Probing bundle-wise abnormalities in patients infected with human immunodeficiency virus using fixel-based analysis: new insights into neurocognitive impairments

Abstract

Background:

Changes in white matter (WM) underlie the neurocognitive damages induced by a human immunodeficiency virus (HIV) infection. This study aimed to examine using a bundle-associated fixel-based analysis (FBA) pipeline for investigating the microstructural and macrostructural alterations in the WM of the brain of HIV patients.

Methods:

This study collected 93 HIV infected patients and 45 age/education/handedness matched healthy controls (HCs) at the Beijing Youan Hospital between January 1, 2016 and December 30, 2016.All HIV patients underwent neurocognitive evaluation and laboratory testing followed by magnetic resonance imaging (MRI) scanning. In order to detect the bundle-wise WM abnormalities accurately, a specific WM bundle template with 56 tracts of interest was firstly generated by an automated fiber clustering method using a subset of subjects. Fixel-based analysis was used to investigate bundle-wise differences between HIV patients and HCs in three perspectives: fiber density (FD), fiber cross-section (FC), and fiber density and cross-section (FDC). The between-group differences were detected by a two-sample t-test with the false discovery rate (FDR) correction (P <0.05). Furthermore, the covarying relationship in FD, FC and FDC between any pair of bundles was also accessed by the constructed covariance networks, which was subsequently compared between HIV and HCs via permutation t-tests. The correlations between abnormal WM metrics and the cognitive functions of HIV patients were explored via partial correlation analysis after controlling age and gender.

Results:

Among FD, FC and FDC, FD was the only metric that showed significant bundle-wise alterations in HIV patients compared to HCs. Increased FD values were observed in the bilateral fronto pontine tract, corona radiata frontal, left arcuate fasciculus, left corona radiata parietal, left superior longitudinal fasciculus III, and right superficial frontal parietal (SFP) (all FDR P <0.05). In bundle-wise covariance network, HIV patients displayed decreased FD and increased FC covarying patterns in comparison to HC (P <0.05) , especially between associated pathways. Finally, the FCs of several tracts exhibited a significant correlation with language and attention-related functions.

Conclusions:

Our study demonstrated the utility of FBA on detecting the WM alterations related to HIV infection. The bundle-wise FBA method provides a new perspective for investigating HIV-induced microstructural and macrostructural WM-related changes, which may help to understand cognitive dysfunction in HIV patients thoroughly.

Introduction

Human immunodeficiency virus (HIV)-associated neurocognitive disorder (HAND) commonly occurs in HIV patients and induces cognitive deficiencies in various cognitive domains, eventually leading to a high risk of dementia. A previous study revealed that attention/working memory deficit, motor slowness, and apathy were widespread in HIV patients.^[1] These cognitive symptoms may be attributed to chronic neuroinflammation, gliosis, perivascular macrophages or microglial cell associated with diffuse pallor,^[1] and HIV-induced axonal injury.^[2]

Some studies have confirmed that white matter (WM) integrity may contribute to cognitive dysfunctions in early non-demented HIV patients.^[3] Moreover, the early detection of WM abnormalities in HIV patients is advantageous for investigating sustained inflammation and poor cognitive performance.^[4,5] The findings of these studies indicated that WM abnormalities might be a strong indicator and biomarker of early HIV disease progression.^[6] The association of WM abnormalities with cognitive functions has been explored in healthy individuals and various neurological patients.^{[7,8,9,10,11]} Diffusion magnetic resonance imaging (dMRI) allows the non-invasive detection of changes in WM via the quantification of water levels diffused in brain tissue.^{[12,13,14,15]} Previous studies have revealed that HIV patients exhibited lower levels of fractional anisotropy (FA) and higher levels of diffusivity in intracerebral WM tracts and projection pathways,^[16,17,18] and these WM abnormalities helped predict cognitive impairments, especially in the attentional domain.^[19] However, most studies used voxel-wise analysis methods to analyze dMRI data, which may be influenced by fibers crossing in one voxel.^{[20,21,22,23]} Moreover, conventional voxel-based analysis is also limited by assuming a single-fiber configuration within a voxel,^[24] which reduces the sensitivity of detection of microstructure-related phenomena, such as axon degeneration, fiber density (FD), and demyelination. Recent developments in dMRI acquisition and analysis have increased the ability to overcome these non-specific measurements as they help to determine more specific microstructural features in the neuronal tissue, such as the apparent FD.^[25] More recently, it was proposed that fixel-based analysis (FBA) could delineate fiber bundles with different orientations within one voxel.^[22] This provides an effective way to map the FD and investigate axonal degeneration in HIV+ individuals,^[26] which found that the FBA metrics of the entire brain decreased in the internal capsule and were positively associated with the attention ability.

Therefore, this study aimed to investigate the changes of fiber bundles in HIV patients and to determine the relationship between the altered fiber bundles and various cognitive functions.

Methods

Ethical approval

The Research Ethics Review Board of Beijing Youan Hospital approved this study (No. 2017032), and it was conducted in accordance with the Declaration of Helsinki. All participants were recruited at Beijing Youan Hospital and provided written consent.

Subjects and neurocognitive tests

This study enrolled 183 subjects, including 93 HIV-infected patients and 45 age/education/handness-matched healthy controls (HC). HIV patients were infected through sexual transmission.

The exclusion criteria included neurological disorders, a history of opportunistic central nervous system infection, traumatic brain injury, substance or alcohol abuse, depression, and hepatitis B/C coinfection. The exclusion criteria for subjects with a history of substance or alcohol abuse were as follows: (1) negative urine screen for substance abuse (alcohol, marijuana, methamphetamine, and amphetamine); and (2) no history of drug or alcohol abuse (the amount of alcohol consumed was more than 35 standard drinks per week).

All HIV+ patients underwent clinical evaluations and laboratory tests, including tests to assess the cluster of differentiation 4 (CD4)+ T cell count, CD4+ and CD8+ T cell ratio, and duration of combination anti-retroviral therapy (cART) at the time-point of MRI scan.

In addition, each HIV-infected participant underwent neurocognitive and motor function evaluations before the MRI scan, including the following sub-domains: language fluency (animal verbal fluency test [AFT]), attention (continuous performance test-identical pair [CPT-IP]; Wechsler memory scale [WMS-III]; paced auditory serial addition test [PASAT]), executive function (Wisconsin card sorting tests [WCST-64]), memory (Hopking verbal learning test [HVLT-R]; brief visuospatial memory test [BVMT-R]), speed of information processing (trail marking test A [TMT-A]), and motor function (grooved pegboard test, dominant and non-dominant). All neurocognitive and motor function assessments lasted 40–60 min. An experienced neurologist had received rigorous neuropsychological training before the study to guarantee the quality of these evaluations.

MRI acquisition

MRI scanning of participants was performed using a 3T Siemens scanner (TimTrio, Erlangen, Germany) and a 32-channel head coil. Importantly, all scans were performed on the same machine at the same site, thereby eliminating the inconsistencies and artifacts resulting from using multiple scanners.^[27] An optimized diffusion sequence was used to acquire dMRI data for each participant using the following parameters: 25 axial slices; field of view (FOV) = 23 cm × 23 cm; repetition time (TR)/echo time (TE) = 3300/90 ms; slice thickness = 4 mm; acquisition matrix = 128 × 128; and voxel size = 2 mm × 2 mm × 4 mm. Diffusion-sensitizing gradients were applied along 60 non-collinear directions (b = 1000 s/mm²), and 3 acquisitions were performed without diffusion weighting (b = 0). The total acquisition time for diffusion imaging was 3 min and 39 s. T1-weighted images were collected using the following: 3-dimensional magnetization-prepared rapid acquisition gradient echo (MPRAGE) sequence; TR/TE = 1900/2.52 ms; inversion time (TI) = 900 ms; slice thickness = 1 mm; voxel size = 1.0 mm × 1.0 mm × 1.0 mm; flip angle = 9°.

Data processing

The dMRI data were corrected thoroughly to exclude potential artifacts using the following steps. First, magnetic field inhomogeneity-induced distortions (leading to intensity loss and voxel shifts) were corrected using a software library tool for the functional MRI of the brain (https://fsl.fmrib.ox.ac.uk/fsl).^[28] Next, the previously proposed distortion correction method^[29] was applied with two main steps: (1) eddy current-induced artifact correction and (2) susceptibility-induced artifact correction. The eddy current-induced distortions of diffusion-weighted (DW) images acquired with opposite phase encoding directions (PEDs) were corrected by affine registration, followed by the coarse non-linear registration of all DW images to their corresponding non-DW image. We then used the spherical mean images (SMIs) of DW images corrected for eddy current distortions in a multi-contrast registration framework to correct susceptibility-induced distortions. After registration, the signal pileup and signal dropout were updated by modulating the undistorted images with the Jacobian determinant of the estimated displacement fields. Because we could not acquire q-space samples with opposing PEDs, we fitted the spherical harmonics (SHs) to the DW images acquired for a PED, and then acquired the DW images of the opposing PED. Finally, the corrected DW images of the opposing PEDs were combined by computing the harmonic mean. A semi-automated quality control (using in-house developed Matlab scripts) process was conducted on all diffusion images. Individuals with diffusion images exhibiting apparent signal drops were excluded from the following analyses. All gradient directions were retained for analysis for the remaining subjects.

Fixel-based analysis

A study-specific, group-wise, FBA was conducted using a well-established pipeline to parcellate the whole-brain fiber configuration into multiple fixel configurations.^[23] Before conducting this analysis, the diffusion images were subjected to the following pre-processing steps: (1) Bias field correction was implemented to eliminate low-frequency and low-intensity non-homogeneities across the image. Field correction of DWI bias was performed by first estimating the bias field from the DWI b = 0 data, and applying the field to correct all DW volumes. (2) Global intensity normalization was performed across participants with the median WM intensity value of b = 0 using tools implemented in MRtrix3 (www.mrtrix.org).^[30,31] (3) A group response function was calculated from the fiber response functions of all participants, and it reflected the signal expected from a voxel containing a single, typical fiber bundle.^[32] Individual fiber response functions were estimated using the convenient and reliable "Tournier" algorithm,^[33] and were subsequently averaged to determine the group response function. (4) DW images were subjected to upsampling to an isotropic size of 1 mm × 1 mm × 1 mm, to increase the anatomical contrast and improve downstream template building, registration, tractography, and statistics. (5) Constrained spherical deconvolution (CSD), a technique that uses the response function to estimate the distribution of fiber orientations contained within each voxel,^[32,34] was used to estimate the fiber orientation distribution function (FOD).^[32] A subset of 30 participants (15 HC) individuals and 15 HIV patients) were randomly selected to generate a study-specific FOD template via the linear and non-linear registration of FOD images.^[35] (6) FOD images from all participants were non-linearly registered to this template, and three fixel metrics were calculated, i.e., the apparent FD, fiber cross-section (FC), and fiber density and cross-section (FDC).^[22]

Whole-brain tractography in population template

Whole-brain tractography was performed using a study-specific FOD template to generate a population tractogram. An improved second-order integration over fiber orientation distributions (iFOD2) was employed to facilitate more accurate fiber reconstruction in heavily curved regions and thus enhance the anatomical plausibility of the results.^[36] The direction of each step during the tracking process was determined by sampling from the FOD at the current position. The probability of obtaining a fiber with a particular orientation was proportional to the amplitude of the FOD along that direction. Subsequently, iFOD2-based fiber tracking was performed with the following settings: default step size = 0.8 mm, maximum angle between successive steps = 90°, maximum length = 160 mm, minimum length = 10 mm, cutoff FA value = 0.01, and unidirectional tracking, and it resulted in 20,000,000 streamlines. Next, whole-brain tractograms were processed using the spherical deconvolution informed filtering of tractograms (SIFT)^[37] procedure introduced in MRtrix3, resulting in 2,000,000 streamlines.

Identification of anatomical WM structures

Using a well-established computational pipeline,^[38] we segmented a whole-brain tractogram into WM bundles that included 56 tracts of interest, including those investigated in previous studies.^{[39,40,41,42]} Unsupervised fiber clustering was used to parcellate the whole-brain tractogram. Whole-brain fiber clustering was carried out in an automated manner based on the population tractogram without using any additional information regarding the brain anatomy. Each streamline was mapped to a Hilbert space via parameterization using cosine representation to increase the robustness of clustering without dependence on fiber length. The streamlines were then automatically clustered via the k-medoids clustering of groups, leading to an atlas of streamlines grouped into 2000 clusters. The streamlines in both cerebral hemispheres were clustered bilaterally. Outliers of each streamlined cluster were removed based on whether the Z-score of the distance between the streamline and its corresponding cluster center was over a typical statistical threshold of 2.5. We also excluded common false-positive streamlines that did not exist anatomically. Tractogram clusters (a total of 2000) were annotated using anatomical information defined in previous studies,^{[43,44,45,46]} including regarding (1) major long-range association and projection tracts, commissural tracts, and tracts connected to the brainstem and cerebellum; and (2) short and medium-range tracts organized into categories based on their association with the brain lobes. The following fiber bundles were generated and compared: the cingulum bundle (CB), superficial parietal temporal (SPT), dorso lateral prefrontal cortex (DLPC), superficial parietal occipital (SPO), corona radiata parietal (CRP), cortico bulbar tract (CBT), cortico pontine tract (CPT), superior longitudinal fasciculus I/II/III (SLF I/II/III), corona radiata frontal (CRF), dorso medial prefrontal cortex (DMPC), frontal aslant tract (FAT), fronto pontine tract (FPT), fornix (FX), arcuate fasciculus (AF), reticulo spinal tract (RST), cortico striatal pathway (CSP), cortico thalamic pathway (CTP), dentato rubro thalamic (DRT), cortico spinal tract (CST), external capsule (EC), orbito frontal cortex (OFC), SFP, superior thalamic radiation (STR), ventro lateral prefrontal cortex (VLPC), vertical occipital fasciculus (VOF), and parietal aslant tract (PAT).

Quantitative fixel measurement

FD, FC, and FDC values were computed across all white matter (WM) fixels in the same space for each participant. Re-orientation of fixel directions and correspondence of fixels with the template image was performed, as described previously.^[22] We computed FD, FC, and FDC values based on the FOD generated from multi-tissue multi-shell CSD. The FODs derived from CSD consist of multiple lobes representing individual fiber bundles. The amplitude of the FOD along a given fiber orientation was proportional to the radial DW signal and was, therefore, proportional to the intra-axonal volume of fibers. FD was calculated by integrating the FOD of each lobe.^[22,30] However, a change in intra-axonal volume may not always reflect a change in FD. Volume-related differences can also be accounted for by changes in morphology occurring perpendicular to the fiber orientation (i.e., a reduced area of fiber bundle cross-section, FC). FC is the determinant of the Jacobian matrix that needs to be warped from the subject to template space based on the spatial orientation of fixels. Finally, FDC, the product of FD and FC, provides a more robust measure of axonal integrity as it combines information from both metrics.^[22] We extracted the mean FD, FC, and FDC values for fixels along each tract of interest in the template space for all tracts and all subjects. These metrics served as our biomarkers of microscopic and macroscopic fiber integrity. FD depicts the microscopic morphology, while FC provides information on a macroscale. Furthermore, the measurement of FDC using a combination of parameters facilitates the comprehensive assessment of WM.

Statistical analysis

To analyze the differences in FD, FDC, and FC values between the HIV and HC groups, a permutation t-test^[47] involving 10,000 permutations was conducted for each fiber tract (left-/right-hemisphere), as implemented in Matlab R2021a (The MathWorks, Inc., Natick, MA, USA). Multiple comparison corrections were performed across the identified tract using the false discovery rate (FDR)^[48] to determine the corrected statistical significance, with a significance level of P <0.05.

The FD, FC, and FDC values of altered bundles between HC individuals and HIV patients helped to confirm the relationship with six types of cognitive and motor functions through partial correlation analysis, while the age and sex were regressed. Then, multiple comparison corrections were performed across the identified tract using FDR^[48] to determine the corrected statistical significance, P <0.05.

We further studied the inter-tract correlation between tract-specific average values after computing fixel metrics for all subjects. The covariance network was constructed for each fixel measurement using Pearson partial correlation between each pair of WM tracts while controlling for the effect of age, sex, and cognitive scores.

Estimation of sample sizes and outcomes

The sample size was estimated to distinguish between HIV-infected patients and HC individuals using G^* power software (www.gpower.hhu.de). The following parameters were set: alpha = 0.05, power = 0.9, and ANI/HC allocation ratio = 2:1. A total of 126 individuals were estimated to be sufficient to achieve the effect size.

Results

Patients

One hundred and thirty-eight subjects, including 93 HIV-infected patients and 45 age/education/handedness-matched HCs, were included in the study.

No significant differences were observed in the age and sex between participants with HIV in comparison to HC participants [Supplementary Table 1, http://links.lww.com/CM9/B725]. The 56 tracts of interest identified in the template space were displayed in Figure 1.

Figure 1.

The 56 tracts of interest (28 per hemisphere) were identified from the whole-brain tractogram generated in template space. Fiber tracts are color-coded with anatomical orientation, blue-to-yellow signifies inferior-to-superior orientation in sagittal view. Streamlines visualization was performed using COMEDI. COMEDI: Computational Medical Imaging Toolkit.

Group-wise comparisons

Significant differences occurred between groups in the identified tracts [Figure 2]. Significant differences (FDR P <0.05) were observed between groups, including the FPT, CRF, RST, CBT, CPT, and DRT in both hemispheres, based on the FD measured at the tract level [Supplementary Figure 1, http://links.lww.com/CM9/B693]. No significant differences could be found in FDC and FC values. We also found significant differences (FDR P <0.05) within a single hemisphere, including in the left AF, left CRP, left SLF III, left VLPC, left fornix, left FAT, right PAT, right SFP, and right SPT. The results showed significantly increased FD values in the HIV group in these tracts, but no significant changes could be found in FDC and FC values.

Figure 2.

The FD of some bundles was significantly higher in HIV patients than HCs by a two samples t-test with FDR correction. ^*Indicates a significant level with FDR corrected P <0.01. FD: Fiber density; FDR: False discovery rate; HCs: Healthy controls; HIV: Human immunodeficiency virus.

Inter-tract correlation analysis

The covariance networks were illustrated as heat maps [Figure 3], where the anti-diagonal elements represented the perfect correlation of a particular matrix for the tract with itself. In contrast, the off-diagonal elements represented the correlation of a given metric of one tract with another. There was a general decrease in inter-tract partial correlations for FD from HC individuals to HIV patients for most tracts of interest and a general increase in inter-tract partial correlations for FC from HC individuals to HIV patients. Interestingly, each type of WM pathway had different patterns: association pathways resulted in a more considerable change in the FD (decreased 31.64% over tracts, P <0.01) and FC (increased 19.86%, P <0.01) than the values observed for projection pathways (decreased 20.97% for FD, P <0.01, and increased 10.11% for FC, P <0.05).

Figure 3.

Cross-tract connectivity of FD, FDC, and FC for HIV and HCs. The dark boxes in the connection matrix represent non-significant (P >0.05) partial correlations after FDR correction for multiple comparisons. FC: Fiber cross-section; FD: Fiber density; FDC: Fiber density and cross-section; HCs: Healthy controls; HIV: Human immunodeficiency virus.

Tract-wise language/attention correlations

The tract-wise mean FC exhibited a significantly positive correlation with the language function [Figure 4] and significantly negative correlation with the attention function [Figure 5]. Both EC (L: r = 0.36, P = 0.007, R: r = 0.18, P = 0.038), both FAT (L: r = 0.16, P = 0.043, R: r = 0.19, P = 0.039), both SFP (L: r = 0.17, P = 0.039, R: r = 0.16, P = 0.042), and both SLF III (L: r = 0.21, P = 0.036, R: r = 0.23, P = 0.035) were correlated with the total language score. No significant correlation was found with the mean FD and FDC values of these tracts. The attention score was significantly correlated with both CBT (L: r = -0.21, P = 0.04, R: r = -0.19, P = 0.046), both DMPC (L: r = -0.21, P = 0.046, R: r = -0.2, P = 0.046), and both FPT (L: r = -0.25, P = 0.041, R: r = -0.32, P = 0.048). We also found that both attention and language scores were significantly correlated with both right FAT (attention: r = -0.25, P = 0.048, language: r = 0.19, P = 0.039), left CSP (attention: r = -0.22, P = 0.045, language: r = 0.25, P = 0.036), left STR (attention: r = -0.24, P = 0.044, language: r = 0.15, P = 0.046), and left SLF I (attention: r = -0.23, P = 0.044, language: r = 0.24, P = 0.037). No significant correlation was found between the mean FD and FDC values of these tracts.

Figure 4.

There were significant positive correlations between language scores and tract-wise FC in HIV. ^*Indicates a significant level with FDR corrected P <0.01. FC: Fiber cross-section; FDR: False discovery rate; HIV: Human immunodeficiency virus.

Figure 5.

There were significant negative correlations between attention scores and tract-wise FC in HIV. FC: Fiber cross-section; FDR: False discovery rate; HIV: Human immunodeficiency virus.

Discussion

In this study, we utilized a novel method involving a combination of tract-based and fixel-based analyses to investigate the abnormalities in bundle-associated WM in HIV patients. In comparison with HCs, significant alterations were observed only in the FD of HIV patients, including in the bilateral FPT, CRF, RST, CBT, CPT, and DRT tracts, and unilateral AF, CRP, SLF III VLPC, FX, FAT, PAT, SFP, and SPT. Furthermore, the FC was quantified in several WM tracts, and was found to be significantly correlated with language and attention-related cognitive functions in HIV patients. The inter-tract covariance network of FD and FC metrics was notably different between the HIV and HC groups, especially in association fiber tracts.

dMRI has been used to accurately detect the effect of HIV on brain WM abnormalities and determine the relationship between cognitive functions and other factors, such as smoking, age, and drug use.^{[40,49,50,51]} Our findings not only revealed CPT, FX, and SLF abnormalities in HIV patients, which have been reported in previous DTI studies,^{[52,53,54,55]} but also reported some new findings in FPT, CRF, RST, CBT, CPT and VLPC tracts. Indeed, recent fixel-based analytical approaches have been widely applied for characterizing a specific fiber population within a voxel (i.e., a "fixel").^[22] Modeling individual fibers at the sub-voxel level in this manner could help us make more accurate measurements and understand tissue degeneration in various disorders more effectively.^[26] However, the current "fixel" was still limited to providing local insight within the sub-voxel. In this work, we focused on individuals living with HIV in Asia with variance in HAND because of other specific features.^[56]

Most of the widespread abnormalities in WM tracts are associated with the frontal and temporal lobes, which play key roles in maintaining cognitive functions and have been implicated in HIV patients and animal models.^[57,58,59] Interestingly, AF and SLF are critical portions of the dorsal pathway associated with language function. AF connects the Broca's area to the middle superior temporal gyrus, and SLF is highly correlated with sentence repetition.^[60,61]

Furthermore, we also found that FC values of FAT, CSP, STR, and SLF I were positively correlated with language function and negatively correlated with attention. It has been validated that FAT is crucial for motor behavior, balance, and verbal fluency tasks.^[62] HIV does not infect neurons directly. It leads to cytokine/chemokine production, which might result in neuronal injury in fiber tract pathways and weaken motor network functioning and verbal fluency. In one previous study, HIV+ subjects presented with greater attention and learning deficits that were associated with the immunological molecule levels in the blood and cerebro-spinal fluid (CSF).^[63] Furthermore, the results for FD in our study showed that changes in WM occurred in CST, FPT, SLF, and several other tracts involved in the limbic and motor pathways, which have also been reported in other types of cognitive disorders,^{[64,65,66,67]} but not in HIV. Therefore, this highlights that HIV leads to cognitive deficits in several systems, such as those involving cooperation between the language, motor, and attention systems, instead of a single isolated domain.

This study have several limitations: (1) DW images with a b-value of b = 1000 s/mm² were acquired. The low b-value might lead to insufficient suppression of the extra-axonal compartment, which would complicate the quantification of the FD. (2) Multi-shell and higher b-value data are ideal for accurately estimating fiber orientations, but some studies emphasized that data with a single b-value could also produce relatively robust results.^[24,34] (3) HIV patients had received medication at different levels, which may indirectly affect axon variations. However, there is currently no clear evidence showing that any medication could change the structure of the axon in the short term.

Few studies have focused on the inter-tract correlation of FBA metrics in HIV patients, and our results revealed that association fibers were more severely affected than projection/superficial fibers in HIV patients. Association pathways include several critical tracts that interconnect the frontal lobe with other lobes in one hemisphere, such as the superior longitudinal fasciculus and uncinate fasciculus, which are widely reported abnormalities in HIV patients. However, the inter-tract correlation of FD values was decreased, while that of FC was increased. The decrease in inter-tract FD correlation indicated the occurrence of heterogeneous FD alterations in association tracts, and the increase in the extent of inter-tract FC correlation implied that similar changes in FC occurred in association tracts. These results demonstrated that FC and FD could simultaneously reflect HIV-related alterations in WM. A small animal model^[68] revealed that in comparison with the HC, the HIV transgenic rat model displayed increased extracellular space between WM tracts and axonal degeneration, which may also support inter-tract FC findings. In summary, our study provided a covarying inter-tract view for HIV patients using FBA metrics, which may help to understand the neurological mechanism of HAND more effectively.

Funding

The study was supported by grants from the China's National Natural Science Foundation (Nos. 62201265, 61936013), the Natural Science Foundation of Beijing, China (No. 7212051), the Beijing Municipal Commission of Education (No. KM202010025025), and open fund project of Beijing Key Laboratory of Fundamental Research on Biomechanics in Clinical Application (No. 2023KF05).

Supplementary Material

cm9-136-2178-s001.pdf ^{(254.7KB, pdf)}

cm9-136-2178-s002.doc ^{(21.5KB, doc)}