INTRODUCTION
Neuroimaging is an important diagnostic tool in stroke rehabilitation [1-5]. Among various neuroimaging modalities, diffusion-tensor imaging (DTI) is unique for its ability to evaluate neural fibers in vivo [6]. Of the various parameter estimates obtained from DTI, fractional anisotropy (FA) is frequently employed to assess neural integrity in the brain, particularly as an indicator of Wallerian degeneration [6]. DTI is partially used for outcome prediction in clinical practice [4,7,8].
The vast majority of DTI studies in relation to stroke outcomes have focused on upper extremity functions. However, stroke patients often experience reduced levels of functional independence in their activities of daily living (ADL). There have been a very limited number of DTI studies that focused on ADL as an outcome measure [9-11]. This is partially due to the complexity of ADL, which encompasses ambulatory ability and cognitive functions such as memory.
Owing to the newly developed automated DTI analysis procedure known as XTRACT [12], the assessments of 42 representative neural tracts within the whole brain have become possible within 1 hour per subject. In our previous studies [13-15], we have ensured clinical applicability of this newly developed technique for the assessment of hemiparesis and aphasia [13-15]. This study aimed to further expand this newly developed technology to ADL in addition to extremity functions. Because ADL is evidently influenced by age [11], in this study, we incorporated age in addition to DTI metrics.
METHODS
Patients
The study protocol received approval from the Institutional Review Board of Hyogo Medical University (No. 4546) and informed consent for participation was obtained using the opt-out method. This research utilized a retrospective cohort of medical records. The study database consisted of patients who were admitted to Nishinomiya Kyoritsu Neurosurgical Hospital for the treatment of stroke between April 2022 and September 2023. The diagnosis of stroke was based on conventional computed tomography and/or magnetic resonance imaging scanned in the emergency unit of our hospital [16,17]. The management of stroke aligned with the Japanese Guidelines for the Management of Stroke 2021 [18], including the prescribed rehabilitative regimen.
To mitigate potential confounding factors arising from variations in pre-stroke health status and lesion sites, we restricted our sample to first-ever stroke patients with supratentorial lesions who were functionally independent in ADL prior to the stroke [9,11,19]. We also excluded patients who exhibited subsequent deterioration in neurological manifestations and other medical conditions during acute care [9,11,19]. Moreover, to minimize disparities in the rehabilitative regimen, we specifically collected data from patients exclusively transferred to our affiliated long-term rehabilitation facility, Nishinomiya Kyoritsu Rehabilitation Hospital [9,11,19].
DTI acquisition
DTI scans were typically conducted during the second week following admission to our acute care service [20], utilizing a 3.0-Tesla scanner (MAGNETOM Trio; Siemens AG) equipped with a 32-channel head coil. The acquisition of DTI data employed a single-shot echo-planar imaging sequence in the anterior-to-posterior direction, comprising 30 images with non-collinear diffusion gradients (b=1,500 s/mm2) and one non-DWI scan (b=0 s/mm2). For each patient, 80 contiguous axial slices were obtained with a field of view of 256 mm×256 mm, an acquisition matrix of 128×128, and a slice thickness of 2 mm. The echo time was 96 ms, the repetition time was 10,900 ms, and the flip angle was 90°. To address eddy current-induced and echo-planar imaging-induced distortions, two additional non-DWI scans were acquired in the anterior-to-posterior direction and two in the posterior-to-anterior direction. Additionally, for capturing anatomical details of the patients’ brains, T1-weighted images were acquired using a three-dimensional fast gradient imaging sequence. For each patient, a total of 176 contiguous sagittal slices were acquired with a field of view of 256 mm×256 mm, an acquisition matrix of 256×256, and a slice thickness of 1 mm. The echo time was 2.52 ms, and the repetition time was 1,900 ms, and the flip angle was 10°.
Image processing
The DTI scan image processing pipeline employed the MRIcron, MRtrix, and FSL software packages [21-23]. The details of image processing were previously published [13-15]. In brief, initial steps involved eliminating the Gibbs ringing artifact, correcting distortions induced by eddy currents and echo-planar imaging, and applying bias field corrections. Brain masks were then derived from the bias field-corrected images. Following the preparatory stage, we utilized the XTRACT function within FSL for fiber tracking [12]. This enabled the generation of tractography for 42 predefined sets of neural bundles. The purpose of the present study was to assess the relationships between neural-fiber damages and outcome measurements among stroke patients. According to our previous studies [9,10,24], the corticospinal tract (CST) was strongly associated with extremity function, whereas association fibers such as the superior longitudinal fasciculus (SLF) and the inferior fronto-occipital fasciculus (IFOF)—partially overlapping with the anterior thalamic radiation (ATR), inferior longitudinal fasciculus (ILF), and uncinate fasciculus (UF)—were linked to cognitive functions. Our previous studies reported that neural degeneration was most prominent in these neural tracts among patients following a supratentorial stroke [9,10,24]. To ensure clarity in the statistical procedures, this study focuses on the CST, IFOF, and SLF, as illustrated in Fig. 1. Parameter estimates, including FA values, were then extracted using a threshold set at 0.01, consistent with findings from our prior studies [13-15].
Outcome measurements
In this study, outcomes were assessed using the motor component of the Stroke Impairment Assessment Set (SIAS-motor) [25] and the Functional Independence Measure (FIM) [26]. SIAS-motor comprises five components—arm, finger, hip, knee, and ankle functions—each scored on a scale from null to full (0 to 5). To quantify gross motor function on the paralyzed side for each patient in this study, we calculated the total sum of SIAS-motor scores. This calculation follows the method outlined in our previous studies [13,17]. The FIM, a widely adopted tool for evaluating independence in ADL, consists of two components: motor (13 items) and cognition (5 items). Each item is scored on a 7-point scale (1=total assistance; 7=complete independence). Total scores for both FIM-motor (scale range, 13–91) and FIM-cognition (scale range, 5–35) are commonly used in stroke rehabilitation. SIAS-motor and FIM scores were assessed bi-weekly, with data collection conducted at our affiliated long-term rehabilitation facility upon discharge. Additionally, the total length of hospital stay (LOS) was recorded for each patient.
Statistical analysis
Multivariate regression was used to analyze the data, with separate assessments conducted for SIAS-motor total, FIM-motor, FIM-cognition, and LOS data. In all analyses, age and lesion-side FA values from the CST, the IFOF, and the SLF were designated as explanatory variables. The final regression models were determined through parameter selection using the Akaike information criterion method. To identify potential multicollinearity, Spearman’s correlation test was conducted for all possible pairings of the four explanatory variables. Statistical analyses were performed using the JMP software package (SAS Institute Inc.), and significance was set at a p-value<0.05.
RESULTS
Throughout the study period, a total of 48 patients meeting our inclusion criteria were enrolled. Three patients were excluded due to suboptimal DTI quality, and additional three patients were also excluded because their FIM data were incomplete. Consequently, our final analytical database comprised 42 patients (Table 1). The research cohort comprised 27 cases of ischemic strokes and 15 cases of hemorrhagic strokes, including 28 male and 14 female patients, ranging in age from 42 to 88 years. The calculated FA values within the CST on the lesion side ranged from 0.292 to 0.627. Similarly, the FA values in the IFOF varied from 0.320 to 0.555, while those in the SLF spanned from 0.196 to 0.446. The total score for SIAS-motor ranged from 0 to 25, FIM-motor spanned from 39 to 91, FIM-cognition ranged from 21 to 35, and LOS varied from 36 to 208.
Table 2 presents the outcomes derived from multivariate regression analyses, while Table 3 displays the results of correlation analyses among explanatory variables. For the SIAS-motor total, the CST emerged as the most influential determinant (F=20.340). Although age and the IFOF were identified as statistically significant contributory factors, their impact was comparatively less pronounced. Regarding FIM-motor, three explanatory variables—age, the CST, and the IFOF—were recognized as statistically significant contributory factors. Among these, age emerged as the most prominent determinant factor (F=18.915). In the case of FIM-cognition, age and the IFOF were recognized as statistically significant contributory factors. Consistent with the observations in FIM-motor, age emerged as the most prominent determinant factor (F=21.984). For LOS, age, the IFOF, and the SLF were determined to be statistically significant contributory factors. As similar to the findings for FIM-motor and FIM-cognition, age was recognized as the most powerful determinant (F=8.326). As shown in Table 3, the correlation between the CST and the IFOF (R=0.440) and that between the IFOF and the SLF (R=0.387) were statistically significant.
DISCUSSION
In this study, we investigated the relationships between DTI data and stroke patient outcomes, incorporating age as an additional factor. Regarding extremity function impairment, as indexed by the SIAS-motor total scores, the CST emerged as the most significant determinant, while age played a less prominent role. In contrast, for ADL performance measured by FIM scores, age was the most influential factor, surpassing the impact of damage to neural tracts, as indicated by FA values, including the CST and IFOF. Additionally, LOS was influenced by both age and association fibers, such as the IFOF and SLF.
The findings of this study emphasize that the primary contributing factors to outcomes vary significantly depending on the modality of assessment, such as ADL and extremity functions. Recent neuroimaging studies focusing on CST integrity often utilized upper extremity function as an outcome measure (e.g., Fugl-Meyer Assessment) [3,27]. Notably, age was rarely incorporated as an explanatory factor in these studies [3], which may be acceptable given the nature of extremity function assessments. In contrast, when ADL was evaluated using the FIM scoring system—comprising motor and cognitive components—age and association fibers, including the IFOF, were more prominently associated with outcome scores. Specifically, age emerged as the most influential determinant for ADL outcomes, possibly due to the inclusion of FIM-motor items, such as dressing, that are closely linked to cognitive function, which is naturally influenced by age. The dissociation between extremity functions and FIM-motor components can likely be attributed to this relationship. As demonstrated, the primary contributing factors exhibit significant variation based on the method of outcome measurements.
Our previous study revealed correlations between stroke outcome scores, including FIM-cognition and FA decrease in broader brain areas, such as the ATR, the UF, and the ILF [24]. Visual inspection on the obtained tractography revealed a partial overlap between the anterior part of the IFOF, the ATR, and the UF [10,15]. Additionally, the posterior part of the IFOF exhibited overlap with the ILF [10,15]. Preliminary correlation analyses among these four neural tracts, similar to those shown in Table 3, demonstrated statistically significant relationships in four out of six possible pairs. It is true that different neural tracts reflect different aspects of brain function. Among the four neural tracts, however, the IFOF is the largest and is associated with a wider range of cognitive functions [28]. Considering that the FA values derived from these neural tracts are highly correlated and that, given the sample size, we must limit the number of explanatory variables included in the multivariate regression analyses, we selected the IFOF as the representative tract. The XTRACT function implemented in FSL was employed in this study. Concerning the SLF, it generates three parts; the dorsal (part 1), the intermediate (part 2), and the ventral (part 3). Consistent with the aforementioned reasons, we designated part 3 as the representative for the SLF in this study.
The cognitive symptoms in stroke patients vary depending on the side of the lesions; individuals with left hemisphere lesions often experience aphasia and/or apraxia, while those with right hemisphere lesions may suffer from neglect and/or disorientation. Despite these differences in lesion hemisphere and cognitive symptoms, we did not conduct separate analyses for the right and left hemisphere lesion groups in this study. In our previous study, we used tract-based spatial statistics to assess neural damage in relation to clinical symptoms. We employed the FIM-cognition score to evaluate cognitive decline in patients after stroke [24]. The results indicated that the degree of cognitive decline correlated with FA values within some association fibers, including the ATR, the ILF, the SLF, and the UF. The observed patterns were nearly symmetrical between the right hemisphere lesion group and the left hemisphere lesion group [24]. Consequently, for the present study, we did not perform separate analyses for the right hemisphere lesion group and the left hemisphere lesion group.
Results from multivariate regression analyses indicated that, besides age, LOS was influenced by both the SLF and the IFOF (Table 2). In our previous study [11], we assessed the CST integrity using the ipsilesional-to-contralesional ratio of FA within the cerebral peduncles. We then performed multivariate regression for LOS, setting the CST neural integrity, age, and stroke type (ischemic or hemorrhagic) as explanatory variables. The results from our previous study revealed that age and the CST neural integrity accounted for much of the variability in LOS (adjusted R2=0.420) [11]. Due to technical difficulty, the previous study did not include the IFOF and the SLF data. In this study, as indicated in Table 3, we observed a mild correlation between the CST and the IFOF. Although the results of the LOS analysis did not include the CST as a statistically contributory factor, such an outcome could be expected due to the issue of multicollinearity.
The vast majority of previous DTI studies investigating the association between the CST integrity and motor functions employed the ipsilesional-to-contralesional ratio of FA [3,11,29,30]. This procedure reduces inter-individual differences in the CST-FA. Nevertheless, the application of the ratio FA diminishes the suitability of DTI for individuals experiencing recurrent strokes, as the FA on the non-lesion side is expected to remain undamaged. We have recently investigated the clinical applicability of raw FA values to assess motor functions after a stroke. The findings were promising [13]. Accordingly, in this study, we employed raw FA values to assess the integrity of neural bundles.
In this study, we included both types of ischemic and hemorrhagic strokes in our analytical database. However, we did not include the type of stroke as an explanatory variable in the multivariate regression analysis. In our previous study, we found that the FA decrease in the CST was more evident in hemorrhagic stroke than in ischemic stroke [31]. Nevertheless, the severity of affected extremity functions paralleled the observed FA decreases [31]. In another previous study [11], we investigated the contributions of the type of stroke to the outcome measurements for both the extremity functions and ADL. The obtained results indicated that the contribution of the type of strokes was minimal [11]. Accordingly, in this study, we omitted including the type of stroke as an explanatory variable in the multivariate regression analysis.
Previous studies have indicated that while extremity functions are largely attributed to the integrity of the CST, cognitive functions, such as aphasia and spatial neglect, involve a broader network of associative fibers, including the SLF and the inferior IFOF [9,10,24]. The specific neural substrates responsible for these functions have been partially elucidated through modern neuroimaging techniques. For instance, the SLF is now thought to be associated with phonetic processing, whereas the IFOF is considered to play a role in semantic processing [15,32]. However, in this study, we utilized total scores of FIM-cognition, which provides a general assessment of various cognitive domains, such as memory, expression, and comprehension. Further research is needed to clarify the relationships between specific components of cognitive functions and their associated neural tracts.
In this study, we acquired DTI data during the second week after admission. Ideally, a longer observation period would be preferable to improve the accuracy of detecting FA decreases [20]. However, under the Japanese health insurance system, stroke patients requiring inpatient rehabilitative treatment are typically transferred to long-term convalescent rehabilitation facilities around the third to fourth week after admission [33]. Unfortunately, our affiliated rehabilitation hospital is not equipped with magnetic resonance scanners. Future studies with longitudinal DTI acquisition settings are desirable [20].
This study has several limitations. First, the FA values are influenced by the threshold setting. In the current study we set the threshold as 0.01 in line with our previous reports in a DTI study series [13-15]. However, there is currently no consensus regarding this setting. Second, the study’s population was restricted to first-ever stroke patients who were functionally independent before the onset of stroke. Consequently, the applicability of the study’s findings to geriatric patients requiring assistance in ADL before stroke remains uncertain [11]. Third, the sample size for the present study comprised 42 participants, which may be considered relatively small in retrospective research studies. However, it is noteworthy that a recent systematic review [3], examining a wide array of studies, revealed that, among the 71 studies reviewed, interquartile range of the samples was 15 to 50 (median, 28). While a larger sample size could increase statistical power and potentially enhance the generalizability of findings, it is crucial to acknowledge that smaller sample sizes are not uncommon in clinical neuroimaging studies.
Conclusion
This study indicated that, incorporating age, the FA derived from automated tractography is associated with levels of independence in ADL and functional capacity of the affected extremity. However, relative contributions of FA and age were different among the modality of outcomes. These findings suggest that this combination may be useful for predicting outcomes in stroke patients.
