjusted volume	0.01	0.98	0.21	0.45
Pulvinar and lateral posterior nuclei complex cortical projections_adjusted volume_laterality index	0.03	0.91	0.33	0.21

Correlations that remained significant after the correction for multiple comparisons are highlighted in bold.

Early stroke group

We found no significant correlations between BNT score and DTI scalars or laterality indices for any structure in the early stroke group even before FDR correction (Table 4 and Fig. 5B–F).

Late stroke group

Dorsomedial and ventral anterior nuclei complex

There was no significant correlation between BNT score and DMVAC FA values (Table 4). The BNT score showed a significant correlation with MD of LI_DMVAC (r = −0.53, p = 0.002, q = 0.02) (Fig. 5B), and this significant correlation was driven by the correlation trend between BNT score and the left DMVAC_MD (r = −0.42, p = 0.02, q = 0.09), as the correlation between BNT score and right DMVAC was not significant (r = 0.13, p = 0.49). Likewise, left-sided smaller volumes significantly correlated with lower BNT score (r = 0.53, p = 0.0025, q = 0.02) (Fig. 5E) and this led to a significant correlation between LI_DMVAC_volume and BNT score (r = 0.51, p = 0.006, q = 0.03).

Posterior nuclei complex

There was no significant correlation between BNT score and PLC FA and MD values (Table 4). However, there were significant correlations between BNT score and the volume of left PLC (r = 0.63, p = 0.00016, q = 0.001) (Fig. 5F) and LI_PLC_vol (r = 0.49, p = 0.009, q = 0.04). There was no significant correlation between the volume of right PLC and BNT score.

DMVAC cortical projections

There was no significant correlation between BNT and DMVAC cortical projections' FA values (Table 4). The MD of left DMVAC cortical projections showed a significant association with the BNT score (r = −0.50, p = 0.007, q = 0.04). This significant correlation led to a significant correlation between the MD of the LI_DMVAC cortical projections and BNT (r = −0.55, p = 0.001, q = 0.01) (Fig. 5C). The correlation between the MD of right DMVAC cortical projections and BNT score was not significant (r = 0.13, p = 0.49). There was no significant correlation between the tract volumes and BNT score.

PLC cortical projections

The MD of the left PLC cortical projections showed a significant correlation with BNT score (r = −0.5, p = 0.005, q = 0.03) (Fig. 5D). There was no significant correlation between BNT and the remaining tract values of the PLC cortical projections (L_PLC cortical projections_FA: r = 0.32, p = 0.09; R_PLC cortical projections_FA: r = 0.21, p = 0.29; R_PLC cortical projections_MD: r = −0.32, p = 0.09). Likewise, the corresponding LI were not significant (LI_PLC cortical projections_FA: r = 0.15, p = 0.44; LI_PLC cortical projections_MD: r = −0.17, p = 0.37). There was no significant correlation between the tract volumes and BNT score.

Discussion

In this study, we showed that the MD of thalamic pathways arising from DMVAC and PLC was higher on the left than the right in both early stroke and late stroke groups, and this pattern was significantly more pronounced in the late stroke group for PLC cortical projections. Increased MD in both thalamocortical pathways arising from DMVAC and PLC significantly correlated with poorer picture naming in the late stroke group but not in the early stroke group, speculatively due to progressive Wallerian degeneration. The MD values reflected changes related to microstructural integrity of the tracts more consistently than FA values, and picture naming correlated well with MD values but not FA values. Left-sided DMVAC had significantly smaller volumes compared with the right in both early and late groups. It is unclear to us why this pattern was not observed in PLC. The left-sided thalamic nuclei volumes also correlated well with the BNT score. However, there was no significant difference between right- and left-sided tract volumes nor was there a significant correlation between tract volumes and BNT score. Although this was a cross-sectional study that included stroke patients with different disease chronicity, the two groups were matched in terms of age, sex, lesion load, education, and z scores from BDAE-3 or WAB.

Thalamocortical connections provide constant feedback and feedforward interactions between the thalamus and cortical regions that are critical for language-related functions (Crosson, 2019). Any lesion in the cortex or thalamus can potentially induce downstream secondary effects and degeneration (Griffis et al., 2019; Price et al., 2017). Interestingly, a DTI study previously showed that the tissue integrity of ipsilateral thalamus decreased 3 months after a corona radiata infarction, signaling Wallerian degeneration (Li et al., 2011). In parallel with these previous studies, we demonstrated that nonlesioned relevant thalamic nuclei implicated in language processing and the fiber pathways connecting them to the cortex had diminished tissue integrity compared with the contralesional side. For PLC cortical projections, the secondary degeneration was more pronounced in the late stroke group. This pattern also showed a significant relation to picture naming abilities. In addition to excluding subjects with thalamic infarcts and removing the lesions from the tracts with thresholding, we also controlled our analyses for overall lesion load as the lesioned tissues were previously found to predict poor outcome (Hillis et al., 2018).

The study lacks healthy controls to affirm right-left lateralization patterns. Previous DTI studies in healthy subjects showed either no right-left asymmetry in the thalamocortical connections (Wakana et al., 2007) or the left thalamus having higher FA than the right (Ardekani et al., 2007; Takao et al., 2011). In our study, we observed an opposite pattern even in the absence of stroke lesions in the thalamus. This is speculatively secondary to Wallerian degeneration that took place after the injury. Another mechanism besides Wallerian generation that could explain these results is reduced structural connectivity between thalamic nuclei and cortical structures due to damage to the projecting fibers themselves, even in the face of intact nuclei. Studies with healthy controls are needed to confirm our findings.

In the early stroke group, we have not observed any association between DTI metrics of any thalamic nuclei or pathways and picture naming scores. Although we do not have long-term follow-up for this group, speculatively, secondary effects of the cortical lesions on the thalamus take longer to become apparent and have an association with picture naming. In addition, although the right hemispheric structures might potentially be associated with language functions (Wright et al., 2018), we have not observed any relationship between right thalamic nuclei/pathways and picture naming in our cohort. This finding is in parallel with the previous studies showing language lateralization to the left thalamus (De Witte et al., 2011; Fritsch et al., 2020; Schmahmann, 2003).

The DMVAC mainly projects to the prefrontal and frontal areas, whereas PLC projects to parietal-temporal-occipital association cortices as well as to the speech motor cortices (Nadeau and Crosson, 1997; Stemmer and Whitaker, 2008). In our study, the fiber pathways arising from these two complexes revealed similar frontal cortex terminations, but PLC has more diffuse cortical projections including the posterior temporoparietal association areas. Although we expected a stronger association between naming and PLC/PLC cortical projections compared to DMVAC/DMVAC due to these more diffuse cortical projections, both nuclei complexes/tracts showed a similar degree of association with picture naming. This finding could possibly be explained by the role of speech motor areas as the final output hub in picture naming, but further studies are needed to confirm this possibility.

In our cohort, we observed that MD values were more sensitive in reflecting changes related to microstructural integrity of the tracts than FA values, and picture naming correlated well with MD values but not FA values. This observation is related to a concomitant increase in axial diffusivity (AD) and RD due to the downstream effects of the ischemic infarcts (Visser et al., 2019). The changes in FA and MD are also influenced by the time since stroke as well. The AD and RD have additive effects in MD calculation but subtractive effects in FA calculation. Note that AD and RD values are not reported in this study to avoid redundancy of metrics, as FA and MD values are derivates of RD and AD. Our finding of MD being more sensitive in capturing changes aligns well with a previous longitudinal stroke study showing an increase in all the diffusivity measures in perilesional tissue over time (Umarova et al., 2017) as well as another study in multiple sclerosis revealing an increase in all diffusivity measures distal to a demyelinating lesion in the optic nerve (Klistorner et al., 2015).

Limitations

Our study is limited by a small sample size and cross-sectional design. In general, many patients presented acutely with language impairments that resolved by 1 year poststroke. The two groups may represent two distinct populations independent of stroke acuity, as most participants in the late group have chronic language impairments, and many participants in the early group may never have chronic language impairment. We attempt to study all patients at the late stage, whether or not they have clinical aphasia. However, we believe that two homogenous groups from different stages of stroke with matched baseline characteristics provide valuable insights into the temporal and longitudinal evolution of secondary degeneration after the initial insult. Larger samples, age-matched healthy controls, and multiple time points are needed to confirm our findings and further characterize dynamic changes in the thalamocortical interplay. Despite these limitations, to the best of our knowledge, our study is the first study in the poststroke aphasia literature suggesting the secondary degeneration in the relevant thalamic nuclei and thalamocortical connections in relation to language-related processes.

Another limitation of our study is the acquisition of nonisometric DWI voxels (0.82 × 0.82 × 2.2 mm), which bias tractography regardless of resampling, even in a healthy brain. This is particularly notable due to the stronger possibility of partial volume effects in the lesioned hemisphere that could be more likely to increase MD when upsampling.

Also, we have utilized neuroanatomical guidance to guide our deterministic tractography that might have led to inclusion of the fiber tracts that may have made sense anatomically but still have been false positives. This is an inherent limitation to deterministic tractography algorithms.

Conclusion

In conclusion, we show that nonlesioned thalamic nuclei and thalamocortical pathways on the left show progressive Wallerian degeneration over time after a left hemispheric stroke, and this pattern is associated with worse picture naming.

Authors' Contributions

Z.K.: design or conceptualization of the study, drafting and revising of the article for intellectual content and analysis and interpretation of the data. E.L.M., M.D.S., B.L.S., R.S., and A.E.H.: design or conceptualization of the study, revising the article for intellectual content and interpretation of the data.

Author Disclosure Statement

No competing financial interests exist.

Funding Information

This work was made possible by the National Institutes of Health (National Institute of Deafness and Communication Disorders) through awards R01 DC05375 (PI: A.E.H.), R01 DC015466 (PI: A.E.H.), P50 DC011739 (PI: Fridriksson), and R00 DC015554 (PI: R.S.). The authors gratefully acknowledge this support.