New Results

Improving stroke detection with Correlation Tensor MRI

Rita Alves, Rafael Neto Henriques, Leevi Kerkelä, Cristina Chavarrías, Sune N Jespersen, Noam Shemesh
doi: https://doi.org/10.1101/2021.02.20.432088

ABSTRACT

Noninvasively detecting and characterizing modulations in cellular scale micro-architecture is a desideratum for contemporary neuroimaging. Diffusion MRI (dMRI) has become the mainstay methodology for probing microstructure, and, in ischemia, its contrasts have revolutionized stroke management. However, the biological underpinnings of the contrasts observed in conventional dMRI in general and in ischemia in particular are still highly debated since the markers only indirectly reporter on microstructure. Here, we present Correlation Tensor MRI (CTI), a method that rather than measuring diffusion, harnesses diffusion correlations as its source of contrast. We show that CTI can resolve the sources of diffusional kurtosis, which in turn, provide dramatically enhanced specificity and sensitivity towards ischemia. In particular, the sensitivity towards ischemia nearly doubles, both in grey matter and white matter, and unique signatures for neurite beading, cell swelling, and edema are inferred from CTI. The enhanced sensitivity and specificity endowed by CTI bodes well for future applications in biomedicine, basic neuroscience, and in the clinic.

INTRODUCTION

Progressive modulation in tissue micro-architecture is associated with diverse natural neural processes including development1, plasticity2, memory3, learning4,5, connectivity between brain areas6,7, ageing8, and recovery from injury9. Adverse micro-architectural alterations in the neural tissue milieu are also associated with psychiatric disorders such as depression10, neurodegenerative diseases such as Parkinson’s disease11 and Alzheimer’s disease12, and injuries such as ischemic stroke9 and traumatic brain injury13. In ischemic stroke – one of the leading causes of disability and death worldwide14 – a complex cascade of micro-architectural events occurs acutely following a blood vessel occlusion and the ensuing metabolic and aerobic deprivation15. These micro-architectural modifications include neurite beading16, intracellular swelling due to loss of ion homeostasis, cytotoxic edema, and cell death17 (Fig. 1a) followed later by disruptions in the blood-brain barrier, vasogenic edema and tissue clearance15. The extent of these processes later determines the prognosis, the potential for functional recovery18 and success of treatment.

Figure 1.
Figure 1.

Acute ischemic mechanisms, CTI elements illustration and stroke experimental set up. (a) At an acute ischemic phase – after vascular occlusion – microscopic disturbances such as ionic imbalance (dashed), cell swelling (light blue arrows account for an excessive cellular influx of water in endothelial and neuronal cells), with a subsequent neurite beading (purple) are occurring (cytotoxic edema). This event is followed by ionic edema – transcapillary flux of Na+ and water efflux (dark blue arrows). The blood brain barrier remains intact at this stage. (b) Mean diffusivity (MD) and total kurtosis (MK) are presented as (e) conventional diffusion metrics. Although conventional diffusion metrics obtained from single diffusion encoding (SDE) techniques assess quantitative information on tissue heterogeneity, these cannot resolve the different kurtosis sources, hence lacking specificity and hampering an accurate association with biophysical alterations. On the other hand, correlation tensor imaging (CTI), relying on the cumulant expansion of Double Diffusion Encoding (DDE) signal with (c) 90° and 0° combinations of wavevectors, enables the disentanglement of the kurtosis sources (d) (Anisotropic Kurtosis, Kaniso; Isotropic Kurtosis, Kiso; and Microscopic Kurtosis, µK), providing a more specific characterization of the microscopic alterations occurring in ischemic tissue. (f) A photothrombotic model was used in order to induce a well delimited focal infarct. Upon longitudinal incision and target region coordinates identification (S1bf), irradiation on the left hemisphere was performed for 15 minutes. At 3 h post irradiation offset, the brains were fixed via transcardial perfusion, extracted and placed in an NMR tube with Fluorinert for MRI imaging.

Given the progressive nature of micro-architectural modulations in neural tissue, non-invasive mapping of tissue microstructure19,20 plays a pivotal role in assessing enhancements in microstructure e.g., due to learning, as well as adverse effects upon disease or neural injury. Diffusion-weighted magnetic resonance imaging (dMRI) has become the mainstay in contemporary neuroimaging for probing tissue microstructure. The dMRI methodology non-invasively imparts contrast sensitive to diffusion-driven displacement of water molecules in the tissue21, thereby indirectly probing their interaction with the microscopic boundaries imposed by the cellular environment22. In stroke, dMRI provided the first effective means of early stroke detection23,24, thereby facilitating the administration of treatment within the limited time window of operation25. Effectively, dMRI is sensitive to the displacement distribution induced by the microstructural features in the tissue. The displacement distribution can be characterized by an apparent diffusion coefficient (ADC) along a single direction. Diffusion Tensor MRI (DTI26) has been developed to estimate the 3-dimensional diffusion tensor. However, the diffusion tensor represents an approximation of the displacement distribution to a 3D normal distribution, and therefore it can only capture Gaussian diffusion. Diffusion Kurtosis MRI (DKI)27, was thus introduced to quantify the amount of non-Gaussian diffusion in the tissue, thereby providing increased sensitivity towards microstructure. Effectively, DTI quantifies the mean of the distribution of apparent displacement distribution and DKI estimates the non-Gaussianity of the distribution. Indeed, dMRI has completely revolutionized our ability to follow microstructural modulations over extended periods of time.

However, the non-Gaussianity of the displacement distribution can arise from numerous microstructural sources, which are all conflated in conventional dMRI. The main limitation stems from the use of “Single Diffusion Encoding” (SDE28,29): acquisitions imparting sensitivity towards molecular displacement in a single epoch for a specific single orientation. In such a scenario mesoscopic effects such as orientation dispersion of anisotropic microenvironments, the size distribution of microscopic restricted systems, and the microscopic kurtosis itself (arising from structural disorder or restricted diffusion) – all potential consequences of microstructural modulations – are conflated and cannot be resolved without extensive assumptions. The SDE signal thus does not carry sufficient information for distinguishing different microstructural features and resolving these contributions remains elusive. In stroke, for example, effects arising from cross-sectional variations due to neurite beading16,30 (Fig. 1a, purple), cell body swelling15 (Fig. 1a, light blue) or the transcapillary ionic edema17 (Fig. 1a dark blue), are notoriously difficult to decipher31. Notably, the biophysical underpinnings of even the simplest ADC changes observed in stroke32,33 have been vigorously debated for the last three decades16,34,35. The parameters estimated from such SDE approaches, including mean diffusivity (MD), fractional anisotropy (FA), or total kurtosis36 (KT) thus suffer from lack of specificity.

Here, we depart from the convention of mapping the diffusion and kurtosis tensors using SDE, and rather quantitatively map the displacement correlation tensor37 using Double Diffusion Encoding – a method applying two diffusion-sensitizing epochs that probe the correlation between diffusion-driven displacements in different directions29,38,39. The ensuing Correlation Tensor MRI (CTI40) approach facilitates the disentanglement of the different underlying non-Gaussian diffusion sources, namely: anisotropic kurtosis41, Kaniso; isotropic kurtosis42, Kiso; and microscopic kurtosis43, μK (Fig.1c). In particular, Kaniso quantifies the degree of anisotropy in the tissue irrespective of orientation dispersion effects; Kiso quantifies the variance in the sizes of the microscopic diffusion environments; and µK quantifies the degree of non-gassian effects induced by structural disorder30,44 and restricted diffusion45. Each of these sources can represent a fundamentally different property of the underlying microstructure, and therefore their measurement can potentially greatly enhance specificity. We present the first CTI experiments in a stroke model, which evidence unique signatures for the different underlying micro-architectural modulations, including beading, swelling and edema. Importantly, the sensitivity towards stroke detection is dramatically increased by this approach. CTI thus delivers the sought-after specificity and sensitivity towards microstructural features, which can impact our understanding the progressive modulations occurring in tissue in health and disease.

RESULTS

Focal thrombi induction

Upon injection with photosensitive dye and irradiation with the appropriate wavelength (Fig.1f) for generation of reactive oxygen species, that in turn produce severe endothelial damage and thrombi formation, a well-delineated extensive stroke was evidenced in T2 weighted images (Fig. 2b) 3h post-ischemia. The affected area appeared with strong hyperintense contrast, unilaterally covering the barrel cortex and to some extent the hippocampus. The dMRI signals averaged across all acquired directions (powder-averaged39) also clearly delineated the stroke area as hyperintense signals (Fig. 2b). For all five mice that underwent ischemic induction, the stroke area was highly reproducible (Fig. S1). By contrast, no interhemispheric differences could be observed in the control mice that were injected with the photosensitive dye but did not undergo irradiation (Fig.2c-d). Both T2-weighted images and powder-averaged diffusion-weighted images appear symmetric and without abnormal contrasts.

Figure 2.
Figure 2.

Raw T2-weighted, powder-averaged data and histological validation of the lesion. (a, c) T2-weighted images and (b, d) the powder-averaged signal decays computed by averaging the diffusion-weighted signals decays over 135 directions of diffusion wave vectors from representative extracted brains from both stroke and control groups (total b-value = 3000 s/mm2). 9 representative coronal slices are presented from rostral to caudal direction. The T2-weighted contrast shows elevated intensity values around the infarcted region (left hemisphere) gray matter, whereas the powder-averaged contrast presents elevated values for both gray and white matter within the infarcted region. (e) Histological analysis was performed to assess cell damage and validate the lesion. A Nissl Cresyl violet stained brain section is presented, showing the lesion in the subcortical region of the left hemisphere from a representative mouse brain perfused at 3 hours post ischemic stroke. A critical selected GM region in the ipsilesional hemisphere was magnified (53x), presenting cell damage associated with necrosis and a brighter coloured and less packed region in the lesion site associated with cell loss, when compared to the intact contralesional region (left).

A histological evaluation of the stroked area (Fig.2e) clearly demonstrated abnormalities in Nissl Cresyl violet staining. The zoomed in view of the ipsilesional cortex (Fig.2e, right) also shows the reduced staining and density, compared with the contralesional cortex (Fig.2e, left).

Conventional contrast in stroke

Total kurtosis and Mean diffusivity. As in prior studies46,47, we observed strongly elevated total kurtosis values in the stroked group. Fig.3a shows the clearly elevated values of total kurtosis (KT) in a representative mouse brain in the affected hemisphere, as well as reduced mean diffusivity and slightly reduced fractional anisotropy (Fig.S2). The Kt values appeared elevated in both gray matter (GM) and white matter (WM), and the contrast was more apparent compared with the mean diffusivity (MD) contrast. On the other hand, no interhemispheric differences were observed in Kt in the control group (c.f. Fig.3b for a representative control brain) and other diffusion tensor metrics (MD, FA were also symmetric (Fig.S2). However, it is difficult to draw conclusions on the sources for stroke-induced the changes in Kt or MD which could involve any of the mechanisms described in Fig.1a, as also represented in the illustration of kurtosis sources (Fig.3c).

Figure 3.
Figure 3.

Kurtosis sources maps for stroked and control brains. Kurtosis maps ((a, b) KT, (d, e) µK, (g, h) Kiso and (j, k) Kaniso) for three slices from representative brains from the stroke (left) and control (right) groups are presented with respective illustration of each kurtosis source (c, f, i, l). In the stroke group, the ipsilesional hemisphere shows higher KT intensity values in gray matter and a clear distinction between hemispheres is observed in µK values, showing greater intensities in both white and gray matter. Kaniso presents lower values in both white and gray matter within the ipsilesional hemisphere when compared to the contralesional hemisphere.

CTI contrast in stroke

Given that CTI can further disentangle the different sources, we next turn to evaluate the differences observed in CTI’s metrics.

Microscopic Kurtosis source (µK). The µK metric derived from the CTI analysis in the stroked brains – which represents contributions from structural disorder and restricted diffusion to non-Gaussian diffusion – was dramatically elevated in the affected area (Fig.3d). This observation was consistent for all N = 5 mice (Fig. S3), and a visual examination shows that the elevated µK appeared higher both in gray matter and white matter tissues. It is further noteworthy that the stroke was much better delineated in the µK maps compared with the conventional total kurtosis counterparts (c.f. below for quantitative analyses). In the control group, again the µK maps showed no striking interhemispheric differences (Fig.3e), suggesting that the elevated µK in stroke, representing structural disorder / restriction (Fig.3f), is not due to some asymmetric left-right (or other) imaging artifact.

Isotropic kurtosis source (Kiso). The isotropic kurtosis contrast, which represents the variance in diffusion tensor traces in the tissue derived from CTI analysis exhibited increases, mainly evident in white matter, and much less noticeable contrast differences in gray matter (Fig.3g). No apparent differences in white or gray matter tissues were observed between hemispheres for Kiso in the control group (Fig.3h), suggesting that the variance in isotropic tensor magnitudes (Fig.3i) in the stroked group is again not related with some imaging source.

Anisotropic kurtosis source (Kaniso). The anisotropic kurtosis (Kaniso) contrast in stroke, which is proportional to the degree of intravoxel anisotropy irrespective of orientation dispersion – evidenced strong reductions, more evident in the ipsilesional gray matter (Fig.3j) but also showing some reductions in white matter. In the control group, the same Kaniso contrast did not appear different between the hemispheres (Fig.3k). In other words, the gray matter in the stroked hemisphere was characterized by decreased local anisotropy, independent of local orientation (Fig.3l).

Quantitative sensitivity analyses of different kurtosis sources in stroke

To assess whether the CTI metrics provide a more sensitive evaluation of stroke, we analysed the percent changes for all affected voxels (gray matter and white matter combined (Fig.S7)) as well as for gray matter and white matter separately. Fig.5a shows the magnitude of the effects (both with respect to the contralesional hemisphere in the stroked group as well as the ipsilateral hemisphere in control mice). While total kurtosis changes ∼40% of its nominal value in stroke, the µK contrast in GM+WM was nearly double, with ∼80% change of its nominal values. While Kt changes were quite similar between GM and WM, the µK changed more dramatically in WM (∼100%) and still very strongly in GM (∼75%). Kiso on the other hand, changed only by ∼40% in general, with higher affinity to WM (∼70% change but with large standard error). Finally, Kaniso decreased by over 50% overall, with nearly −90% changes in GM and −30% changes in WM.

We then asked whether CTI metrics show a higher count of involved voxels compared with the conventional Kt. The analysis (Fig.5b) clearly demonstrates that µK is able to detect more affected voxels (∼17% more) overall, with a higher affinity towards GM, while Kt is more sensitive in WM. Kiso and Kaniso do not detect more affected voxels than Kt (Fig.5b). Note that metrics were compared between ipsi and contralateral hemisphere as well as with controls that did not undergo inschemic induction, to ensure that putative inter-hemispheric effects at 3h post ischemia are not a confounding factor.

Figure 4.
Figure 4.

3D rendering maps for KT and µK. Volume render maps using tricubic smooth interpolation for KT and µK sources in a representative stroke group brain (Supplementary Videos). µK shows a more elevated contrast in the ipsilesional hemisphere (delineating the region affected region by the stroke) in comparison to KT.

Figure 5.
Figure 5.

Interhemispheric percent change according to the stroked hemisphere and quantitative analysis. (a) Mean ipsilesional-contralesional ratio (unstriped pattern bars [ic]) accounting for the mean values of all voxels for different ROIs – gray matter (GM) and white matter (WM) – from five stroked brains (N = 5) and mean ipsilesional-ipsilateral control ratio (striped pattern bars [ii]) accounting for the mean of all voxels for different ROIs from five stroked and five control group hemispheres (mean ± standard error of the mean) were assessed for KT, µK, Kiso, Kaniso and Mean Diffusivity (MD). Percentage change was distinctly more elevated for both GM and WM in µK. No statistically significant differences were reported (p < 0.05); (b) A quantitative analysis was performed based on the number of voxels which present abnormal contrast in the ipsilesional hemisphere when compared to the contralesional hemisphere. Significant differences were reported between µK and Kaniso in total lesion (GM + WM); between µK and Kiso in GM; between Kiso and Kaniso in GM (p < 0.05). µK presented in our study more abnormal voxels accounting for total lesion and GM, whereas KT presented more abnormal voxels in WM.

In vivo CTI in stroke

To ensure that the effects above were not dominated by fixation effects upon perfusion of the brains, and to show the in vivo feasibility of the approach we repeated the same experiments at the 3h time point post-ischemia in an in vivo setting. Fig. S8 shows that the same trends as observed in the ex vivo results reported above persist; most importantly, the dramatic elevation of µK persists in all tissues, Kiso is increased in WM, and Kaniso decreases in GM. Therefore, all the trends above are reproduced in vivo.

DISCUSSION

Alterations in neural tissue micro-architectural features accompany a multitude of biological processes ranging from learning, development and plasticity to aging, neurodegeneration and neural injury. Therefore, longitudinally and noninvasively characterizing such time-dependent processes has been a desideratum for contemporary neuroimaging. Indeed, diffusion-driven methods such as DWI, DTI and DKI have played a crucial role in stroke evaluation owing to the microscopic-scale sensitivity arising from water diffusion within cellular-scale barriers. For example, the diffusion coefficient (or tensor) based methods can detect the stroke at very early stages48, while the DKI counterparts show larger lesion sizes47. However, these methods suffer from a degeneracy that prevents them from being specific towards microstructural features.

Here, the CTI methodology was shown to impart both sensitivity and specificity towards microscopic modulations within the imaged voxels upon ischemia. By relying on the measurement of the 4th order displacement correlation tensor using DDE, CTI disentangles the features contributing to non-Gaussian diffusion into the underlying sources: anisotropic, isotropic and microscopic kurtosis sources (while also providing the more conventional metrics typically used for stroke imaging (e.g., MD, KT, FA)). Indeed, the CTI approach dramatically improved the detection and characterization of the stroked region. In particular, (some of) its metrics provided a much more sensitive detection of the ischemic features, while the combination of its parameters provided enhanced specificity and insights into the underlying micro-architectural modulations in the tissue.

For instance, µK showed a dramatic increase and much higher sensitivity to the ischemia compared to its conventional KT counterpart. Indeed, the µK maps revealed both much stronger contrast than the conventional metrics as well as larger areas affected by the stroke. The reason for this enhanced sensitivity can be ascribed to the positive (µK, Kiso) and negative (Kaniso) effects that negate each other in Kt, but not in µK, which reflects the microstructure without the confounds of local anisotropy effects. Therefore, µK is a strong candidate for earlier detection of ischemia and better assessment of potential stroke severity, functional outcomes and prognosis.

From a more biological perspective, the subacute phase of ischemia investigated in this study is characterized by three major micro-architectural modulations occurring downstream of the disruption in oxygen supply. In particular (1) neurite beading16 due to ongoing excitotoxicity and activated microglia; and (2) edema formation17 (more free water in the tissue) and disruption of extracellular/intracellular ionic balance; and (3) cell swelling15 (enlargement of cellular structures). It is thus instructive to assess how these processes could affect the CTI metrics, at least qualitatively.

Figure 6 shows simulations for different micro-architectural scenarios. The upper panel describes the different CTI metrics for different degrees of beading. Interestingly, we find clear signatures for increased beading in CTI signals. The MD decreases (consistent with Budde et al.16), while total kurtosis increases, which can now be explained by strong increases in µK accompanied by small decreases in Kaniso (Kiso remains zero for all beaded scenarios by definition40). This, importantly, contrasts with the scenario of edema formation (c.f. Fig.6b), which shows that as the edema fraction increases (from only beads to only free water, 0 < fedema < 1), µK and Kaniso decrease in a similar fashion, while MD increases and total kurtosis decreases. In this case, Kiso first increases as the diffusivity difference between the beads and the more freely diffusing water is initially large but as the free fraction begins to dominate, the variance in diffusivities is strongly skewed to free diffusion, culminating in totally free diffusion, which has zero variance (Kiso = 0). Finally, if we consider edema formation in the presence of “swelling” of disconnected spherical objects (e.g., representing spheres, Fig. 6c), Kaniso remains constant and zero while KT and Kiso evidence maxima in their values, and µK is negative, and increasing in value as edema fraction increases. Hence, CTI can provide unique insights into the underlying modes of micro-architectural modulations and edema formation.

Figure 6.
Figure 6.

Kurtosis sources estimates with beading and restricted environment (beading and swelling with free water fraction) scenarios. (a) MC random walk simulations to study the effect of axon beading on the diffusion kurtosis parameters. Five simulated axons were generated by increasing A (which controls the amplitude of beading) from 0, i.e., modulation of axon without beading, to 0.8 (length = 31.4 μm). The respective conventional and novel diffusion kurtosis metrics are presented for all simulated axons scenarios (from left to right: MD [µm2/ms], KT, µK, Kiso, Kaniso). (b) Resulting kurtosis estimates from simulated beading effects (A = 0.8) as the free water fraction increases (modulating tissue beading as edema increases). Note that Kiso is zero by definition. (c) Resulting kurtosis estimates from simulated fully restricted spheres compartments with a GPD approximation size distribution (µ = 0.75 ± 0.2 µm) over a range of 100 b-values as free water fraction increases (modulating cell swelling as edema increases).

Interestingly, the introduction of a more rapidly diffusing component into the system (e.g., edema) also explains the stronger WM changes observed in Kiso: in WM, the microstructure comprises much smaller objects (e.g., axons, myelin etc.) that are characterized by a very low diffusivity. When water with higher diffusivity is introduced to the system, the isotropic variance will be strongly impacted as noted in Fig.6. By contrast, in gray matter, the initial diffusivity is higher, thereby the variance in diffusivities upon the introduction of a more rapidly diffusing component (e.g., edema) would be smaller than in WM, thereby also making the changes in Kiso lower in GM upon ischemia, consistent with our observations.

The anisotropic kurtosis source, Kaniso showed a large decrease both in GM and in WM, with much larger decreases in the former. Kaniso reflects the microscopic anisotropy without the conflating effects of orientation dispersion suggesting that the spins diffuse in less anisotropic structures upon ischemia, and as shown in Figure 6, a decrease in Kaniso would be commensurate with increased beading and edema formation.

Our findings also have several implications for dMRI modelling in health and disease. First, the often-ignored µK, clearly needs to be taken into account when modelling dMRI signals in general, as its contribution could clearly be important when looking for changes upon e.g. neurodegeneration. Ignoring µK will result in its contributions becoming conflated into the other kurtosis sources; this in turn could mislead the interpretation of the observed changes in micro-architecture. Finally, it is important to state that characterization of penumbra49 – which is a major goal for stroke imaging, could strongly benefit from CTI, as mismatches between the different metrics could be plotted and contrasted with, e.g., changes in perfusion. In addition, the enhanced sensitivity and specificity towards microstructural and edema effects could be used for developing and assessing novel therapies for stroke. More generally, we expect that the CTI metrics will transcend the particular model used here, namely experimental ischemia, and can be generalized towards other important biological mechanisms underlying different diseases, aging, or changes in microstructure associated with development, learning and maturation. All these features bode well for the application of CTI in basic and applied research in the future.

METHODS

All animal experiments were preapproved by the competent institutional and national authorities and carried out according to European Directive 2010/63.

Animals

Adult male C57BL/6 J mice (aged 11 weeks, weights 24-29 g, grown with a 12 h /12 h light/dark cycle with ad libitum access to food and water) were used in this study.

Surgical procedures

A photothcrombotic Rose Bengal stroke model50 was used to induce a focal infarction in the barrel cortex (S1bf), previously identified upon anatomic mapping, with a solution of Rose Bengal dye (Sigma Aldrich, 95%) dissolved in sterile saline (15 mg/ml) and filtrated through a 0.2 μm sterile filter. Each mouse was injected with meloxicam (Nonsteroidal anti-inflammatory drug) 30 min prior to surgery and anesthetized with isoflurane (∼ 2.5% in 28% oxygen). Temperature was monitored and maintained at 36 – 37°C with a heating pad. The skull was exposed by a median incision of the skin at the dorsal aspect of the head and the periosteum was removed.

The solution was delivered intravenously by a retroorbital injection (10µl/g animal weight). The animals were then irradiated with a cold light source and a fiber optic light guide (0.89 mm tip) – reaching a colour temperature of 3200 K and a beam light intensity of 10 W/cm2 – in the barrel cortex (1.67 mm posterior; 3 mm lateral to bregma51) for 15 minutes. A sham group (N = 5) underwent the same conditions except for the lesion-inducing illumination, yet the 15-minute interval post injection was respected.

Brain extraction and sample preparation. Brain specimens were fixed via transcardial perfusion with 4% Paraformaldehyde (PFA) and extracted from ten adult mice (N = 10) at 3 h post ischemic onset. After extraction from the skull, the brains were immersed in a 4% PFA solution for 24 h and washed in a Phosphate-Buffered Saline solution afterwards, preserved in for at least 24 h. The specimens were subsequently placed in a 10-mm NMR tube filled with Fluorinert (Sigma Aldrich, Lisbon, PT), secured with a stopper to prevent from floating. The tube was then sealed with paraffin film.

Diffusion MRI acquisition

All ex vivo MRI scans were performed on a 16.4 T Aeon Ascend Bruker scanner (Karlsruhe, Germany) equipped with an AVANCE IIIHD console and a Micro5 probe with gradient coils capable of producing up to 3000 mT/m in all directions and a birdcage RF volume coil. Once inserted, the samples were maintained at 37°C due to the probe’s variable temperature capability and were allowed to acclimatize with the surroundings for at least 2 h prior to the beginning of diffusion MRI experiments. In favour of shimming, a B0 map was also acquired covering the volume of the brain.

Double diffusion encoding (DDE) data were subsequently acquired for 25 coronal slices using an in house written EPI-based DDE pulse sequence implemented in Paravision 6.0.1 (Bruker BioSpin MRI GmbH, Ettlingen, Germany) and TopSpin 3.1. The diffusion encoding gradient pulse separation Δ and mixing time τm were set to 10 ms, and the pulsed gradient duration δ was set to 1.5 ms. Acquisitions were repeated for five q1-q2 magnitude combinations (1498 - 0, 1059.2 - 1059.2, 1059.2 - 0, 749.0 - 749.0 and 749.0 - 0 mT/m) corresponding to b-values of 3000 - 0, 1500 - 1500, 1500 - 0, 750 - 750 and 750 - 0 s/mm2. Acquisition with q2=0 are acquired for 135 directions for q1, while equal magnitude q1-q2 combinations were repeated for 135 parallel and perpendicular directions (Fig. 3.10) as described by Henriques et al.52. These DDE acquisition-protocol are adjusted according to the requirements for CTI reconstruction40,52. In addition, twenty acquisitions without any diffusion-weighted sensitization (zero b-value) were performed for each q1-q2 magnitude combinations to guarantee a high ratio between the number of non-diffusion and diffusion-weighted acquisitions. For all experiments, the following parameters were used: TR/TE = 3000/49 ms, Field of View = 11 × 11 mm2, matrix size 78 × 78, resulting in an in-plane voxel resolution of 141 × 141 μm2, slice thickness = 0.5 mm, number of segments = 2, number of averages = 8. For every bmax value, the total acquisition time was approximately 1 h and 57 min.

Structural MRI acquisition

Axial and sagittal T2-weighted images with high resolution and high SNR were acquired for anatomical reference. These data were acquired using RARE pulse sequences with the following parameters: TR = 4000 ms, TE = 50 ms, RARE factor = 12, number of averages = 6. Concerning the axial images, the Field of View (FOV) was 18 × 18 mm2, and the matrix size was 240 × 132, resulting in an in-plane voxel resolution of 75 × 76 μm2. For the sagittal images, FOV was set to 18 × 9 mm2, matrix size to 240 × 120, and subsequently in-plane voxel resolution resulted in 75 × 75 μm2 (sagittal). Both axial and sagittal acquisitions sampled the total of 33 slices with a thickness of 0.3 mm. For the coronal images, FOV was established as 10.5 × 10.5 mm2, the matrix size as 140 × 140, hence an in-plane voxel resolution of 75 × 75 μm2, for 72 slices 0.225 mm thick. The MRI infarct volumes were calculated by the amount of voxels within the designated masks multiplied by the volume of each voxel.

Diffusion data pre-processing

Before starting the diffusion data analysis, masks for delineation were manually drawn slice-by-slice using Matlab (Matlab R2018b). All datasets were corrected for Gibbs ringing artifacts53,54 in Python (Dipy, version 1.055) which suppresses the Gibbs oscillations effects based on sub-voxel Fourier shifts (total variance analysis across three adjacent points for each voxel used to access Gibbs oscillation suppression). All diffusion-weighted datasets underwent realignment via a sub-pixel registration method56 in which each set of data for every total diffusion b-value would be realigned to a counter defined dataset with similar DDE gradient pattern combinations.

CTI reconstruction

CTI was then directly fitted to the data using a linear-least squares fitting procedure implemented to the following equation40: Embedded Image where Dij, Wijkl and Cijkl correspond to the diffusion, kurtosis and covariance tensors, respectively. The sources of kurtosis can be extracted from these tensors in the following way (Henriques et al., 2020): 1) the total kurtosis KT can be computed from Wijkl; 2) the two inter-compartmental kurtosis sources (Kaniso and Kiso) can be extracted from Cijkl; and 3) an intra-compartmental kurtosis source Kintra can be estimated as KT − Kaniso − Kiso.

Region of Interest Analysis

A region of Interest (ROI) analysis was performed by manual selection of the most relevant areas for every diffusion data slice in both hemispheres in all brains, containing the total lesion and counterpart region in the opposite hemisphere, and subsequent threshold selection of gray and white matter within the manually selected ROIs. For each brain sample which had previously undergone ischemic insult and considering the previously selected ROI covering the total lesion area in the MD measure map, an identically sized ROI (and symmetrical in location to the former) was manually drawn. Identical regions were then delineated on the sham group hemispheres. Within these more general ROIs, GM and WM regions were further selected according to threshold values pre-established for FA and MD measures (Sensitivity assessment) To assess which metrics were more sensitive to stroke, total ROIs were also manually drawn for each considered metric according to interhemispheric visual asymmetry (Fig.S6-7). White matter ROIs were manually drawn within the total ROI according to FA maps. All ROI analyses were performed in Matlab (Matlab R2018b).

Statistical analysis features

A statistical specificity analysis was conducted to assess whether the differences between mean values in each ROI for each metric averaged across mice (from each group) were significant through a one-way ANOVA test, and a Bonferroni correction was performed for multiple comparisons across the different diffusion metrics. For the sensitivity analysis, a one-way ANOVA test with a subsequent Bonferroni correction were also performed to test if the differences in interhemispheric ratio percentage changes between ipsilesional-contralesional (stroked brain) and ipsilesional (stroked brain)-ipsilateral (control brain), as calculated below, were significant. Embedded Image A similar analysis was performed to assess significant differences between the number of voxels which, based on manually drawn ROIs, were abnormal when compared with the contralesional hemisphere of the stroked group and for all kurtosis sources. All statistical analysis were performed in Matlab (Matlab R2018b).

Data visualization

All volumetric datasets were rendered with ImageJ. Volume render maps were produced using the plugin Volume Viewer57, with tricubic smooth interpolation for KT and µK sources.

Simulations

Monte Carlo random walk simulations were performed to study the effect of axon beading on the diffusion kurtosis parameters. An axon without beading was modelled as an impermeable cylinder (length = 31.4 μm, radius = 1 μm) aligned with the z-axis. Beading was introduced by making the radius depend on the location along the z-axis: r(z) = r0 + A · sin(B · z), where r0 is the mean radius, A ∈ [0,1] controls the amplitude of beading, and B controls the frequency and location of beading. Five simulated axons, shown in Figure X, were generated by increasing A from 0 to 0.8 in equal steps and adjusting r0 to keep the volume constant. B = 1 μm−1 for all simulated axons (Fig.6a).

Simulated data was generated using Disimpy58 with 105 random walkers, 104 time steps, and diffusivity of 2 μm2/ms. The initial positions of the random walkers were randomly sampled from a uniform distribution inside the axons. MR signals were generated using pulsed gradient single diffusion encoding with δ = 1.5 ms and Δ = 10 ms, 60 diffusion encoding directions uniformly distributed over the surface of half a sphere, and 10 b-values uniformly distributed between 0 and 3 ms/μm2. The diffusion and kurtosis tensors were estimated from the simulated signals using a weighted linear least squares fit in Dipy55. MD, μK, Kiso, and Kaniso were calculated from the diffusion tensor eigenvalues and the elements of the kurtosis tensor40 for individual components (Fig.6a).

Analytically computed pulsed gradient spin echo sequence signals were simulated for two different models using the MISST toolbox59–62: beading (to model beaded axons) with formation of edema, assuming the most beaded previously simulated scenario with A = 0.8 (Fig.6b); spherical compartments with restricted diffusion (to model cell swelling) and formation of edema, with combinations of radii between 0.8 µm and 10 µm, with a diffusivity value of 2 μm2/ms (Fig.6c).

Nissl Cresyl Violet staining

Histological analysis was performed in one of the stroked ex vivo samples. Slices were obtained through Vibratome sectioning with a thickness of 0.04 mm and Mowiol containing 2.5 % 1,4 diazobicyclo-[2.2.2]-octane (DABCO, Sigma, D2522) was used as the mounting media. The brain sections were then fixed with 10% formalin and processed with Nissl-Cresyl Violet staining for microscopy in order to assess tissue damage and cell loss within the infarcted region.

Optical imaging

Histological imaging was performed with a ZEISS Axio Scan.Z1 (Zeiss, Germany) coupled to a Hitachi 3 CCD colour camera and processed with QuPath 0.2.3 (Fig.2e). Images were magnified (53×) for both ipsi- and contralesional hemispheres.

Animal monitoring for in vivo MRI imaging

For anaesthesia induction, the body temperature of the mouse was kept constant by placing the animal on top of an electrical heating pad. Anaesthesia with a mixture of medical air and 4% isoflurane (Vetflurane, Virbac, France) was maintained until the animal righting reflex and any reaction to firm foot squeeze were lost. The isoflurane concentration was regulated and reduced to 2.5%. The mouse was then weighed and transferred to the animal bed, prone positioned above a heated water pad – in order for the mouse body temperature not to oscillate during the experiments –, having its head placed with its upper incisors held on to a mouth bite bar. Oxygen concentrations were kept between 27% and 28%, monitored by a portable oxygen monitor (MX300-I, Viamed, United Kingdom). Ear bars were used for a safe and efficient head fixation (into external meatus) and eye ointment (Bepanthen Eye Drops, Bepanthen, Germany) was applied to prevent the corneas from drying. A rectal temperature probe and a respiration sensor (Model 1030 Monitoring Gating System, SAII, United States of America) were placed for real-time monitoring of these physiological measurements to guarantee the animal’s welfare and immobilization. Considering the water molecules sensitivity towards temperature alterations, the waterbed temperature was cautiously monitored and controlled to avoid oscillations. Respiration rates were also monitored and maintained at physiological levels throughout dMRI scanning.

In vivo MRI acquisiontions The in vivo MRI data were acquired on a 9.4 T horizontal MRI scanner (BioSpec 94/20 USR, Bruker BioSpin, Germany) equipped a gradient system able to produce up to 660 mT/m in every direction, an 86 mm quadrature coil for transmission and a 4-element array surface cryocoil for reception.

Sagittal T2-weighted images were acquired for anatomical reference using a RARE pulse sequence with the following parameters: TR = 2000 ms, TE = 36 ms, RARE factor = 8, number of averages = 8. The field of view was 24 × 16.1 mm2, the matrix size was 256 × 256, resulting in an in-plane voxel resolution of 150 × 150 μm2. The slice thickness was 0.5 mm, and 21 slices were sampled.

Following an optimized protocol when compared to the ex vivo experiment, DDE data were acquired for 5 coronal slices using our in house written EPI-based DDE pulse sequence. The diffusion encoding gradient pulse separation Δ and mixing time τm were set to 10 ms, and the pulsed gradient duration δ was set to 4 ms. Acquisitions were repeated for five q1-q2 magnitude combinations (518.79 - 0, 366.84 – 366.84.2, 366.84 - 0, 259.4 – 259.4 and 259.4 - 0 mT/m) corresponding to b-values of 3000 - 0, 1500 - 1500, 1500 - 0, 750 - 750 and 750 - 0 s/mm2. For each gradient combination, experiments are repeated for the same directions in the ex vivo experiments (Figure 3.10) and described by Henriques et al.52. In addition, twenty acquisitions without any diffusion-weighted sensitization (b-value = 0) were performed. For all experiments, the following parameters were used: TR/TE = 2800/44.5 ms, FOV = 12 × 12 mm2, matrix size 78 × 78, resulting in an in-plane voxel resolution of 181 × 181 μm2, slice thickness = 0.85 mm, number of segments = 1, number of averages = 1 and partial Fourier acceleration of 1.25. For every bmax value, the total acquisition time was approximately 7 min.

Data availability

The data sets generated and analysed during the current study are available from the corresponding author upon reasonable request.

Code availability

Custom MATLAB code for dMRI pre- and post-processing of data is available from the corresponding author upon reasonable request.

ACKNOWLEDGMENTS

This study was funded by the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (Starting Grant, agreement No. 679058). The authors acknowledge the vivarium of the Champalimaud Centre for the Unknow, a facility of CONGENTO financed by Lisboa Regional Operational Programme (Lisboa 2020), project LISBOA01-0145-FEDER-022170, and also the Champalimaud Histopathology and the Champalimaud ABBE Platforms.

REFERENCES

  1. 1.↵
    Suter, T. A. C. S. & Jaworski, A. Cell migration and axon guidance at the border between central and peripheral nervous system. Science (80-). 365, eaaw8231 (2019).
    OpenUrlAbstract/FREE Full Text
  2. 2.↵
    Hughes, E. G., Orthmann-Murphy, J. L., Langseth, A. J. & Bergles, D. E. Myelin remodeling through experience-dependent oligodendrogenesis in the adult somatosensory cortex. Nat. Neurosci. 21, 696–706 (2018).
    OpenUrlCrossRefPubMed
  3. 3.↵
    Brodt, S. et al. Fast track to the neocortex: A memory engram in the posterior parietal cortex. Science (80-.). 362, 1045–1048 (2018).
    OpenUrlAbstract/FREE Full Text
  4. 4.↵
    Zatorre, R. J., Fields, R. D. & Johansen-Berg, H. Plasticity in gray and white: neuroimaging changes in brain structure during learning. Nat. Neurosci. 15, 528–536 (2012).
    OpenUrlCrossRefPubMed
  5. 5.↵
    Scholz, J., Klein, M. C., Behrens, T. E. J. & Johansen-Berg, H. Training induces changes in white-matter architecture. Nat. Neurosci. 12, 1370–1371 (2009).
    OpenUrlCrossRefPubMedWeb of Science
  6. 6.↵
    Craddock, R. C. et al. Imaging human connectomes at the macroscale. Nat. Methods 10, 524–539 (2013).
    OpenUrlCrossRefPubMedWeb of Science
  7. 7.↵
    Pestilli, F., Yeatman, J. D., Rokem, A., Kay, K. N. & Wandell, B. A. Evaluation and statistical inference for human connectomes. Nat. Methods 11, 1058–1063 (2014).
    OpenUrlCrossRefPubMedWeb of Science
  8. 8.↵
    Hill, R. A., Li, A. M. & Grutzendler, J. Lifelong cortical myelin plasticity and age-related degeneration in the live mammalian brain. Nat. Neurosci. 21, 683–695 (2018).
    OpenUrlCrossRefPubMed
  9. 9.↵
    Murphy, T. H. & Corbett, D. Plasticity during stroke recovery: From synapse to behaviour. Nat. Rev. Neurosci. 10, 861–872 (2009).
    OpenUrlCrossRefPubMedWeb of Science
  10. 10.↵
    Li, N. et al. mTOR-Dependent Synapse Formation Underlies the Rapid Antidepressant Effects of NMDA Antagonists. Science (80-.). 329, 959–964 (2010).
    OpenUrlAbstract/FREE Full Text
  11. 11.↵
    Peelaerts, W. et al. α-Synuclein strains cause distinct synucleinopathies after local and systemic administration. Nature 522, 340–344 (2015).
    OpenUrlCrossRefPubMed
  12. 12.↵
    De Strooper, B. & Karran, E. The Cellular Phase of Alzheimer’s Disease. Cell 164, 603–615 (2016).
    OpenUrlCrossRefPubMed
  13. 13.↵
    Budde, M. D., Janes, L., Gold, E., Turtzo, L. C. & Frank, J. A. The contribution of gliosis to diffusion tensor anisotropy and tractography following traumatic brain injury: validation in the rat using Fourier analysis of stained tissue sections. Brain 134, 2248– 2260 (2011).
    OpenUrlCrossRefPubMedWeb of Science
  14. 14.↵
    Mathers, C. et al. WHO methods and data sources for country-level causes of death 2000-2015. (2017). Available at: https://www.who.int/healthinfo/global_burden_disease/GlobalCOD_method_2000_2015.pdf?ua=1.
  15. 15.↵
    Xing, C., Arai, K., Lo, E. H. & Hommel, M. Pathophysiologic cascades in ischemic stroke. Int. J. Stroke 7, 378–385 (2012).
    OpenUrlCrossRefPubMedWeb of Science
  16. 16.↵
    Budde, M. D. & Frank, J. A. Neurite beading is sufficient to decrease the apparent diffusion coefficient after ischemic stroke. Proc. Natl. Acad. Sci. U. S. A. 107, 14472– 14477 (2010).
    OpenUrlAbstract/FREE Full Text
  17. 17.↵
    Stokum, J. A., Gerzanich, V. & Simard, J. M. Molecular pathophysiology of cerebral edema. J. Cereb. Blood Flow Metab. 36, 513–538 (2016).
    OpenUrlCrossRefPubMed
  18. 18.↵
    Zeiler, S. R. et al. Paradoxical Motor Recovery From a First Stroke After Induction of a Second Stroke. Neurorehabil. Neural Repair 30, 794–800 (2016).
    OpenUrlCrossRefPubMed
  19. 19.↵
    Veraart, J. et al. Noninvasive quantification of axon radii using diffusion MRI. Elife 9, (2020).
  20. 20.↵
    Andersson, M. et al. Axon morphology is modulated by the local environment and impacts the noninvasive investigation of its structure–function relationship. Proc. Natl. Acad. Sci. 117, 33649–33659 (2020).
    OpenUrlAbstract/FREE Full Text
  21. 21.↵
    Le Bihan, D. Looking into the functional architecture of the brain with diffusion MRI. Nat. Rev. Neurosci. 4, 469–480 (2003).
    OpenUrlCrossRefPubMedWeb of Science
  22. 22.↵
    Novikov, D. S., Jensen, J. H., Helpern, J. A. & Fieremans, E. Revealing mesoscopic structural universality with diffusion. Proc. Natl. Acad. Sci. 111, 5088–5093 (2014).
    OpenUrlAbstract/FREE Full Text
  23. 23.↵
    Moseley, M. E. et al. Early detection of regional cerebral ischemia in cats: Comparison of diffusion- and T2-weighted MRI and spectroscopy. Magn. Reson. Med. 14, 330–346 (1990).
    OpenUrlCrossRefPubMedWeb of Science
  24. 24.↵
    Karbasforoushan, H., Cohen-Adad, J. & Dewald, J. P. A. Brainstem and spinal cord MRI identifies altered sensorimotor pathways post-stroke. Nat. Commun. 10, 3524 (2019).
    OpenUrl
  25. 25.↵
    Clark, W. M. et al. Recombinant Tissue-Type Plasminogen Activator (Alteplase) for Ischemic Stroke 3 to 5 Hours After Symptom Onset. JAMA 282, 2019 (1999).
    OpenUrlCrossRefPubMedWeb of Science
  26. 26.↵
    Basser, P. J., Mattiello, J. & LeBihan, D. MR diffusion tensor spectroscopy and imaging. Biophys. J. 66, 259–267 (1994).
    OpenUrlCrossRefPubMedWeb of Science
  27. 27.↵
    Jensen, J. H., Helpern, J. A., Ramani, A., Lu, H. & Kaczynski, K. Diffusional kurtosis imaging: The quantification of non-gaussian water diffusion by means of magnetic resonance imaging. Magn. Reson. Med. 53, 1432–1440 (2005).
    OpenUrlCrossRefPubMedWeb of Science
  28. 28.↵
    Stejskal, E. O. & Tanner, J. E. Spin diffusion measurements: Spin echoes in the presence of a time-dependent field gradient. J. Chem. Phys. 42, 288–292 (1965).
    OpenUrlCrossRefWeb of Science
  29. 29.↵
    Shemesh, N. et al. Conventions and Nomenclature for Double Diffusion Encoding NMR and MRI. Magn. Reson. Med. 87, 82–87 (2016).
    OpenUrl
  30. 30.↵
    Lee, H. H., Papaioannou, A., Kim, S. L., Novikov, D. S. & Fieremans, E. A time-dependent diffusion MRI signature of axon caliber variations and beading. Commun. Biol. 3, (2020).
  31. 31.↵
    Le Bihan, D. & Iima, M. Diffusion Magnetic Resonance Imaging: What Water Tells Us about Biological Tissues. PLOS Biol. 13, e1002203 (2015).
    OpenUrlCrossRefPubMed
  32. 32.↵
    Moseley, M. E. et al. Diffusion-weighted MR imaging of acute stroke: Correlation with T2-weighted and magnetic susceptibility-enhanced MR imaging in cats. Am. J. Neuroradiol. 11, 423–429 (1990).
    OpenUrlAbstract/FREE Full Text
  33. 33.↵
    Baron, C. A. et al. Reduction of Diffusion-Weighted Imaging Contrast of Acute Ischemic Stroke at Short Diffusion Times. Stroke 46, 2136–2141 (2015).
    OpenUrlAbstract/FREE Full Text
  34. 34.↵
    Novikov, D. S., Fieremans, E., Jespersen, S. N. & Kiselev, V. G. Quantifying brain microstructure with diffusion MRI: Theory and parameter estimation. NMR Biomed. 32, e3998 (2019).
    OpenUrl
  35. 35.↵
    Szafer, A., Zhong, J. & Gore, J. C. Theoretical Model for Water Diffusion in Tissues. Magn. Reson. Med. 33, 697–712 (1995).
    OpenUrlCrossRefPubMedWeb of Science
  36. 36.↵
    Weber, R. A. et al. Diffusional kurtosis and diffusion tensor imaging reveal different time-sensitive stroke-induced microstructural changes. Stroke 46, 545–550 (2015).
    OpenUrlAbstract/FREE Full Text
  37. 37.↵
    Nørhøj, S. & Buhl, N. The displacement correlation tensor: Microstructure, ensemble anisotropy and curving fibers. J. Magn. Reson. Imaging 208, 34–43 (2011).
    OpenUrl
  38. 38.↵
    Mitra, P. P. Multiple wave-vector extensions of the NMR pulsed-field-gradient spin-echo diffusion measurement. Phys. Rev. 51, 74–78 (1995).
    OpenUrl
  39. 39.↵
    Jespersen, S. N., Lundell, H., Sønderby, C. K. & Dyrby, T. B. Orientationally invariant metrics of apparent compartment eccentricity from double pulsed field gradient diffusion experiments. NMR Biomed. 26, 1647–1662 (2013).
    OpenUrl
  40. 40.↵
    Henriques, R. N., Jespersen, S. N. & Shemesh, N. Correlation tensor magnetic resonance imaging. Neuroimage 211, (2020).
  41. 41.↵
    Kerkelä, L., Henriques, R. N., Hall, M. G., Clark, C. A. & Shemesh, N. Validation and noise robustness assessment of microscopic anisotropy estimation with clinically feasible double diffusion encoding MRI. Magn. Reson. Med. 1–13 (2019). doi:10.1002/mrm.28048
    OpenUrlCrossRef
  42. 42.↵
    Nilsson, M. et al. Tensor-valued diffusion MRI in under 3 minutes: an initial survey of microscopic anisotropy and tissue heterogeneity in intracranial tumors. Magn. Reson. Med. 83, 608–620 (2020).
    OpenUrl
  43. 43.↵
    Callaghan, P. T. Pulsed-Gradient Spin-Echo NMR for Planar, Cylindrical, and Spherical Pores under Conditions of Wall Relaxation. J. Magn. Reson. Ser. A 113, 53–59 (1995).
    OpenUrl
  44. 44.↵
    Novikov, D. S. & Kiselev, V. G. Effective medium theory of a diffusion-weighted signal. NMR Biomed. 23, 682–697 (2010).
    OpenUrlCrossRefPubMedWeb of Science
  45. 45.↵
    Callaghan, P. T., Coy, A., MacGowan, D., Packer, K. J. & Zelaya, F. O. Diffraction-like effects in NMR diffusion studies of fluids in porous solids. Nature 351, 467–469 (1991).
    OpenUrlCrossRefWeb of Science
  46. 46.↵
    Hui, E. S. et al. Stroke assessment with diffusional kurtosis imaging. Stroke 43, 2968– 2973 (2012).
    OpenUrlAbstract/FREE Full Text
  47. 47.↵
    Umesh Rudrapatna, S. et al. Can diffusion kurtosis imaging improve the sensitivity and specificity of detecting microstructural alterations in brain tissue chronically after experimental stroke? Comparisons with diffusion tensor imaging and histology. Neuroimage 97, 363–373 (2014).
    OpenUrlCrossRefPubMed
  48. 48.↵
    Sotak, C. H. The role of diffusion tensor imaging in the evaluation of ischemic brain - A review. NMR Biomed. 15, 561–569 (2002).
    OpenUrlCrossRefPubMedWeb of Science
  49. 49.↵
    Astrup, J., Siesjö, B. K. & Symon, L. Thresholds in cerebral ischemia - the ischemic penumbra. Stroke 12, 723–725 (1981).
    OpenUrlFREE Full Text
  50. 50.↵
    Watson, B. D., Dietrich, W. D., Busto, R., Wachtel, M. S. & Ginsberg, M. D. Induction of reproducible brain infarction by photochemically initiated thrombosis. Ann. Neurol. 17, 497–504 (1985).
    OpenUrlCrossRefPubMedWeb of Science
  51. 51.↵
    Franklin, K. & Paxinos, G. Paxinos and Franklin’s the Mouse Brain in Stereotaxic Coordinates. (2019).
  52. 52.↵
    Henriques, R. N., Jespersen, S. N. & Shemesh, N. Evidence for microscopic kurtosis in neural tissue revealed by Correlation Tensor MRI. (2021).
  53. 53.↵
    Henriques, R. N. Advanced Methods for Diffusion MRI Data Analysis and their Application to the Healthy Ageing Brain. (Cambridge University, 2018). doi:10.17863/CAM.29356
    OpenUrlCrossRef
  54. 54.↵
    Kellner, E., Dhital, B., Kiselev, V. G. & Reisert, M. Gibbs-ringing artifact removal based on local subvoxel-shifts. Magn. Reson. Med. 76, 1574–1581 (2016).
    OpenUrlCrossRefPubMed
  55. 55.↵
    Garyfallidis, E. et al. Dipy, a library for the analysis of diffusion MRI data. Front. Neuroinform. 8, (2014).
  56. 56.↵
    Guizar-sicairos, M., Thurman, S. T. & Fienup, J. R. Guizar_Efficent_subpixel_. Opt. Lett. 33, 156–158 (2008).
    OpenUrlCrossRefPubMedWeb of Science
  57. 57.↵
    Barthel, K. U. Volume Viewer. (2012). Available at: https://imagej.net/plugins/volume-viewer.html.
  58. 58.↵
    Kerkelä, L., Nery, F., Hall, M. & Clark, C. Disimpy: A massively parallel Monte Carlo simulator for generating diffusion-weighted MRI data in Python. J. Open Source Softw. 5, 2527 (2020).
    OpenUrl
  59. 59.↵
    Drobnjak, I., Siow, B. & Alexander, D. C. Optimizing gradient waveforms for microstructure sensitivity in diffusion-weighted MR. J. Magn. Reson. 206, 41–51 (2010).
    OpenUrlPubMed
  60. 60.
    Drobnjak, I., Zhang, H., Hall, M. G. & Alexander, D. C. The matrix formalism for generalised gradients with time-varying orientation in diffusion NMR. J. Magn. Reson. 210, 151–157 (2011).
    OpenUrlPubMed
  61. 61.
    Ianuş, A., Siow, B., Drobnjak, I., Zhang, H. & Alexander, D. C. Gaussian phase distribution approximations for oscillating gradient spin echo diffusion MRI. J. Magn. Reson. 227, 25–34 (2013).
    OpenUrl
  62. 62.↵
    Ianuş, A., Alexander, D. C. & Drobnjak, I. Microstructure Imaging Sequence Simulation Toolbox. in 34–44 (2016). doi:10.1007/978-3-319-46630-9_4
    OpenUrlCrossRef
Back to top
Posted February 26, 2021.
Download PDF

Supplementary Material

Email
Improving stroke detection with Correlation Tensor MRI
Rita Alves, Rafael Neto Henriques, Leevi Kerkelä, Cristina Chavarrías, Sune N Jespersen, Noam Shemesh
bioRxiv 2021.02.20.432088; doi: https://doi.org/10.1101/2021.02.20.432088
Reddit logo Twitter logo Facebook logo LinkedIn logo Mendeley logo
Citation Tools

Subject Area