Skip to main content
Journal of the American Heart Association: Cardiovascular and Cerebrovascular Disease logoLink to Journal of the American Heart Association: Cardiovascular and Cerebrovascular Disease
. 2026 Aug 30;15(19):e046071. doi: 10.1161/JAHA.125.046071

In Vivo Tensor‐Valued Diffusion Magnetic Resonance Imaging Reveals Isotropic‐Anisotropic Kurtosis Mismatch in a Rat Middle Cerebral Artery Occlusion Stroke Model

Mingyao Liang 1,2,3,#, Jianyu Yuan 2,3,4,#, Chaogang Tang 1,#, Pingfu Wang 2,3,4,5,#, Tingting Gu 6, Zhao‐Yu Deng 1, Weiming Sun 1, Xinming Yang 1, Yaohui Tang 6,✉, Ye Li 2,3,4,5,✉, Yi He 2,3,4,5,✉
PMCID: PMC13745115  PMID: 42669929

Abstract

Background

Tensor‐valued diffusion magnetic resonance imaging enables the separation of total kurtosis into isotropic and anisotropic components, offering improved specificity over conventional diffusion kurtosis imaging. In this study, we introduce a novel framework for detecting anisotropic‐isotropic kurtosis mismatch and evaluate its relevance in experimental ischemic stroke.

Methods

Tensor‐valued diffusion magnetic resonance imaging was performed in a rat model of middle cerebral artery occlusion. Metrics, including anisotropic mean kurtosis and isotropic mean kurtosis, were quantified to assess microstructural tissue heterogeneity. Histologic validation was conducted using coregistered tissue sections.

Results

We observed significant mismatches between mean diffusivity, mean kurtosis, as well as specifically in anisotropic mean kurtosis and isotropic mean kurtosis in ischemic regions. Notably, anisotropic mean kurtosis and isotropic mean kurtosis exhibited statistically significant alterations compared with contralateral regions. Moreover, these metrics strongly correlated with pathologic findings. The observed mismatch highlighted microstructurally distinct subregions within the lesion.

Conclusions

This study establishes an in vivo framework for characterizing anisotropic‐isotropic kurtosis mismatch in an experimental stroke model. By quantifying the isotropic mean kurtosis/anisotropic mean kurtosis mismatch, tensor‐valued diffusion magnetic resonance imaging provides a more specific mechanistic characterization of diffusion heterogeneity than conventional diffusion tensor imaging and diffusion kurtosis imaging metrics, supporting its use as a preclinical tool for investigating microstructural alterations in ischemic brain tissue.

Keywords: anisotropic‐isotropic mismatch, diffusion kurtosis imaging, ischemic stroke, middle cerebral artery occlusion, tensor‐valued diffusion magnetic resonance imaging

Subject Categories: Magnetic Resonance Imaging (MRI), Ischemia, Imaging


Nonstandard Abbreviations and Acronyms

DKI

diffusion kurtosis imaging

DWI

diffusion‐weighted imaging

GM

gray matter

HA

histologic anisotropy

HE

hematoxylin and eosin

HI

histologic isotropy

LFB

Luxol fast blue

LTE

linear tensor encoding

MCAO

middle cerebral artery occlusion

MD

mean diffusivity

MK

mean kurtosis

MKA

anisotropic mean kurtosis

MKI

isotropic mean kurtosis

ROI

region of interest

STE

spherical tensor encoding

WM

white matter

Research Perspective.

What ls New?

  • Tensor‐valued diffusion magnetic resonance imaging enables in vivo decomposition of mean kurtosis into isotropic and anisotropic components, providing greater microstructural specificity than conventional diffusion or kurtosis imaging in acute ischemic stroke.

  • Quantifying the mismatch between isotropic mean kurtosis and anisotropic mean kurtosis provides complementary insight into microstructurally distinct compartments within ischemic tissue, as supported by histologic validation.

What Question Should Be Addressed Next?

  • Future research should optimize acquisition strategies to improve sensitivity and reproducibility, and integrate the metrics into clinical mismatch paradigms to refine stroke‐imaging–based treatment selection.

Ischemic stroke is a leading cause of disability and death worldwide, underscoring the need for precise neuroimaging to support early diagnosis, guide therapy, and improve patient outcomes. 1 , 2 Diffusion‐weighted magnetic resonance imaging (DWI) has revolutionized acute stroke assessment by enabling early detection of ischemic lesions within minutes of symptom onset. 3 , 4 Conventional diffusion tensor imaging (DTI), which derives metrics such as mean diffusivity (MD) and fractional anisotropy (FA), provides insights into water mobility and directional tissue organization. However, DTI relies on the assumption of Gaussian diffusion, limiting its sensitivity to the complex microstructural changes after stroke. 5 Although diffusion restriction (manifested as decreased MD) indicates cytotoxic edema in acute ischemia, DTI cannot fully resolve tissue heterogeneity or distinguish reversible injury. 6

To overcome these limitations, advanced diffusion magnetic resonance imaging (MRI) techniques have been developed to characterize ischemic tissue beyond the capabilities of DTI. 5 , 7 Diffusion kurtosis imaging (DKI), an extension of DTI, quantifies non‐Gaussian diffusion of water and is more sensitive to microstructural heterogeneity. 3 , 8 Specifically, kurtosis measures the degree to which diffusion is restricted or multicompartmental, a property sensitive to microstructural barriers, such as cell membranes, neurites, and edema. DKI has demonstrated improved sensitivity in detecting acute ischemic regions compared with DTI. 7 Mean kurtosis (MK) is typically increased in infarcted areas, providing additional contrast with the contralateral hemisphere. 9 Notably, infarct volumes defined by MK often appear smaller than those delineated by DWI or MD, suggesting that kurtosis imaging may more precisely identify the damaged infarct core within the broader diffusion‐restricted region. 10

However, DKI cannot distinguish contributions from different microstructural features, limiting biological specificity. To address this limitation, tensor‐valued diffusion MRI has emerged as a powerful technique that leverages advanced b‐tensor encoding schemes. 11 , 12 , 13 Unlike conventional linear encoding (single direction), tensor‐valued encoding uses multiple b‐tensor shapes (such as linear, planar, and spherical) to probe diffusion properties in a multidimensional manner. 14 , 15 This multidimensional approach enables the decomposition of total kurtosis into anisotropic MK (MKA), which reflects orientation‐dependent microstructural features, such as axonal coherence or dispersion, and isotropic MK (MKI), which captures orientation‐independent restrictions caused by phenomena such as cell swelling or high cell density. 16 , 17 These indexes do not imply that brain parenchyma is globally isotropic or anisotropic, but rather provide quantitative, region‐specific measures that allow comparison with diffusion‐derived kurtosis metrics. 14 , 15 This decomposition offers a more nuanced view of tissue microstructure, potentially enhancing the specificity of diffusion MRI in pathologic conditions.

In the context of acute ischemic stroke, the brain can be divided into a damaged infarct core and a surrounding ischemic penumbra, which is functionally impaired but potentially salvageable. 18 , 19 , 20 Several mismatch paradigms have been developed to delineate these tissue states. Classical relaxometric approaches include the DWI/T2–fluid‐attenuated inversion recovery mismatch, which has been widely adopted as a surrogate marker of stroke onset time and patient selection for reperfusion therapy beyond conventional windows. Similarly, mismatch between Apparent Diffusion Coefficient (ADC) and absolute T2 has been used to capture evolving ischemic changes at an early stage. These relaxation‐based methods primarily reflect changes in water content and tissue relaxation properties, providing macroscopic information about lesion evolution. 21 , 22 , 23

At the microscopic level, diffusion‐based approaches have further evolved: Two major imaging mismatch paradigms are widely recognized: (1) perfusion‐diffusion mismatch (Perfusion‐Weighted Imaging (PWI) Animal Research: Reporting of In Vivo Experiments‐DWI mismatch) 21 , 24 , 25 : a macroscopic concept identifying areas with reduced blood flow but preserved tissue integrity; and (2) diffusivity‐kurtosis mismatch 10 (MD‐MK mismatch): a microscopic concept based on diffusion kurtosis imaging to assess microstructural damage. In particular, a mismatch between MD and MK has been proposed to delineate damaged core tissue from potentially salvageable penumbra. 10 However, a significant limitation of DKI is its lack of specificity (its aggregate MK measure cannot distinguish between distinct microstructural alterations). 26

To our knowledge, no in vivo study has yet applied tensor‐valued diffusion MRI in a rodent model of cerebral ischemia to explicitly quantify the mismatch between anisotropic and isotropic kurtosis. Identifying this kurtosis mismatch may enhance the delineation of injured tissue and provide a novel imaging biomarker for stroke. In this study, we addressed this gap by applying tensor‐valued diffusion MRI (dMRI) in a rat model of middle cerebral artery occlusion (MCAO), which closely resembles the pathophysiological features of heterogeneous tissue injury in humans. 3 , 27 We hypothesized that in vivo quantification of MKA and MKI would sensitively detect microstructural alterations within ischemic brain tissue and reveal a measurable mismatch between MKA and MKI within the lesion. We further evaluated the diagnostic utility of this mismatch in comparison with standard DKI metric MK and DTI metric MD, and correlated imaging findings with histopathologic markers of tissue injury. Our ultimate goal is to establish whether tensor‐valued dMRI can provide a more specific imaging biomarker, particularly the anisotropic‐isotropic kurtosis mismatch for stroke characterization, with the potential to guide treatment decisions and improve prognostication.

METHODS

The data that support the findings of this study are available from corresponding author, Yi He, on reasonable request. This study is reported in accordance with the Animal Research: Reporting of In Vivo Experiments (ARRIVE) 2.0 guidelines. 28

Animal Model

All animal procedures were conducted in accordance with a protocol approved by the Animal Ethics Committee of The Fifth Affiliated Hospital, Sun Yat‐sen University (Protocol No. 00328), and adhered to institutional and national guidelines for laboratory animal care and use. Male Sprague–Dawley rats (n=6; 8–12 weeks old) were housed under a 12‐hour light/dark cycle with ad libitum access to food and water. An MCAO rodent model was used in the study. 29 , 30 Cerebral ischemia was induced using the intraluminal filament model of MCAO. A silicone‐coated monofilament was inserted into the internal carotid artery via the external carotid artery and advanced to occlude the origin of the middle cerebral artery. After 2 hours of occlusion, the filament was withdrawn to allow reperfusion, thus establishing a reproducible model of transient ischemia–reperfusion injury (Figure 1A). Conventional dMRI revealed microstructural alterations consistent with acute ischemia, including reduced MD and elevated MK within the affected hemisphere (Figure 1B). The spatial reproducibility and anatomic consistency of infarcts across animals were validated by both T2‐weighted imaging and 2,3,5‐triphenyltetrazolium chloride staining, which confirmed lesion location and volume (Figure 1C).

Figure 1. Experimental setup and validation of infarction in the MCAO stroke model.

Figure 1

A, Focal cerebral ischemia was induced using the MCAO model. Reperfusion was initiated 2 hours after occlusion, followed by imaging on a 9.4‐T MRI scanner. B, Conventional diffusion MRI maps show changes in tissue microstructure. Left: Mean diffusivity map. Right: Mean kurtosis map. C, Infarct localization and size were confirmed by TTC staining and T2‐weighted imaging, demonstrating consistent lesion patterns across animals. MCAO indicates middle cerebral artery occlusion; MRI, magnetic resonance imaging; and TTC, 2,3,5‐triphenyltetrazolium chloride.

Real‐time monitoring of cerebral blood flow was conducted using laser speckle imaging, a wide‐field, high‐resolution technique capable of mapping relative changes in cortical surface perfusion. After cranial exposure, baseline laser speckle imaging images were acquired and continuously recorded throughout the transient middle cerebral artery occlusion procedure. Ischemia was considered successful when surface cerebral blood flow decreased to ≤20% of baseline, and reperfusion was confirmed when cerebral blood flow recovered to ≥70% of baseline following filament withdrawal. These thresholds are widely accepted in preclinical stroke models. 29 , 30 Only animals that met both occlusion and reperfusion criteria were included in subsequent MRI and histologic analyses, ensuring a reliable and quantifiable model of ischemia–reperfusion injury. With this 2‐hour MCAO protocol, delayed mortality attributable to malignant cerebral edema typically occurs at 24 to 48 hours after occlusion. Consistent with our prior reports using the same model, 29 , 30 , 31 the expected mortality rate is ≈30%. In the present study, all animals included in the imaging analysis survived through the hyperacute imaging period, and no deaths occurred before or during MRI acquisition.

MRI Acquisition and Data Processing

All MR images were acquired 2 hours after reperfusion to examine early subacute ischemic changes (Figure 1A), using a 9.4‐T small animal MRI system (BioSpec 94/30, Bruker, Germany). T2‐weighted images were acquired by using a rapid acquisition with relaxation enhancement sequence with the following parameters: field of view: 32×32 mm2; repetition time: 5000 ms; echo time: 33 ms; slices: 32 slices; image size: 256×256; resolution: 0.125×0.125; and slice thickness=800 μm. Tensor‐valued dMRI acquisitions included both linear tensor encoding (LTE) and spherical tensor encoding (STE) schemes, with the following parameters: b‐values=200, 700, 1400, and 2000 s/mm2; repetition time: 6500 ms; echo time: 45 ms; voxel size: 250×250×800 μm3; field of view: 32×32 mm2; slices: 32 slices; number of segments=8; and slice thickness=800 μm. For standard DTI/DKI acquisition, the following diffusion scheme was used: b‐values=1000 s/mm2 with 30 directions, b‐values=2000 s/mm2 with 60 directions, b=0 s/mm2 with 5 non–diffusion‐weighted images.

The tensor‐valued diffusion sequence was adapted from the MCW sequences originally developed by Dr. Matthew Budde, with modifications made to port them from Paravision 6 to Paravision 360 (https://github.com/mdbudde/mcw_Preclinical_MRIsequences). The b‐tensor waveforms, including both LTE and STE encoding shapes, were generated using custom in‐house MATLAB scripts. Additionally, these waveforms can also be reproduced using the NOW (Numerical Optimization of Waveforms) open‐source MATLAB toolbox (https://github.com/jsjol/NOW). Tensor‐valued dMRI uses gradient waveforms to probe the shape of water diffusion within tissue, extending beyond conventional single‐direction encoding. In traditional DWI/DTI, 1 gradient lobe is applied before and after the 180° pulse, sensitizing signal to diffusion along a single direction (Figure 2A and 2B). In tensor‐valued encoding, more complex waveforms are used (eg, linear versus spherical encoding), shown as differently shaped gradient blocks in the sequence diagram. By including both LTE and STE in 1 protocol (visualized as gradient‐filled waveform diagrams), tensor‐valued dMRI separates microscopic anisotropy from isotropic heterogeneity.

Figure 2. Tensor‐valued diffusion MRI gradient waveforms and powder‐averaged signal analysis.

Figure 2

A, Gradient waveform schematics for LTE and STE, illustrating diffusion sensitization along single and multiple axes, respectively. B, Conventional pulsed gradient spin‐echo waveform used in standard diffusion MRI. C, Representative b0 images showing regions of interest on the contralateral (normal) and ipsilateral (lesion) hemispheres. Corresponding powder‐averaged diffusion signal decay curves from LTE (green) and STE (gray) are plotted, with derived metrics including MD, MKA and MKI, and μFA. μFA indicates microscopic fractional anisotropy; CDE, Conventional Diffusion Encoding; L, left; LTE, linear tensor encoding; MD, mean diffusivity; MK, mean kurtosis; MKA, anisotropic MK; MKI, isotropic MK; MRI, magnetic resonance imaging; R, right; and STE, spherical tensor encoding.

MRI data were processed and analyzed using MATLAB, FSL (https://fsl.fmrib.ox.ac.uk/fsl/fslwiki), MRtrix3 (https://www.mrtrix.org/), and Advanced Normalization Tools. Preprocessing included denosing, 32 distortion correction, 33 skull stripping, and spatial smoothing. Minimal smoothing (σ=0.5 voxel) was applied solely to reduce quantization noise without altering structural detail. Multidimensional diffusion parameter estimation was conducted with the md‐dmri toolbox in MATLAB (https://github.com/markus‐nilsson/md‐dmri). Tensor‐valued dMRI metric maps were generated, including MD, MK, MKI, and MKA. Lesion areas were segmented using a K‐means clustering algorithm and overlaid on coregistered T2‐weighted images.

Lesion Segmentation and Clustering

To delineate ischemic regions at this early stage of injury (2 hours after MCAO reperfusion) and conduct subsequent analyses, we established a semiautomatic lesion segmentation workflow for diffusion parameter maps based on K‐means clustering. First, a signal‐intensity–based thresholding method was applied to each diffusion‐derived map to generate an initial lesion mask, which was then manually corrected to ensure its anatomic accuracy. These masks constrained the abnormal regions in each image and served as regions of interest (ROIs) for clustering analysis.

Subsequently, 3‐dimensional K‐means clustering was performed on each individual diffusion parameter map within the ROI using built‐in MATLAB functions. The number of clusters (K) was adaptively determined according to the lesion volume and signal complexity. Each parameter map (MD, MK, MKI, and MKA) was clustered independently, without joint processing, to preserve the unique sensitivity of each metric to microstructural alterations. Following clustering, the class best representing infarction characteristics was identified, binarized, and intersected with the anatomic lesion mask to restrict the final segmentation within a biologically reasonable region. A representative example of this segmentation workflow is provided in Figure S1. All resulting masks were saved in NIfTI format and visualized slice by slice to ensure quality control before quantitative analysis.

Histologic Preparation

Brains were collected immediately after MRI via transcardiac perfusion with saline followed by 4% paraformaldehyde, ensuring that histology reflects the same physiological state as imaging. Brains were postfixed, paraffin embedded, and sectioned at 5‐μm thickness. Sections were stained with hematoxylin and eosin (HE) for general cytoarchitecture and Luxol fast blue (LFB) for myelin integrity assessment.

HE Staining

Paraffin sections were deparaffinized, rehydrated, and stained with hematoxylin for 5 minutes, rinsed under running tap water for 1 minute, followed by eosin staining for 5 minutes. Sections were dehydrated through graded ethanol, cleared in xylene for 10 minutes, and mounted with resin.

LFB Staining

Sections were deparaffinized and rehydrated, then incubated in 0.1% LFB solution at 60 °C for 4 hours. After staining, sections were rinsed and differentiated in lithium carbonate for 5 seconds and then in 70% ethanol for 10 seconds. After thorough rinsing, slides were dehydrated in absolute ethanol (3×5 minutes), cleared in xylene for 5 minutes, and sealed with mounting resin.

Histology‐MRI Coregistration and ROI Definition

Histologic and MRI images were spatially coregistered using MRtrix3 and FSL software. Histology images were identified corresponding to MRI slices based on slicing depth markers and anatomic landmarks. High‐resolution histologic images (0.1214 μm/pixel) were aligned to the MRI data using custom‐written MATLAB scripts. To achieve precise alignment between histologic sections and MRI slices, we implemented a robust 30‐point landmark registration workflow based on in‐house code. Thirty anatomically homologous landmarks were manually annotated on both MRI slices and their corresponding histologic images. We used MATLAB's fitgeotform2d function with a third‐degree polynomial model, performing both global alignment and smooth local deformations in a single transformation. This approach, combining carefully selected anatomic landmarks with a flexible third‐degree polynomial model, ensures robust alignment between histology and MRI despite nonlinear tissue distortions during histologic processing. Image resolutions were normalized to ensure consistent scale across modalities. Identical ROIs were manually delineated on both histologic and MRI data sets to enable quantitative comparison of microstructural features across imaging modalities. This precise coregistration represents a key methodological innovation, supporting mechanistic interpretation of tensor‐valued diffusion MRI contrasts in hyperacute ischemia.

Histologic Quantification of HI and HA

To validate microstructural parameters, histologic isotropy (HI) and histologic anisotropy (HA) were quantitatively assessed from HE and LFB‐stained tissue sections using in‐house MATLAB algorithms. Each HE‐ and LFB‐stained image was divided into 125×125‐μm2 regions, further subdivided into twenty‐five 25×25‐μm2 subregions. Cell nuclei were segmented through grayscale thresholding and morphologic operations, and the number of labeled nuclei was used to determine local cell density (cells/μm2). HI was calculated as the normalized variance of cell densities derived from microscopy. 14 LFB‐stained images were converted to grayscale and analyzed using structure tensor analysis to quantify directional microstructural organization. 15 , 34 , 35 Local image gradients were computed using Gaussian derivative filters (σ 1=1 μm) in x and y directions. The resulting tensor elements (Sxx, Sxy, and Syy) were smoothed with a Gaussian filter (σ 2=20 μm), and the dominant eigenvalues (λ₁ and λ2) were extracted within each region. HA was defined as the normalized variance of these eigenvalues (3×variance/mean2), reflecting directional coherence. MRI‐defined ROIs were transposed onto the corresponding registered histology images to ensure voxelwise correspondence after exclusion of artifacts. Finally, HA was computed across regions, and defined as the normalized variance of these eigenvalues, reflecting the degree of directional coherence 14 (Figure S2). This ensures a true voxelwise correspondence between imaging and ground‐truth histologic data, thus validating MRI‐based lesion identification.

Statistical Analysis

All statistical analyses and data visualizations were performed using GraphPad Prism 10.0 (https://www.graphpad‐prism.cn/) and MATLAB. Linear regression analyses were used to quantify the relationship between dMRI‐derived parameters and histologically defined microstructural metrics, such as MKA versus HA, and MKI versus HI. To evaluate group‐level differences, 2‐sample t‐tests were used to compare dMRI parameters between infarcted and contralateral brain regions. Pearson correlation coefficient (r) was used to assess the strength of associations, with statistical significance set at α=0.05. P<0.05 was considered statistically significant. Statistically significant correlations are marked with asterisks (*), whereas nonsignificant correlations are denoted as “NS.”

RESULTS

In Vivo Diffusion Encoding and Acquisition Using Tensor‐Valued MRI in Stroke

The tensor‐valued protocol included LTE and STE acquisitions (Figure 2A), whereas conventional DWI used standard single‐direction diffusion encoding (Figure 2B). We analyzed powder‐averaged signal decays from LTE and STE acquisitions in both the ipsilateral (lesion) and contralateral (normal) hemispheres. Orientation‐invariant diffusion signals were obtained by averaging the diffusion‐weighted volumes for each b‐value shell, yielding metrics free from fiber orientation dispersion.

Representative signal decays and derived microstructural metrics from 6 brain ROIs in 1 MCAO rat are shown in Figure 2C. In the contralateral (nonischemic) hemisphere, diffusion profiles showed higher MD values (0.81–0.82 μm2/ms) and relatively lower MKI values, with MKA ranging from 0.5 to 1.0, and microscopic FA (μFA) ranging from 0.66 to 0.82. In contrast, ischemic ROIs in the ipsilateral hemisphere demonstrated reduced MD values between 0.62 and 0.66 μm2/ms, elevated MKI, and decreased MKA and μFA values.

Regional Kurtosis Alterations Revealed by Tensor‐Valued dMRI

In the MCAO rat model, comparison of the ipsilateral (ischemic regions) and contralateral (normal) hemispheres revealed distinct alterations across tensor‐valued dMRI metrics. As shown in Figure 3A, ischemic regions exhibited markedly reduced MD and MKA, alongside elevated MK and MKI. These patterns reflect restricted diffusion and increased microstructural complexity in the infarct core. Importantly, these diffusion‐derived abnormalities represent evolving ischemic regions at 2 hours after MCAO rather than a fully developed infarct. The spatial mismatch between MKA and MKI further underscores the utility of tensor‐valued decomposition in identifying microstructural heterogeneity within early ischemic tissue.

Figure 3. Mismatch between MKI and MKA in ischemic brain tissue 2 hours after reperfusion.

Figure 3

A, Representative coronal maps of MD, MK, MKI, and MKA from P to A brain slices. The lesion region demonstrates decreased MD (green arrow) and MKA (red arrow), whereas MK (orange arrow) and MKI (purple arrow) are notably elevated, indicating a mismatch between the kurtosis sources. B, Overlay of lesion regions defined by each diffusion metric (color coded) on b0 images from a representative animal. ROIs were segmented based on metric‐specific abnormal voxels: MD (green), MK (yellow), MKI (red), and MKA (blue). C–F, Voxel‐wise comparisons between lesion and contralateral hemispheres for MD, MK, MKA, and MKI. Lesion ROIs showed significantly lower MD (Δ=−0.00018 [95% CI, −0.000181 to −0.000179], d=−1.54, FDR‐adjusted P<0.0001) and MKA (Δ=−0.4863 [95% CI, −0.4967 to −0.4758], d=−1.17, FDR‐adjusted P<0.0001), and significantly higher MK (Δ=0.4966 [95% CI, 0.4917−0.5015], d=1.40, FDR‐adjusted P<0.0001) and MKI (Δ=0.6206 [95% CI, 0.6086−0.6326], d=0.89, FDR‐adjusted P<0.001), supporting differential sensitivity of these metrics to ischemic tissue microstructural changes. A indicates anterior; FDR, false discovery rate; L, left; MD, mean diffusivity; MK, mean kurtosis; MKA, anisotropic MK; MKI, isotropic MK; P, posterior; R, right; and ROI, region of interest.

To accurately delineate these variations, K‐means clustering was applied to the multimetric data (Figure S3), enabling data‐driven segmentation of infarcted regions based on voxelwise diffusion properties. 3 , 36 We then overlaid lesion maps derived from each metric (Figure 3B) with the specific K‐means clustering workflow illustrated in Figure S1, revealing that MKI‐defined lesions were spatially more confined than MK‐defined regions, whereas MKA reductions extended variably across perilesional tissue. This provides further support for a kurtosis‐based microstructural mismatch framework, where elevated MK may result from either increased isotropic restriction or retained directional coherence. Quantitative analysis of mean values across segmented lesion ROIs demonstrated statistically significant differences between hemispheres (Figure 3C–3F). Compared with the contralateral side, lesion ROIs showed significantly lower MD (Δ=−0.00018 [95% CI, −0.000181 to −0.000179], d=−1.54, false discovery rate–adjusted P<0.0001) and MKA (Δ=−0.4863 [95% CI, −0.4967 to −0.4758], d=−1.17, false discovery rate–adjusted P<0.0001), and significantly higher MK (Δ=0.4966 [95% CI, 0.4917–0.5015], d=1.40, false discovery rate–adjusted P<0.0001) and MKI (Δ=0.6206 [95% CI, 0.6086–0.6326], d=0.89, false discovery rate–adjusted P<0.0001), confirming both diffusion restriction and kurtosis heterogeneity within ischemic tissue.

Influence of White and Gray Matter Composition on MKI /MKA Changes

Given the inherent anatomic contrast between white matter (WM) and gray matter (GM) in the MKI and MKA maps, we evaluated whether tissue type influenced the observed kurtosis alterations in ischemic regions. Diffusion data sets were spatially normalized to a standard rat brain atlas with predefined WM and GM masks. Using these masks, we assessed the composition of MKI‐ and MKA‐defined clusters with respect to tissue type.

The MKI‐related cluster consisted predominantly of GM voxels (mean±SD: 85.7%±6.5%) with a smaller proportion of WM voxels (14.3%±6.5%). Similarly, the MKA‐related cluster contained 88.2%±9.1% GM and 11.8%±9.1% WM voxels. To determine whether the degree of kurtosis change differed by tissue type, we compared MKI and MKA values in WM‐ versus GM‐dominated clusters. No statistically significant differences were found (MKI: P=0.489; MKA: P=0.476; paired t‐tests), suggesting that the observed alterations were not primarily driven by WM/GM composition (Figure S4). These findings support that although MKI and MKA maps exhibit intrinsic WM/GM contrast, the ischemia‐related alterations captured by tensor‐valued dMRI are largely independent of tissue type.

Histologic Correlation of Tensor‐Valued dMRI Metrics in Stroke

To interpret dMRI findings at the microstructural level, we performed ROI analysis by precisely coregistering in vivo dMRI maps with corresponding histologic sections. Histologic samples were collected immediately after MRI (within 20 minutes), ensuring direct temporal correspondence. dMRI lacks chemical specificity to myelin; observed diffusion contrasts reflect microstructural modulation of water displacement by barriers such as cell membranes, axons, and myelin sheaths, rather than direct sensitivity to myelin as a molecular entity. HE staining and LFB staining were used to identify cytological and myelin‐specific changes, respectively, in infarcted versus contralateral brain tissue. In GM, the contributions of myelin to diffusion metrics are minimal; thus, observed changes in tensor‐encoded metrics in our rat MCAO model are most likely driven by cytotoxic processes (such as neurite integrity) rather than myelination. For each dMRI‐derived metric, symmetric ROIs were selected in both hemispheres to quantify MKA and MKI kurtosis, alongside corresponding histologic indexes of structural integrity (HA and HI). As shown in Figure 4A and 4C, the infarcted regions demonstrated pronounced changes in both kurtosis metrics and their histologic correlates.

Figure 4. Histologic validation of tensor‐valued diffusion MRI metrics in ischemic brain tissue.

Figure 4

A, Coregistration of MRI‐derived MKI maps with hematoxylin and eosin staining. HI was assessed in the lesion and contralateral regions and compared with corresponding MKI values. B, Violin plots show significantly elevated MKI and histology‐derived cell density index (HI) in the lesion side compared with the contralateral side (P<0.0001). Linear regression reveals a strong positive correlation between MKI and HI (r=0.776 [95% CI, 0.672−0.850], P<10−5). C, Coregistration of MKA maps with Luxol fast blue staining, used to quantify myelin density. D, MKA and histology‐derived anisotropy index (HA) were both significantly lower in the lesion side (P<0.0001), and showed a strong positive correlation (r=0.724 [95% CI, 0.667−0.773], P<10−5), supporting MKA as a marker of microstructural anisotropy loss. HA indicates histologic anisotropy; HI, histologic isotropy; L, left; MKA, anisotropic mean kurtosis; MKI, isotropic mean kurtosis; MRI, magnetic resonance imaging; and R, right.

Quantitative comparisons revealed significantly elevated MKI and HI values on the lesion side compared with the contralateral side (P<0.0001, n=6 rats; Figure 4B). Moreover, MKI was strongly correlated with HI (r=0.776 [95% CI, 0.672–0.850], P<10−5, n=6 rats), suggesting that isotropic kurtosis reflects variable cell density induced by cytotoxic edema. Similarly, MKA and HA were significantly reduced in the infarcted hemisphere (P<0.0001, n=6 rats; Figure 4D), and a robust positive correlation was observed between MKA and histologic anisotropy HA (r=0.724 [95% CI, 0.667–0.773], P<0.0001, n=6 rats), In contrast, cross‐comparisons between mismatched parameters (MKI versus HA, and MKA versus HI) showed no significant correlation (Figure S5), further confirming that MKI and MKA selectively reflect distinct microstructural domains rather than overlapping histologic features. Together, these data demonstrate that tensor‐valued dMRI metrics, particularly MKI and MKA, reliably correspond to histologic alterations in ischemic brain tissue.

Voxel‐Wise Quantification of Isotropic‐Anisotropic Mismatch

To explore these microstructural differences, lesion masks were analyzed for MKI and MKA components, highlighting distinct diffusion patterns within ischemic tissue. In Figure 5A, we reconstructed and visualized 3‐dimensional lesion masks based on each diffusion metric: MD, MK, MKI, and MKA. These reconstructions revealed distinct spatial distributions within the infarcted hemisphere. MKI‐based lesions appeared more confined and medially located compared with the broader areas delineated by MD or MK. In contrast, MKA highlighted regions with disrupted directional coherence. Moreover, we illustrated the conceptual basis for these tensor‐valued metrics (Figure 5B).

Figure 5. Tensor‐valued dMRI reveals mismatches in isotropic and anisotropic kurtosis components.

Figure 5

A, Three‐dimensional rendering of lesion volumes from a representative animal 2 hours after MCAO, segmented by diffusion and kurtosis metrics: MD (green), MK (yellow), MKI (red), and MKA (blue). B, Schematic representations of isotropic (spherical) and anisotropic (columnar) microstructural environments, corresponding to MKI and MKA components. C, Venn‐type overlap diagram summarize lesion fractions relative to MD across animals. D, Proportions of MKI (mean±SD: 33.6%±19.4%), MKA (17.3%±6.3%), MKI∩MKA (7.6%±4.0%), and MK (54.4%±15.1%) voxels within the MD lesion, showing MK covered the largest fraction. A indicates anterior; dMRI, diffusion magnetic resonance imaging; L, left; MCAO, middle cerebral artery occlusion; MD, mean diffusivity; MK, mean kurtosis; MKA, anisotropic MK; MKI, isotropic MK; P, posterior; and R, right.

To quantify the mismatch between MKI and MKA within ischemic lesions, we performed a voxel‐wise overlap analysis relative to the MD‐defined lesion mask. As shown in Figure 5C, MKI and MKA exhibited partially distinct distributions with limited overlap, indicating that they capture complementary rather than redundant aspects of ischemic pathology. Group‐level analysis across animals (Figure 5D) revealed that MK encompassed the largest fraction of the MD‐defined lesion, whereas MKI consistently occupied a significantly greater portion than MKA (P<0.05). The overlap between MKI and MKA remained small across subjects, further reinforcing their distinct and nonredundant contributions.

Collectively, these results establish a voxel‐wise quantitative framework for characterizing anisotropic–isotropic mismatches in ischemic lesions. MK alterations were predominantly driven by MKI, whereas MKA highlighted a smaller but spatially distinct subset of abnormal voxels. This mismatch quantification advances lesion characterization beyond conventional MD or MK metrics, providing novel insights into the heterogeneous microstructural pathology of ischemic stroke.

DISCUSSION

In this study, we applied tensor‐valued dMRI to a rat model of acute ischemic stroke, enabling in vivo decomposition of MK into MKI and MKA components. This approach enhances sensitivity to distinct microstructural features underlying diffusion kurtosis and allows a more nuanced characterization of tissue heterogeneity associated with acute cerebral ischemia in a controlled experimental setting. The present findings indicate that tensor‐valued dMRI provides complementary mechanistic insight beyond conventional DTI and standard DKI, rather than replacing established diffusion metrics.

All measurements were performed at a single hyperacute time point. Therefore, the present study does not address the temporal evolution of tensor‐valued diffusion metrics or their relationship to long‐term tissue outcome. Consistent with prior stroke imaging literature, MRI measurements acquired during the earliest phases of ischemia, beyond established metrics, such as ADC or T2 relaxation, generally provide limited prognostic information about tissue fate. Moreover, advanced diffusion encoding schemes, such as tensor‐valued dMRI, are currently impractical for routine hyperacute clinical stroke workflows. Accordingly, the present results should be interpreted strictly as mechanistic observations under controlled preclinical conditions, rather than as evidence supporting immediate clinical translation or prognostic utility.

Although MD decline is a well‐established hallmark of early ischemic injury, it does not distinguish cytotoxic edema from irreversible structural damage. This limitation motivates the exploration of diffusion‐based metrics that are more sensitive to microstructural heterogeneity during acute ischemia. In the present study, histologic correlation suggests that tensor‐valued diffusion metrics capture distinct aspects of microstructural heterogeneity during acute ischemia. Importantly, interpretations of MKI and MKA alterations should be regarded as hypotheses supported by correlative histologic evidence, rather than definitive mechanistic conclusions. By separating MK into MKI and MKA, the technique enables differential assessment of isotropic versus anisotropic microstructural alterations, a distinction that aligns with known cytological responses to acute ischemia.

Elevated MKI was primarily localized to the infarct core, correlating with histologic evidence of cytotoxic edema and cell density variations. 37 However, these associations should be interpreted as suggestive, reflecting possible cytotoxic changes rather than proven microstructural mechanisms. In contrast, MKA was significantly reduced in regions exhibiting disrupted tissue organization, as supported by LFB staining. This reduction likely reflects altered water displacement arising from microstructural disorganization, rather than direct sensitivity to myelin as a molecular constituent, consistent with the known lack of chemical specificity in dMRI. In GM, including both cortical regions and deep GM, anisotropic kurtosis reflects the orientation and coherence of local neuritic processes, such as dendrites and thin intracortical or subcortical axons, rather than the large myelinated fiber bundles characteristic of WM. Thus, reductions in MKA in gray matter are interpreted as disruptions to the organization of these oriented neuritic processes and their spatial coherence. Consistent with histologic structure tensor theory, anisotropic kurtosis parallels orientation‐dependent variance, and isotropic kurtosis parallels local cellular heterogeneity. 38 The ability to decouple these diffusion components allowed us to map both spherically restricted and directionally impaired compartments, enhancing specificity beyond that of DKI. Importantly, although MKI‐MKA mismatch provides insights into microstructural heterogeneity, the present study is preclinical, and any implications for clinical stroke imaging should be considered speculative rather than validated biomarkers.

Our findings are broadly consistent with a recent human study 39 that applied b‐tensor diffusion encoding in patients with acute stroke. Their work demonstrated that diffusional variance metrics, including MKA and MKI, were sensitive to ischemic changes and modulated by diffusion time. Specifically, longer diffusion times revealed elevated microscopic μFA and MKA in lesions compared with contralateral regions, whereas shorter diffusion times yielded reduced values. The detectability of these structural changes as either anisotropic or isotropic signals depends critically on the diffusion time window. In our animal study, diffusion data were acquired at a single diffusion time during a stage likely characterized by advanced cytotoxic edema and axonal damage, which may explain the predominance of isotropic kurtosis elevation and anisotropic kurtosis reduction.

Prior experimental studies in both animal and human stroke models have linked such kurtosis patterns, specifically increased MKI and decreased MKA, to neurite beading, a hallmark of early microstructural disruption. Using Monte Carlo simulations, these studies demonstrated that MKI elevation reflects increased isotropic diffusion heterogeneity caused by swollen or beaded neurites, whereas MKA reduction indicates diminished anisotropic directional coherence attributable to axonal injury. 7 , 39 This supports the biophysical interpretation that MKI and MKA provide distinct, yet complementary, insights into ischemic microstructural remodeling. Moreover, given the low baseline anisotropy of cortical GM in rodent stroke models, MKA and μFA changes are expected to be small; MKI increases dominate. In human studies where WM is prominently affected, the magnitude and interpretability of MKA/μFA changes are substantially higher. Thus, translating findings requires careful consideration of tissue composition differences between species. Despite differences in timing and protocol, both studies suggest that tensor‐valued dMRI can disentangle isotropic and anisotropic tissue features. Together, these findings support the interpretation that MKI and MKA provide distinct, yet complementary, information about ischemic microstructural remodeling.

Importantly, although the observed MKI‐MKA mismatch reflects underlying microstructural heterogeneity within ischemic lesions, it should not be interpreted as a validated clinical biomarker. Instead, this mismatch provides a conceptual and descriptive framework for interpreting the coexistence of distinct diffusion compartments, including spherically restricted and directionally disorganized tissue components, in experimental ischemia. Any extrapolation to infarct core delineation, tissue viability, or clinical decision‐making remains speculative and requires rigorous longitudinal, outcome‐based, and multimodal validation.

When considered alongside established relaxometric mismatch paradigms, such as DWI/fluid‐attenuated inversion recovery and ADC/T2, tensor‐valued dMRI does not seek to replace these clinically validated approaches. Rather, it offers a complementary preclinical framework for probing microstructural heterogeneity that remains indistinct on conventional diffusion imaging. Any relationship to clinical paradigms requires dedicated longitudinal investigation and is beyond the scope of the present study. Elevated MKI coupled with reduced MKA may reflect distinct patterns of diffusion heterogeneity within ischemic tissue, whereas areas with preserved MKA but mild MKI elevation may represent functionally impaired yet structurally viable tissue. Importantly, these patterns should not be interpreted as direct indicators of tissue fate, infarct core, or penumbral status. Rather, the observed heterogeneity highlights the coexistence of multiple diffusion compartments within ischemic lesions that remain indistinct when characterized using conventional diffusion metrics alone.

Despite these findings, several important limitations should be acknowledged. First, the relatively small sample size (n=6) may limit generalizability, although interanimal consistency was high. Second, the study was performed at ultrahigh field (9.4 T) with small voxel size, which may not directly translate to clinical settings; however, tensor‐valued diffusion encoding is feasible at clinical field strengths and has been demonstrated in human studies. 39 Third, although histology provides ground truth, coregistration between MRI and histologic sections remains technically challenging and may introduce minor spatial misalignments. 36 , 40

A potential concern is that higher‐order diffusion kurtosis metrics (MKI and MKA) are inherently more noise sensitive than lower‐order metrics, such as MD or MK. Increased fitting error and reduced contrast/noise ratio could therefore contribute to apparent mismatches. We acknowledge this limitation explicitly. Nevertheless, the consistent detection of MKI‐only and MKA‐only voxels across all animals suggests that the mismatch cannot be fully explained by noise or fitting sensitivity. Instead, these findings likely reflect genuine biological variability, with isotropic and anisotropic microstructural alterations contributing differently to kurtosis within the ischemic lesion. Future work incorporating test–retest reliability, explicit noise modeling, and advanced acquisition strategies will be critical to further disentangle technical versus biological sources of the observed mismatch.

Future research should focus on longitudinal and multimodal investigations to characterize the temporal dynamics of isotropic‐anisotropic kurtosis patterns and to further clarify their biological underpinnings across different stages and models of cerebrovascular injury. Such longitudinal analysis could offer critical insights into the progression, severity, and potential reversibility of tissue injury. 41 In addition, extending this approach to other stroke models, including hemorrhagic stroke and various ischemic subtypes, would help elucidate both shared and distinct patterns of microstructural disruption. These efforts would improve the robustness and interpretability of tensor‐valued dMRI for mechanistic studies across diverse cerebrovascular injury models.

CONCLUSIONS

We have demonstrated that in vivo tensor‐valued dMRI can separately quantify isotropic and anisotropic diffusion kurtosis in acute stroke, revealing a distinct kurtosis mismatch in ischemic tissue that is not discernible with conventional techniques. Although tensor‐valued dMRI has previously been applied in other contexts, to our knowledge, it has not yet been systematically investigated in experimental stroke models. Building on these prior advances, the present study focuses on quantitative in vivo assessment of the mismatch between isotropic and anisotropic kurtosis as a framework for characterizing microstructural heterogeneity in a rodent MCAO model. This approach provides a more specific imaging readout of diffusion heterogeneity associated with ischemic pathology in a controlled preclinical setting. Compared with standard DTI and DKI, tensor‐valued dMRI offers enhanced specificity for distinguishing isotropic and anisotropic contributions to diffusion kurtosis, without implying direct identification of tissue fate. These findings contribute to growing evidence that advanced dMRI can capture the complexity of diffusion heterogeneity in ischemic tissue.

In summary, tensor‐valued dMRI enables decomposition of diffusion kurtosis into isotropic and anisotropic components, providing enhanced sensitivity to microstructural heterogeneity in experimental stroke. The ability to map anisotropic and isotropic kurtosis in vivo opens new possibilities for mechanistic investigation of ischemic microstructural alterations, including the study of edema‐related and cytoskeletal changes in experimental models. Ultimately, this work underscores the potential of advanced dMRI for decoding microstructural signatures underlying ischemic injury, providing a foundation for future longitudinal and translational investigations.

Sources of Funding

This work was supported by grants from the National Natural Science Foundation of China (No. 82201447, No. RLZY20231001‐01, and No. 82371307, Y.T.), Shenzhen Science and Technology Program (JSGGKQTD20210831174329010), Shenzhen Outstanding Talents Training Fund, the Strategic Priority Research Program of the Chinese Academy of Sciences (XDB0930000), National Key Research and Development Program of China (2023YFF0714200 and 2024YFC3406602), the Chinese Academy of Sciences Project (No. Y2021098), the CAS Project for Young Scientists in Basic Research (YSBR‐114), the Fundamental Research Funds for the Central Universities (YG2025ZD18, Y.T.), Young Leading Scientists Cultivation Plan supported by Shanghai Municipal Education Commission (ZXWH1082101, Y.T.), Key Laboratory for Magnetic Resonance and Multimodality Imaging of Guangdong Province (2023B1212060052), the Medical Scientific Research Foundation of Guangdong Province of China (A2023080), the Young Scientists Innovation Research Project on Cerebral Small Vessel Disease (Y.T.) and Shenzhen Key Laboratory for Magnetic Resonance Imaging (SYSPG20241211173949066).

Disclosures

None.

Supporting information

Figures S1–S5

ARRIVE checklist

JAH3-15-e046071-s001.pdf (153.3KB, pdf)

Preprint posted on bioRxiv, June 27, 2025. doi:https://doi.org/10.1101/2025.06.23.661205.

This manuscript was sent to Neel Singhal, MD, PhD, Associate Editor, for review by expert referees, editorial decision, and final disposition.

For Sources of Funding and Disclosures, see page 13.

Contributor Information

Yaohui Tang, Email: yaohuitang@sjtu.edu.cn.

Ye Li, Email: liye1@siat.ac.cn.

Yi He, Email: y.he3@siat.ac.cn.

References

  • 1. Campbell BCV , De Silva DA, Macleod MR, Coutts SB, Schwamm LH, Davis SM, Donnan GA. Ischaemic stroke. Nat Rev Dis Primers. 2019;5:70. doi: 10.1038/s41572-019-0118-8 [DOI] [PubMed] [Google Scholar]
  • 2. Global, regional, and national burden of stroke and its risk factors, 1990‐2019: a systematic analysis for the global burden of disease study 2019. Lancet Neurol. 2021;20:795–820. doi: 10.1016/s1474-4422(21)00252-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3. Lu D, Jiang Y, Ji Y, Zhou IY, Mandeville E, Lo EH, Wang X, Sun PZ. Evaluation of diffusion kurtosis imaging of stroke lesion with hemodynamic and metabolic MRI in a rodent model of acute stroke. AJR Am J Roentgenol. 2018;210:720–727. doi: 10.2214/ajr.17.19134 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4. Cheung J, Doerr M, Hu R, Sun PZ. Refined ischemic penumbra imaging with tissue pH and diffusion kurtosis magnetic resonance imaging. Transl Stroke Res. 2021;12:742–753. doi: 10.1007/s12975-020-00868-z [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5. DiBella EVR, Sharma A, Richards L, Prabhakaran V, Majersik JJ, HashemizadehKolowri SK. Beyond diffusion tensor MRI methods for improved characterization of the brain after ischemic stroke: a review. AJNR Am J Neuroradiol. 2022;43:661–669. doi: 10.3174/ajnr.A7414 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6. Yin J, Sun H, Wang Z, Ni H, Shen W, Sun PZ. Diffusion kurtosis imaging of acute infarction: comparison with routine diffusion and follow‐up MR imaging. Radiology. 2018;287:651–657. doi: 10.1148/radiol.2017170553 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7. Alves R, Henriques RN, Kerkelä L, Chavarrías C, Jespersen SN, Shemesh N. Correlation tensor MRI deciphers underlying kurtosis sources in stroke. NeuroImage. 2022;247:118833. doi: 10.1016/j.neuroimage.2021.118833 [DOI] [PubMed] [Google Scholar]
  • 8. Bay V, Kjølby BF, Iversen NK, Mikkelsen IK, Ardalan M, Nyengaard JR, Jespersen SN, Drasbek KR, Østergaard L, Hansen B. Stroke infarct volume estimation in fixed tissue: comparison of diffusion kurtosis imaging to diffusion weighted imaging and histology in a rodent MCAO model. PLoS One. 2018;13:e0196161. doi: 10.1371/journal.pone.0196161 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9. Hu R, Kim H, Kim J, Allen JW, Sun PZ. Fast diffusion kurtosis imaging in acute ischemic stroke shows mean kurtosis‐diffusivity mismatch. J Neuroimaging. 2022;32:941–946. doi: 10.1111/jon.13000 [DOI] [PubMed] [Google Scholar]
  • 10. Cheung JS, Wang E, Lo EH, Sun PZ. Stratification of heterogeneous diffusion MRI ischemic lesion with kurtosis imaging: evaluation of mean diffusion and kurtosis MRI mismatch in an animal model of transient focal ischemia. Stroke. 2012;43:2252–2254. doi: 10.1161/strokeaha.112.661926 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11. He Y, Aznar S, Siebner HR, Dyrby TB. In vivo tensor‐valued diffusion MRI of focal demyelination in white and deep grey matter of rodents. Neuroimage Clin. 2021;30:102675. doi: 10.1016/j.nicl.2021.102675 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12. Nilsson M, Szczepankiewicz F, Brabec J, Taylor M, Westin CF, Golby A, van Westen D, Sundgren PC. Tensor‐valued diffusion MRI in under 3 minutes: an initial survey of microscopic anisotropy and tissue heterogeneity in intracranial tumors. Magn Reson Med. 2020;83:608–620. doi: 10.1002/mrm.27959 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13. Uddin MN, Singh MV, Faiyaz A, Szczepankiewicz F, Nilsson M, Boodoo ZD, Sutton KR, Tivarus ME, Zhong J, Wang L, et al. Tensor‐valued diffusion MRI detects brain microstructural abnormalities in HIV infected individuals with cognitive impairment. Sci Rep. 2024;14:28839. doi: 10.1038/s41598-024-80372-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14. Szczepankiewicz F, van Westen D, Englund E, Westin CF, Ståhlberg F, Lätt J, Sundgren PC, Nilsson M. The link between diffusion MRI and tumor heterogeneity: mapping cell eccentricity and density by diffusional variance decomposition (DIVIDE). NeuroImage. 2016;142:522–532. doi: 10.1016/j.neuroimage.2016.07.038 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15. Szczepankiewicz F, Sjölund J, Ståhlberg F, Lätt J, Nilsson M. Tensor‐valued diffusion encoding for diffusional variance decomposition (DIVIDE): technical feasibility in clinical MRI systems. PLoS One. 2019;14:e0214238. doi: 10.1371/journal.pone.0214238 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16. Chen X, Lin L, Wu J, Yang G, Zhong T, Du X, Chen Z, Xu G, Song Y, Xue Y, et al. Histogram analysis in predicting the grade and histological subtype of meningiomas based on diffusion kurtosis imaging. Acta Radiol. 2020;61:1228–1239. doi: 10.1177/0284185119898656 [DOI] [PubMed] [Google Scholar]
  • 17. Li S, Zheng Y, Sun W, Lasič S, Szczepankiewicz F, Wei Q, Han S, Zhang S, Zhong X, Wang L, et al. Glioma grading, molecular feature classification, and microstructural characterization using MR diffusional variance decomposition (DIVIDE) imaging. Eur Radiol. 2021;31:8197–8207. doi: 10.1007/s00330-021-07959-x [DOI] [PubMed] [Google Scholar]
  • 18. Astrup J, Siesjö BK, Symon L. Thresholds in cerebral ischemia ‐ the ischemic penumbra. Stroke. 1981;12:723–725. doi: 10.1161/01.str.12.6.723 [DOI] [PubMed] [Google Scholar]
  • 19. Paciaroni M, Caso V, Agnelli G. The concept of ischemic penumbra in acute stroke and therapeutic opportunities. Eur Neurol. 2009;61:321–330. doi: 10.1159/000210544 [DOI] [PubMed] [Google Scholar]
  • 20. Leigh R, Knutsson L, Zhou J, van Zijl PC. Imaging the physiological evolution of the ischemic penumbra in acute ischemic stroke. J Cereb Blood Flow Metab. 2018;38:1500–1516. doi: 10.1177/0271678x17700913 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21. Chen F, Ni YC. Magnetic resonance diffusion‐perfusion mismatch in acute ischemic stroke: an update. World J Radiol. 2012;4:63–74. doi: 10.4329/wjr.v4.i3.63 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22. Thomalla G, Cheng B, Ebinger M, Hao Q, Tourdias T, Wu O, Kim JS, Breuer L, Singer OC, Warach S, et al. DWI‐FLAIR mismatch for the identification of patients with acute ischaemic stroke within 4·5 h of symptom onset (PRE‐FLAIR): a multicentre observational study. Lancet Neurol. 2011;10:978–986. doi: 10.1016/S1474-4422(11)70192-2 [DOI] [PubMed] [Google Scholar]
  • 23. Lansberg MG, Straka M, Kemp S, Mlynash M, Wechsler LR, Jovin TG, Wilder MJ, Lutsep HL, Czartoski TJ, Bernstein RA, et al. MRI profile and response to endovascular reperfusion after stroke (DEFUSE 2): a prospective cohort study. Lancet Neurol. 2012;11:860–867. doi: 10.1016/S1474-4422(12)70203-X [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24. Neumann‐Haefelin T, Wittsack H‐J, Wenserski F, Siebler M, Seitz RJ, Mödder U, Freund H‐J. Diffusion‐ and perfusion‐weighted MRI. Stroke. 1999;30:1591–1597. doi: 10.1161/01.STR.30.8.1591 [DOI] [PubMed] [Google Scholar]
  • 25. Butcher K, Parsons M, Allport L, Lee SB, Barber PA, Tress B, Donnan GA, Davis SM. Rapid assessment of perfusion–diffusion mismatch. Stroke. 2008;39:75–81. doi: 10.1161/STROKEAHA.107.490524 [DOI] [PubMed] [Google Scholar]
  • 26. Henriques RN, Jespersen SN, Shemesh N. Correlation tensor magnetic resonance imaging. NeuroImage. 2020;211:116605. doi: 10.1016/j.neuroimage.2020.116605 [DOI] [PubMed] [Google Scholar]
  • 27. Liu F, McCullough LD. Middle cerebral artery occlusion model in rodents: methods and potential pitfalls. Biomed Res Int. 2011;2011:464701. doi: 10.1155/2011/464701 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28. Percie du Sert N, Hurst V, Ahluwalia A, Alam S, Avey MT, Baker M, Browne WJ, Clark A, Cuthill IC, Dirnagl U. The ARRIVE guidelines 2.0: updated guidelines for reporting animal research. J Cereb Blood Flow Metab. 2020;40:1769–1777. doi: 10.1177/0271678X20943823 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29. Tang Y, Wang J, Lin X, Wang L, Shao B, Jin K, Wang Y, Yang GY. Neural stem cell protects aged rat brain from ischemia‐reperfusion injury through neurogenesis and angiogenesis. J Cereb Blood Flow Metab. 2014;34:1138–1147. doi: 10.1038/jcbfm.2014.61 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30. Li J, Tang Y, Wang Y, Tang R, Jiang W, Yang GY, Gao WQ. Neurovascular recovery via co‐transplanted neural and vascular progenitors leads to improved functional restoration after ischemic stroke in rats. Stem Cell Reports. 2014;3:101–114. doi: 10.1016/j.stemcr.2014.05.012 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31. Tang Y, Cai B, Yuan F, He X, Lin X, Wang J, Wang Y, Yang G‐Y. Melatonin pretreatment improves the survival and function of transplanted mesenchymal stem cells after focal cerebral ischemia. Cell Transplant. 2014;23:1279–1291. doi: 10.3727/096368913X667510 [DOI] [PubMed] [Google Scholar]
  • 32. Veraart J, Novikov DS, Christiaens D, Ades‐aron B, Sijbers J, Fieremans E. Denoising of diffusion MRI using random matrix theory. NeuroImage. 2016;142:394–406. doi: 10.1016/j.neuroimage.2016.08.016 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33. Andersson JLR, Sotiropoulos SN. An integrated approach to correction for off‐resonance effects and subject movement in diffusion MR imaging. NeuroImage. 2016;125:1063–1078. doi: 10.1016/j.neuroimage.2015.10.019 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34. de Almeida Martins JP, Nilsson M, Lampinen B, Palombo M, While PT, Westin C‐F, Szczepankiewicz F. Neural networks for parameter estimation in microstructural MRI: application to a diffusion‐relaxation model of white matter. NeuroImage. 2021;244:118601. doi: 10.1016/j.neuroimage.2021.118601 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35. Budde MD, Annese J. Quantification of anisotropy and fiber orientation in human brain histological sections. Front Integr Neurosci. 2013;7:3. doi: 10.3389/fnint.2013.00003 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36. Zhou IY, Guo Y, Igarashi T, Wang Y, Mandeville E, Chan ST, Wen L, Vangel M, Lo EH, Ji X, et al. Fast diffusion kurtosis imaging (DKI) with inherent COrrelation‐based normalization (ICON) enhances automatic segmentation of heterogeneous diffusion MRI lesion in acute stroke. NMR Biomed. 2016;29:1670–1677. doi: 10.1002/nbm.3617 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37. Nilsson M, Eklund G, Szczepankiewicz F, Skorpil M, Bryskhe K, Westin CF, Lindh C, Blomqvist L, Jäderling F. Mapping prostatic microscopic anisotropy using linear and spherical b‐tensor encoding: a preliminary study. Magn Reson Med. 2021;86:2025–2033. doi: 10.1002/mrm.28856 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38. Brabec J, Szczepankiewicz F, Lennartsson F, Englund E, Pebdani H, Bengzon J, Knutsson L, Westin CF, Sundgren PC, Nilsson M. Histogram analysis of tensor‐valued diffusion MRI in meningiomas: relation to consistency, histological grade and type. Neuroimage Clin. 2022;33:102912. doi: 10.1016/j.nicl.2021.102912 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39. Zhou M, Stobbe R, Szczepankiewicz F, Budde M, Buck B, Kate M, Lloret M, Fairall P, Butcher K, Shuaib A, et al. Tensor‐valued diffusion MRI of human acute stroke. Magn Reson Med. 2024;91:2126–2141. doi: 10.1002/mrm.29975 [DOI] [PubMed] [Google Scholar]
  • 40. Wang E, Wu Y, Cheung JS, Igarashi T, Wu L, Zhang X, Sun PZ. Mapping tissue pH in an experimental model of acute stroke ‐ determination of graded regional tissue pH changes with non‐invasive quantitative amide proton transfer MRI. NeuroImage. 2019;191:610–617. doi: 10.1016/j.neuroimage.2019.02.022 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41. Gubskiy IL, Namestnikova DD, Cherkashova EA, Chekhonin VP, Baklaushev VP, Gubsky LV, Yarygin KN. MRI guiding of the middle cerebral artery occlusion in rats aimed to improve stroke modeling. Transl Stroke Res. 2018;9:417–425. doi: 10.1007/s12975-017-0590-y [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Figures S1–S5

ARRIVE checklist

JAH3-15-e046071-s001.pdf (153.3KB, pdf)

Articles from Journal of the American Heart Association: Cardiovascular and Cerebrovascular Disease are provided here courtesy of Wiley

RESOURCES