White Matter Integrity of the Three Subdivisions of the Superior Longitudinal Fasciculi in First-Episode Psychosis Patients
Article information
Abstract
Objective
The superior longitudinal fasciculus (SLF) is a white matter (WM) bundle connecting the frontal and parietal lobes. SLF subdivisions have distinct functions (e.g., attention, working memory, and language processing), and abnormalities in specific subdivisions may differentially contribute to the pathophysiology of first-episode psychosis (FEP). However, previous diffusion magnetic resonance imaging (MRI) studies have not fully examined SLF subdivisions, limiting our understanding of their role in FEP patients. The aim of this study was to assess whether WM integrity differs in SLF subdivisions between FEP patients and healthy controls (HCs) using advanced diffusion tensor imaging.
Methods
3T diffusion MRI scans were obtained from 39 FEP patients and 110 HCs. Deterministic tractography was used to reconstruct the three SLF subdivisions (SLF I, II, and III). We analyzed the fractional anisotropy (FA), mean diffusivity (MD), and radial diffusivity (RD) of these subdivisions and compared the results between FEP patients and HCs.
Results
There were no significant group differences in the FA, MD, or RD between the FEP patients and HCs.
Conclusion
No significant differences in WM integrity between FEP patients and HCs were found. These results suggest that significant WM changes may develop later in the disease course rather than in the early stages of psychotic disorders such as FEP.
INTRODUCTION
Schizophrenia is a severely disabling psychiatric disorder that affects 0.3% to 0.7% of the global population [1] and is characterized by a wide range of clinical symptoms, including both positive and negative symptoms [2]. The disconnection hypothesis suggests that the symptoms of schizophrenia result from disrupted integration of neural networks rather than from localized brain abnormalities [3]. Specifically, white matter (WM) disruptions, particularly those involving the prefrontal cortex [4], are thought to impair the coordination and synchronization of brain activity, with myelination changes leading to the core pathology of schizophrenia [5]. Several studies have reported alterations in WM bundles, including the superior longitudinal fasciculus (SLF), cingulum bundle, uncinate fasciculus, inferior longitudinal fasciculus, and arcuate fasciculus, in schizophrenia patients [6].
Among the WM bundles in the human brain, the SLF is the largest associated fiber bundle, connecting the frontal and parietal regions within the same hemisphere [7]. Anatomically, this major bundle is further divided into three subdivisions (i.e., SLF I, II, and III), each of which is responsible for different functions targeting independent brain regions [8]. SLF I is involved in attention and working memory, SLF II is involved in visuospatial processing, and SLF III is involved in language and spatial awareness [9]. Additionally, SLF II acts as a modulator between SLF I and SLF III, contributing to their functions by overlapping with both regions [10]. The SLF involves various regions, including the prefrontal cortex, which is a crucial area in patients with schizophrenia [11]; thus, the SLF is thought to function as a bridge between brain regions implicated in the pathophysiology of schizophrenia.
Previous diffusion tensor imaging (DTI) studies investigating WM integrity have reported deficient WM integrity in the SLF of patients with schizophrenia compared with that of healthy controls (HCs), but these results are inconsistent, possibly because of methodological factors such as the reliance on tract-based spatial statistics (TBSS) and voxel-based analysis (VBA) rather than the use of more precise segmentation techniques such as deterministic tractography (DT). Although TBSS is a widely used approach, voxels are often misassigned owing to blurring of adjacent WM tracts, leading to inaccurate tract identification and misinterpretation of fractional anisotropy (FA) values [12]. Furthermore, in many studies, the specific subdivisions of the SLF (i.e., SLF I, II, and III), which have distinct functions and may be differentially affected in patients with schizophrenia, have not been analyzed [13]. Several studies using the TBSS approach have reported impaired WM integrity of the SLF subdivisions in patients with schizophrenia [14] and first-episode psychosis (FEP) [15]; furthermore, other studies employing VBA have reported deficient SLF integrity in FEP patients [16]. In contrast, another TBSS study revealed intact WM integrity in schizophrenia patients, although the entire SLF was analyzed than its subdivisions in this study [17]. These inconsistencies underscore the need for studies employing more delineated imaging techniques, such as DT, to better understand SLF abnormalities in patients with schizophrenia or FEP.
DT is a DTI method that can be used to investigate the SLF by visualizing WM tract anatomy in vivo through the reconstruction of WM fibers [18,19]. Through DT, more precise fiber estimation can be achieved, leading to more accurate anatomical investigations of SLF subdivisions [20]; furthermore, DT is more sensitive to real WM properties than other approaches in terms of diffusion indices [21]. Although DT can provide more reliable and precise reconstructions of tracts in SLF subdivisions, only one tractography study on SLF subdivisions in schizophrenia patients has been conducted. Leroux et al. [22] reported that schizophrenia patients presented with reduced FA and increased mean diffusivity (MD) and radial diffusivity (RD) values in the right SLF II and decreased FA values in the right SLF III. Thus, it is important to use DT to investigate whether the impairment of the SLF subdivisions in FEP patients occurs in the early stage of psychotic disorders, as patients’ brains are less affected by aging, the chronic nature of this disease, and psychotropic medication in early stages [23].
In this study, we aimed to investigate whether impairments in WM integrity occur starting from the early stage of psychosis, such as in patients with FEP, and used DT to reconstruct the SLF subdivisions (i.e., SLF I, II, and III) while resolving crossing fibers [20]. On the basis of previous studies, which have consistently reported WM impairments in schizophrenia patients but have presented inconsistent findings for FEP patients [14-17], we hypothesized that WM integrity in FEP patients varies across the SLF subdivisions. Specifically, given the role of SLF II as a modulator between SLF I and SLF III [10], we expect that FEP patients would show more prominent WM disruption in SLF II than would HCs.
METHODS
Participants
One hundred and ten HCs and 39 FEP patients participated in this study. FEP patients were recruited from the inpatient and outpatient clinics of the Department of Neuropsychiatry at Seoul National University Hospital (SNUH). For patients with FEP, diagnosis was performed using the Structured Interview for the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition Axis I Disorders (SCID-I), and the severity of the clinical symptoms was assessed by experienced psychiatrists using the Positive and Negative Syndrome Scale (PANSS) [24]. The FEP patients included in this study were between 15 and 40 years of age, had a schizophreniform disorder, schizophrenia or schizoaffective disorder, and had a duration of illness (DOI) of less than 2 years. HCs were recruited through internet advertisements and were screened via the SCID, nonpatient edition. Potential HCs who had any previous or current diagnosis of psychiatric disorders and any first-to third-degree biological relative with a diagnosed psychotic disorder were excluded. The intelligence quotient (IQ) was measured using the abbreviated version of the Korean version of the Wechsler Adult Intelligence Scale, Fourth Edition. The following exclusion criteria were applied for both FEP patients and HCs: intellectual disability (IQ<70); a diagnosis of a substance use disorder (except nicotine); a previous history of significant head trauma leading to loss of consciousness; history of seizures, meningitis, or previous neurosurgical interventions; or any other medical problems that affect cognitive functions.
Written informed consent was obtained from all participants after thorough explanation of the study procedure (HIRB no. H-1110-009-380). For participants younger than 18 years of age, informed consent was also obtained from their parents. This study was conducted in accordance with the Declaration of Helsinki (2013) and approved by the Institutional Review Board of SNUH (IRB no. H-2306-212-1445).
Image acquisition and preprocessing
DWI
Diffusion-weighted images (DWIs) in the axial plane were acquired using an echo-planar imaging sequence with a Siemens 3T Magnetom Trio Tim Syngo MR B17 scanner with a 12-channel head coil from a total of 149 participants with a repetition time of 11,400 ms, an echo time of 88 ms, a matrix size of 128×128, a field of view of 240 mm, and a voxel size of 1.9×1.9×3.5 mm3. Diffusion-sensitizing gradient echo encoding was applied in 64 directions using a diffusion-weighting b factor of 1,000 s/mm2. One volume without a gradient (i.e., b factor of 0 s/mm2) was acquired (B0 image) at the beginning of each scan.
DWI data were preprocessed using the FMRIB Software Library (FSL version 6.0.6.4; https://fsl.fmrib.ox.ac.uk/fsl/) to correct eddy current-induced distortions and artifacts due to bulk motions of the participant [25], followed by brain extraction [26] and visual inspection for major artifacts.
FA, MD, and RD maps
FA, MD, and RD data were generated by applying dtifit, a tool in the FSL Diffusion Toolkit, to the preprocessed DWI data. This process involved fitting tensors through a nonlinear least square fitting process to ensure accurate estimation of the diffusion parameters.
Whole-brain tractography
Spherical deconvolution
To investigate the specific fiber bundle of interest in each participant, we performed whole-brain tractography using DT, a DTI methodology, via the latest version of a program called StarTrack (https://mr-startrack.com/). With this program, whole-brain tractography is performed by analyzing DWI data with spherical deconvolution (SD) operations and computing fiber orientation distributions. We used the damped Richardson-Lucy algorithm to reduce isotropic background effects in the SD operations [20], and the parameters were as follows: ALFA=1.5, number of iterations=400, n=0.0015, and r=16.
Whole-brain DT
To perform whole-brain DT, modified Euler algorithms were used [27], and the parameters were as follows: number of runs=5, ABS threshold=0.0025, step size=1 mm, angle threshold=45°, minimum length=20 mm, and maximum length=300 mm [20,21].
Dissection and diffusivity measurements
Dissection of SLF I, II, and III
In TrackVis version 0.6.1 (http://trackvis.org/), the SLFs were dissected into a total of six sub-bundles for each participant (i.e., three sub-bundles bilaterally) in the native space of each participant on the basis of de Schotten et al. [8], who defined the SLF subdivisions on the basis of the following regions of interest: the left SLF I (SLF IL), right SLF I (SLF IR), left SLF II (SLF IIL), right SLF II (SLF IIR), left SLF III (SLF IIIL), and right SLF III (SLF IIIR).
The initial dataset included data from a total of 149 participants (39 FEP patients and 110 HCs). However, if any of the six subfibers could not be reconstructed, the data of the corresponding participant were excluded. Consequently, the final dataset for analysis consisted of data from 113 participants (32 FEP patients and 81 HCs). All dissection results were verified by two experienced raters, PHK and PIK.
Diffusion index calculation
In TrackVis, we registered the FA/MD/RD maps obtained from FSL with the High Angular Resolution Diffusion Imaging maps to increase accuracy [28]. These maps were overlaid with the fiber-traced SLF I, II, and III to calculate the diffusivity measurements. Consequently, we obtained diffusion indices for each hemisphere for each participant, resulting in a total of 18 diffusion indices.
Statistical analysis
All the statistical analyses were conducted using R version 4.2.3 (https://www.r-project.org/). Demographic differences between groups in terms of clinical characteristics (i.e., age, IQ, and years of education) were assessed using independent samples t tests, whereas categorical data (i.e., sex and handedness) were analyzed with chi-square tests. p<0.05 was considered to indicate statistical significance in all statistical analyses.
For the main effects (i.e., group differences in the diffusion index values extracted for each fiber), we conducted analyses of covariance (ANCOVAs), controlling for age, sex, and the total number of streamlines. These results were then corrected for multiple comparisons of the 18 measurements using the Bonferroni correction.
RESULTS
Demographics and clinical characteristics
No significant differences in demographic characteristics were observed between the groups. However, IQ (χ2=-5.00, p<0.001) and years of education (χ2=-2.26, p=0.026) were significantly different between the FEP patients and HCs, with HCs having higher values than patients with FEP did (Table 1).
Demographic and clinical characteristics of the participants
The number of streamlines
No significant differences in the number of streamlines in SLF I, II, and III in either hemisphere were observed between the groups. Although the number of streamlines in HCs was consistently greater than that in the FEP patients, these differences were not statistically significant (-0.298<χ2<-1.96, with all p>0.33). However, the total number of streamlines (χ2=-2.18, p<0.05) significantly differed between the two groups, with HCs having a greater number of streamlines than did FEP patients; thus, this variable was controlled as a covariate (Figure 1).
Between-group differences in the number of streamlines for the three subdivisions of the SLF (SLF I, II, and III) in both hemispheres. Each graph shows the number of streamlines for SLF I (1st row, cyan), SLF II (2nd row, blue), and SLF III (3rd row, purple) in FEP patients and HCs. Significant differences were obtained only for the total number of streamlines after Bonferroni correction. The black bar within the boxplot represents the group mean. *p<0.05. SLF, superior longitudinal fasciculus; FEP, first-episode psychosis; HCs, healthy controls.
Comparison of the reconstruction and diffusivity of SLF I, II, and III
Reconstructed SLF I, II, and III
In TrackVis, the reconstruction process involves reconstructing the subregions for each participant overlaid on the B0 image. An example of the SLF subregions reconstructed via DT for a representative participant over the brain template (Figure 2A-C) and B0 image (Figure 2D-F) are presented. Each fiber is depicted in different colors: SLF I in cyan, SLF II in blue, and SLF III in purple.
Reconstructing the three subdivisions of the SLF (SLF I, II, and III) through deterministic tractography in an individual. SLF I (cyan), SLF II (blue), and SLF III (purple) are depicted within the axial (superior) (A) and coronal (posterior) (B) views, representing both hemispheres. The focus of the sagittal view (C) is the left hemisphere (D, E, and F) present the same views as above but for the B0 image. SLF, superior longitudinal fasciculus.
Group differences in the 18 diffusion indices
The ANCOVA results for group differences in WM integrity across the six subcomponents in terms of the three diffusion indices (FA, MD, and RD) are summarized in Table 2. Among the 18 diffusion indices, after controlling for age, sex, and the total number of streamlines as covariates, none of the group effects on the three diffusion indices in the bilateral subdivisions of the SLF were determined to be significant following multiple-comparison corrections (Figure 3).
FA, MD, and RD for the three subdivisions of the SLF (SLF I, II, and III) in both hemispheres between the two groups
Between-group differences in the FA, MD, and RD for the three subdivisions of the SLF (SLF I, II, and III) in both hemispheres. Each graph shows the FA, MD, and RD distributions for SLF I (1st row, cyan), SLF II (2nd row, blue), and SLF III (3rd row, purple) in FEP patients and HCs. ANCOVA, with age, sex, and the total number of streamlines as covariates, revealed no significant differences after Bonferroni correction in the diffusion measures across SLF regions. The black bar within the boxplot represents the group mean. FA, fractional anisotropy; MD, mean diffusivity; RD, radial diffusivity; SLF, superior longitudinal fasciculus; FEP, first-episode psychosis; HCs, healthy controls; ANCOVA, analyses of covariance.
DISCUSSION
In this study, we investigated WM alterations in SLF subdivisions (i.e., SLF I, II, and III) using DT and virtual dissection to precisely reconstruct SLF tracts in patients with FEP and HCs. We found no significant changes in WM integrity within the three SLF subdivisions in the FEP patients, as evaluated by diffusivity indices (i.e., FA, MD, and RD). This result may suggest that the WM connectivity among SLF subdivisions is relatively preserved in FEP patients who are in the early stages of psychotic disorders.
In the present study, the WM integrity of FEP patients did not significantly differ across 18 diffusion indices in the SLF subdivisions (SLF I, II, and III) in either hemisphere when the DT method and virtual dissection were used for analysis. However, previous studies using different DTI analysis methods have reported alterations in SLF subdivisions in patients with FEP [15,16]. One study reported reduced WM integrity in both the right SLF II and the left SLF III [16], whereas another study reported changes in only the right SLF II [15]. These specific alterations contrast with our findings, in which none of the 18 diffusion indices, including those in the right SLF II and left SLF III, showed significant group differences. Notably, this study examined all six SLF subdivisions bilaterally, thereby offering a more comprehensive assessment than previous studies. These inconsistencies may stem from methodological differences. In previous studies, TBSS and VBA methods have mainly been used; these approaches are useful for assessing WM integrity at the voxel level but are limited in precisely identifying individual WM tracts, particularly SLF subdivisions [12]. Additionally, variations in symptom severity may contribute to these discrepancies. For example, in the study by Guo et al. [15], the patients with FEP had high PANSS scores across all the subscales, whereas the FEP patients in the present study presented with milder symptoms. Furthermore, Zhou et al. [16] used the Brief Psychiatric Rating Scale in their study, limiting direct comparisons with the present work. More severe symptoms have been linked to more pronounced structural alterations [29], which may explain the absence of significant WM differences in our cohort. Taken together, methodological differences, such as the use of DT with virtual dissection [20,21], and the milder symptoms in our FEP patients may explain the lack of significant SLF alterations [29], although subtle WM changes could emerge with symptom progression.
Previous DTI studies on chronic schizophrenia patients have reported significant alterations in SLF subdivisions [14,22], which contrasts with our findings in FEP patients. Both of these previous studies reported reduced WM integrity in the right SLF II and right SLF III, whereas another study reported changes across both hemispheres in SLF I, II, and III [14]. This discrepancy may be attributed to several key factors, such as age, DOI, medication exposure, and methodological differences (e.g., DT and probabilistic tractography). Given that schizophrenia is a progressive disorder with increasing structural changes over time [30,31], WM, particularly the SLF, may remain dynamic rather than stable, especially in early adulthood. The SLF continues to mature well into the late 20s, with SLF II developing more slowly than the other subdivisions do [32]. This ongoing maturation may make the SLF more susceptible to subtle alterations that are not yet detectable in the early stages of psychotic disorders, such as FEP, but could become more pronounced with the progression of such chronic diseases [33]. Additionally, a longer DOI has been associated with greater reductions in both FA34 and gray matter volume [35,36], suggesting that structural changes in SLF subdivisions may emerge as the disorder progresses. Medication exposure is another critical factor, as studies involving FEP patients have shown that higher doses of antipsychotic drugs are linked to cortical thinning and brain volume loss [37,38], whereas short-term treatment increases FA in the SLF [39]. Since these findings indicate that medication-related structural changes are already detectable in early stages, longer exposure to antipsychotic medication in chronic schizophrenia patients may contribute to greater WM alterations. Furthermore, methodological differences may contribute to inconsistencies across studies. Leroux et al. [22] reported that chronic schizophrenia patients exhibited reduced WM integrity in the right SLF II and III via probabilistic tractography. Probabilistic tractography can be used to detect subtle structural disruptions not captured by DT40; furthermore, more precise fiber estimation [20] and more sensitive detection of real WM properties, including diffusion indices, can be achieved [21]. While DT enables anatomically precise reconstructions, it may be less sensitive than probabilistic tractography for detecting subtle or spatially diffuse WM abnormalities, which could account for the lack of significant group differences observed in this study. These factors suggest that while SLF subdivisions may appear intact in the early stage of psychotic disorders, such as FEP, progressive WM changes may occur with chronic disease progression. Future longitudinal studies are needed to track these structural changes over time and clarify their relationship with chronic illness.
The relative preservation of SLF subdivisions in FEP patients may reflect functional compensation and neurodevelopmental factors. Each SLF subdivision plays a distinct role in cognition, as SLF I is associated with attention and working memory, SLF II is associated with visuospatial processing, and SLF III is associated with language and spatial awareness [9]. The disconnection hypothesis suggests that schizophrenia results from disrupted network integration rather than from immediate structural abnormalities [3]. Thus, in the early stages of psychotic disorders, such as FEP, functional impairments may arise before WM disruptions are detected. Consistent with this, Fornito et al. [41] showed that global network topology remains largely intact in early psychosis, indicating that while subtle network inefficiencies may emerge, large-scale WM structures are preserved until later stages of illness progression. While structural WM integrity appears to be preserved in FEP patients, the presence of subtle network inefficiencies suggests that functional disruptions may precede measurable structural degeneration. Future longitudinal studies are needed to clarify how these early network alterations contribute to long-term structural decline.
This study had several limitations. First, the use of a single B0 image and nonisotropic voxel shapes may have introduced partial volume effects and tract reconstruction bias in the WM tract reconstruction process [42], and uncontrolled cardiac pulsations could have caused body movements. Despite these issues, we visually inspected the DWIs for significant motion artifacts. Second, this work was a cross-sectional study and did not include chronic schizophrenia patients; therefore, we cannot determine whether SLF subdivisions deteriorate as the illness progresses or remain intact in certain subtypes of FEP. While previous studies have reported SLF alterations in chronic schizophrenia patients [14,22], longitudinal research is needed to clarify when such changes emerge and whether they occur across all patients with psychosis. Third, although our sample size was larger than that of similar tractography studies, the final number of participants after exclusions, partly due to unsuccessful tract reconstruction (32 FEP patients and 81 HCs), may still have been insufficient for detecting subtle group differences in diffusion indices, particularly given the inherent variability of DTI measures. Fourth, the study’s statistical power may also have been limited by the extensive multiple comparisons across 18 diffusion indices, increasing the risk of type II errors. Additionally, antipsychotic medication was not included as a covariate in the analysis, despite its known influence on WM microstructure [29]. Fifth, the relatively mild symptom severity in our FEP cohort may have limited our ability to detect group differences, as more pronounced structural alterations are typically observed in patients with higher symptom burdens [29]. Sixth, the FEP group had significantly lower IQ and fewer years of education than HCs, but these variables were not included as covariates in the main ANCOVA because they were considered characteristics of the disease course in psychotic disorders. Although age was controlled as a related factor, the exclusion of IQ and education may have biased the interpretation of group differences in WM integrity [14]. Seventh, we examined SLF as a whole by averaging values across the three subdivisions rather than reconstructing the full tract anatomically. While this approach revealed no significant group differences (0.05<F<1.08, with all p>0.30), it may not fully reflect whole-tract geometry. The correlations between subdivisions and whole-tract metrics were generally strong but varied across measures (r=0.83 or low as r=0.32), suggesting that averaging may obscure localized differences [8]. Finally, although no group differences were significant, we examined the effect sizes to inform future research. The largest observed generalized eta squared for group across all diffusion measures was 0.016, indicating a small effect. This value may assist future studies in estimating appropriate sample sizes. Other factors, such as childhood trauma [43], exercise [44], alcohol use [45], and tobacco use [46], were not the focus of the present study but should be investigated in future work.
In summary, our results suggest that WM integrity in the subdivisions of the SLF remains largely preserved in FEP patients, indicating that structural connectivity may not be significantly disrupted in the early stage of psychotic disorders. Given the role of the SLF in cognitive functions such as attention, working memory, and language processing, the preserved structural integrity of the SLF may imply that early dysfunction in these networks arises from functional rather than anatomical disruptions, as suggested by previous studies. While SLF alterations in chronic schizophrenia patients have been reported in prior studies, our findings suggest that such changes may develop as the disease progresses. Methodological differences across studies, including tractography techniques and sample characteristics, may also contribute to inconsistencies in the reported SLF alterations. To clarify the trajectory of SLF changes in patients with psychotic disorders, future longitudinal studies are essential. By employing more advanced tractography methods and integrating multimodal imaging approaches, these studies can offer deeper insights into the evolving nature of WM alterations in patients with schizophrenia.
Notes
Availability of Data and Material
The datasets generated or analyzed during the study are available from the corresponding author on reasonable request.
Conflicts of Interest
Jun Soo Kwon, a contributing editor of the Psychiatry Investigation, was not involved in the editorial evaluation or decision to publish this article. All remaining authors have declared no conflicts of interest.
Author Contributions
Conceptualization: Jongrak Kim, Minah Kim, Jun Soo Kwon. Data curation: Jongrak Kim, Hyungyou Park, Inkyung Park, Jiseon Jang. Formal analysis: Jongrak Kim, Hyungyou Park, Inkyung Park. Funding acquisition: Jun Soo Kwon. Investigation: Jongrak Kim, Hyungyou Park, Inkyung Park. Methodology: Jongrak Kim, Hyungyou Park, Inkyung Park. Project administration: Minah Kim, Jun Soo Kwon. Supervision: Minah Kim, Jun Soo Kwon. Validation: Hyungyou Park, Inkyung Park, Minah Kim, Jun Soo Kwon. Visualization: Jongrak Kim, Hyungyou Park, Inkyung Park. Writing—original draft: Jongrak Kim, Hyungyou Park, Inkyung Park, Minah Kim. Writing—review & editing: Jiseon Jang, Minah Kim, Jun Soo Kwon.
Funding Statement
This research was supported by the Brain Science Convergence Research Program through the National Research Foundation of Korea (NRF) and the KBRI basic research program through the Korea Brain Research Institute, funded by the Ministry of Science & ICT (Grant Nos. RS- 2023-00266120 and 25-BR-05-05).
Acknowledgments
None