Tract-Specific Correlates of Neuropsychological Deficits in Patients with Subcortical Vascular Cognitive Impairment

INTRODUCTION

Subcortical vascular cognitive impairment (SVCI) is characterized by extensive small vessel disease (SVD) such as white matter hyperintensities (WMH) and lacunes in white matter regions [1, 2]. Previous studies have shown that SVD, which occurs predominantly in frontal and subcortical regions, can affect fronto-subcortical circuits, eventually resulting in frontal executive dysfunction [3, 4]. Furthermore, several studies showed that SVD affected frontal dysfunction regardless of their location [5, 6]. However, increasing evidence suggests that patients with SVCI have multiple cognitive impairments and not just frontal executive dysfunction [7, 8]. In this context, it has been hypothesized that cortico-cortical or cortico-subcortical disconnection might contribute to multiple cognitive impairments in SVCI patients [9]. Although this hypothesis has been widely accepted, evidence is indirect and limited.

Previous studies have primarily focused on correlations between gray matter abnormalities and cognitive impairments [10, 11]. In contrast, the associations of white matter tracts with specific cognitive impairments remain unclear. Previous studies revealed that SVD-related white matter deficits were associated with psychomotor slowing or frontal executive dysfunction rather than other cognitive impairments [4, 12]. It might be explained by the damage to the white matter itself. Alternatively, it would, be reasonable to expect that each white matter tract subserves particular brain functions, considering that it connects cortical or subcortical gray matter regions responsible for specific cognitive functions.

Diffusion weighted imaging (DWI) is a method that can facilitate the investigation of white matter microstructural alterations. A popular method, tract-based spatial statistics (TBSS), attempts to localize these alterations by projecting diffusion parameters onto a white matter skeleton [13]; however, this projection may contain differently oriented fibers [14]. In contrast, tract-specific statistical analysis (TSSA), a method we developed and validated, enables us to analyze tract diffusion coefficients moreprecisely.

In the present study, we explored tract-specific abnormalities in SVCI patients using TSSA. First, we determined which white matter tracts were selectively vulnerable in patients with SVCI. Second, we investigated tract-specific neuropsychological correlates in these patients.

MATERIALS AND METHODS

RESULTS

DISCUSSION

In the present study, we explored tract-specific abnormalities in SVCI and correlates of neuropsychological deficits using our TSSA method. TSSA provides a greater ability to localize deficits in white matter tracts than previous methods, which average all diffusion coefficients of fiber tracts [34–37]. Though it is well known that white matter damage is critical in patients with SVCI, few studies have investigated the associations between white matter tracts and cognitive impairments. Taking advantage of our TSSA method, the major findings of this study were that deficits were more severe in tracts that traverse frontal and parietal white matter regions, and that focal disruptions in specific tracts, rather than widespread disruption, were associated with specific cognitive impairments. Therefore, our findings may extend the understanding of mechanisms by which white matter disruption leads to cognitive impairments in SVCI.

In the present study, relative to NC subjects, SVCI patients showed significant deficits in tracts that traverse frontal and parietal white matter regions. Notably, bilateral ATR, left CG and left SLF showed involvement along more than 50% of each tract. In contrast, no deficits were observed in right ILF and bilateral IFO, tracts that traverse temporo-occipital white matter. This finding corroborates the notion that ischemia may specifically affect white matter in frontal and parietal regions rather than in temporal and occipital regions [38–40]. Our TSSA method, combined with multiple comparisons correction, may enable us to observe more focal findings, particularly compared with previous studies showing that patients with SVCI have widespread white matter microstructural alteration [41–43].

The major finding that disruptions in specific tracts were associated with multiple cognitive impairments was supported by the following observations. First, FA values in CG were associated with scores in language, visuospatial, memory, and frontal functions. Second, FA values in ATR were associated with scores in attention, language, memory, and frontal functions. Third, FA values in SLF were associated with scores in visuospatial and frontal functions. Most previous studies have focused on the relationship between white matter and frontal dysfunction [3, 44]. However, our findings suggest that damage to white matter tracts could cause not only frontal dysfunction but also dysfunction in other cognitive domains such as language, visuospatial, and memory functions.

The mechanisms by which these multiple cognitive impairments developed in patients with SVCI require further investigation. However, specific gray matter regions are responsible for specific cognitive domains, and white matter tracts interconnect these gray matter regions. It is, therefore, reasonable to argue that ischemia affected the white matter tracts of SVCI patients, and the resulting interruption of information transfer between gray matter regions may be associated with these cognitive impairments. Disruption in specific white matter fiber tracts could contribute to multiple cognitive impairments if the corresponding gray matter regions were responsible for multiple cognitive functions or specific fiber tracts connected multiple gray matter regions.

We found that microstructural differences in the middle portion of CG were associated with scores in language, visuospatial, memory, and frontal functions. A previous study showed that FA values in the posterior CG were associated with language, visuospatial, memory, and attention/executive dysfunction, while FA in the anterior CG was associated with memory and attention/executive dysfunction [45]. CG bi-directionally interconnects anterior and posterior cingulate cortices [46, 47]. Disruptions of the anterior and posterior cingulate cortices are responsible for frontal dysfunction, and episodic memory and visuospatial dysfunctions, respectively [46, 49]. Given that the transfer of various forms of information occurs in the middle portions of CG, disruption in the middle of CG may contribute to multiple cognitive impairments. In the present study, deficits in the anterior portion of the left ATR were associated with scores in attention, memory and frontal executive functions, while those in its middle portion were associated with score in language function. This suggests that the profile of cognitive impairment differs according to the location of FA disruptions even within the same fiber tract. Previous studies suggested that WMH and lacunes in ATR were associated with executive dysfunction and slow processing speed [12, 50], though these studies did not evaluate regional relations with cognitive impairments. The ATR connects anterior and dorsomedial thalamic nuclei with the prefrontal cortex [51]. It is well recognized that dorsolateral prefrontal cortex is responsible for attention and frontal functions. Also, damage to ventral anterior and mediodorsal thalamic nuclei consistently leads to language disturbances [52]. Thus, our findings suggest that the specific regions of tracts that neighbor their corresponding gray matter regions are critical to the particular cognitive impairments. In this context, considering that the anterior nucleus of thalamus is a key component of the Papez circuit for episodic memory [53], the relationship between ATR and memory may be expected in the posterior portion of ATR. However, we found that disruption in the anterior portion of ATR was associated with memory impairment. Moreover, this association with memory score was driven mainly by delayed recall rather than recognition, suggesting that it might be related to type of retrieval deficit, a typical finding in SVCI patients. In fact, memory-related regions of ATR overlapped generally with those related to frontal dysfunction.

We found that microstructural differences in the middle or posterior portion of SLF were associated with visuospatial dysfunction. The middle or posterior portion of SLF interconnects the medial posterior parietal lobe, cuneus, and lingual and fusiform gyri, which are associated with visuospatial dysfunction [49]. Importantly, our finding of association between FA in the middle portion of SLF and frontal dysfunction corroborates a previous study from our group using an independent sample [2]. That study showed that frontal executive dysfunction was associated with disruption in posterior white matter [2].

Interestingly, microstructural changes only in the part of fasciculus, but not whole region, were associated with cognitive impairments. There might be several explanations. First, a fasciculus may contain connections with multiple regions which are responsible for corresponding cognitive impairments. Therefore, there might be differences in locations of a fasciculus according to patterns of cognitive impairments. Second, there might be differences in a complex internal organization of the connection in a fasciculus. Therefore, there might be specific regions of a fasciculus which are related to cognitive impairments. Finally, considering the specific regions of tracts neighboring their corresponding gray matter regions are critical to the particular cognitive impairments, Wallerian degeneration following atrophy in gray matter might partially explain our finding that only part of a fasciculus link to cognitive scores.

A strength of our study is that our method can detect tract specific white matter local deficits precisely. The whole brain connectivity analysis may investigate alteration in relationship between brain regions; however, since each connection between the regions was determined by averaging FA values along its whole pathway, it cannot localize which part of a fiber tract was affected (e.g., the middle portion of CG). Comparing to the TBSS, since our tract specific analysis considers fiber directions of tractography results while TBSS does not, it may assess local alteration more precisely than TBSS [14]. Compared to other types of tract specific analyses [14, 54–56], our method has two advantages. First, our white matter tract classification method has improved labeling performance [26]. Since our approach modeled fiber tracts by clustering individual streamlines based on their shapes and trajectories without defining ROIs for masking streamlines, it is more robust to inter-subject variability and registration errors than the ROI-based automatic labeling [27]. Moreover, our approach is more similar to what human experts do conceptually and practically than the other approach [27]. The performance of our approach excelled the other, measured by κ which evaluates agreement of automatic labeling with the expert’s labeling (the ground truth): average performance of our labeling was 0.90 while the other approach was 0.60 [27]. In this study, we employed a spectral matching algorithm to improve performance further, allowing partial matching between template tracts and those of subjects. Second, using CBS, our statistical analysis is robust to noise; this method is popular in VBM but has not been employed in previous tract-specific analyses. CBS provides a single corrected p-value for a clustered region whose FA values were significantly different between groups, or significantly correlated with a test score.

Some methodological issues exist. First, fiber tracts were sometimes lost for a certain subject. Even though our algorithm has improved labeling performance, it is based on fiber shape; when severe atrophy affects the shape of fibers too much, TSSA fails to classify streamlines into known fiber tracts. However, the number of subjects who did not have a certain fiber tract was 0.17 over 55 NC subjects and 2.08 over 114 SVCI subjects on average except both IFO (NC, left: 51% loss, right: 13% ; SVCI, left: 56% , right: 43% ) which did not show any significant results. Second, though CBS is robust to noise, it requires arbitrary initial thresholds. This can be resolved with the threshold-free cluster enhancement method [57]. Third, the results of the correlation analysis should be carefully interpreted, since the regions with significantly reduced FA values did not always overlap with the correlation study results. Finally, our study population was enriched for patients with significant vascular burden, which may limit the generalizability of our data to other populations. However, to the best of our knowledge, our study is the first to investigate the tract-specific correlates of neuropsychological results in patients with SVCI. In future, we would further investigate sub-bundles of a fiber tract. Though our method identifies sub-bundles of a fiber tract, we collated them into one fiber tract for simplicity.

In conclusion, our findings suggest that disconnection of specific fiber tracts, especially in regions that connect to and are neighboring corresponding gray matter regions, contribute to multiple cognitive impairments in patients with SVCI. Furthermore, our findings elucidated the pathological mechanisms of cognitive impairments in patients with SVCI.