Intravoxel incoherent motion imaging in stroke infarct core and penumbra is related to long-term clinical outcome

Abstract

Intravoxel incoherent motion (IVIM) imaging, a contrast agent-free magnetic resonance imaging technique, enables the evaluation of microvascular perfusion abnormalities in acute stroke. Prior research reported reduced IVIM values within the infarct core in acute stroke. However, findings concerning IVIM characteristics in the penumbra have been mixed and the relationship between IVIM and clinical outcomes remains unknown. We employed a longitudinal multimodal imaging approach for ischemic stroke patients (n analyzed=24; pre-/post-treatment and 90-day post-stroke assessments) including IVIM, diffusion-weighted, and contrast-enhanced perfusion-weighted imaging. We evaluated IVIM in relevant stroke areas after endovascular treatment. Reduced post-treatment IVIM perfusion fraction in infarct core and recanalized penumbra was associated with poorer functional recovery at 90-days post-stroke (NIH Stroke Scale [NIHSS]; r=-0.64 and r=-0.69). Including IVIM perfusion fraction increased the explained variance of NIHSS from 42% up to 83% compared to well-known prognostic factors core volume and patient age. Additionally, IVIM perfusion fraction was reduced in the core and recanalized penumbra compared to contralateral healthy tissue, suggesting impaired microvascular reperfusion after endovascular treatment. In conclusion, IVIM characteristics of the infarct core and recanalized penumbra are strong prognostic factors for long-term outcome in stroke patients and IVIM shows promise for characterizing microvascular perfusion in relevant stroke areas.

Introduction

In recent years, intravoxel incoherent motion (IVIM) imaging has emerged as a promising magnetic resonance imaging (MRI) technique for ischemic stroke imaging. IVIM imaging allows to quantify microvascular perfusion without the need for exogenous contrast agents. This method, originally introduced by Le Bihan et al., segregates the signal acquired from a multi-b-value diffusion sequence into microvascular and nonvascular components¹. Their research confirmed that the blood flow in capillaries resembles a diffusion process, affecting the signal in diffusion MRI and this influence can be segregated to facilitate the quantification of microvascular perfusion². This separation of molecular diffusion and microvascular perfusion effects facilitates the calculation of IVIM parameters, encompassing the perfusion-related pseudo-diffusion coefficient D*, characterizing the movement of blood components within the microvasculature and a perfusion fraction f, quantifying the signal proportion originating from the microvascular compartment³. The IVIM parameters, f, D*, and f·D* (the multiplication of the two parameters), also exhibit a theoretical association with the conventional perfusion metrics, including cerebral blood volume (CBV), mean transit time (MTT), and cerebral blood flow (CBF), respectively, as established by Le Bihan^1,2. However, it is important to note that IVIM is not specific to blood microcirculation. Non-Gaussian diffusion behavior, Rician noise, and microstructural changes such as axonal beading following ischemic stroke can influence the estimation of IVIM parameters⁴.

Several studies have offered valuable insights into IVIM parameters in the context of acute ischemic stroke. Notably, studies have consistently observed reduced IVIM f and fD* values within the infarct core^3,5–9. The findings concerning IVIM D* have been mixed, with certain studies indicating reduced D* in the infarct core^5,8, while others have reported no significant differences^7,9. Alterations in penumbral IVIM characteristics have been investigated in only two studies examining stroke patients before treatment. Federau et al. reported that IVIM f in the penumbra was comparable to healthy tissue⁶, whereas Zhu et al. observed reduced f and fD* in the penumbra compared to healthy tissue⁹. However, while both studies found reduced IVIM f and fD* in the core compared to healthy tissue, both studies did not find significant differences in any IVIM parameters between the core and penumbra.

Federau et al. suggested that their findings indicate preserved microvascular perfusion in the penumbra due to collateral blood flow, which might explain the penumbra’s viability in cases of acute stroke resulting from large vessel occlusion⁶. They further reported that all evident confluent regions of hypoperfusion on IVIM f ultimately advanced to infarction even following thrombectomy treatment, suggesting that these characterize non-salvageable penumbral tissue. Additionally, Zhu et al. reported a strong concordance between the volumes of perfusion abnormality delineated via IVIM and those obtained from conventional contrast-enhanced perfusion-weighted imaging (PWI) parametric maps⁹. They suggested that IVIM alone could adequately delineate the ischemic core and penumbra by utilizing IVIM D (the diffusion component of IVIM, analogous to the apparent diffusion coefficient [ADC]) in conjunction with either IVIM f or fD*, and that the mismatch between IVIM D and f or fD* could serve as a criterion for selecting patients suitable for thrombectomy treatment. Chen et al. were the first group that explored the relationship between IVIM parameters within the stroke core and long-term clinical outcomes⁵. However, their analysis did not yield a statistically significant correlation between IVIM perfusion parameters within the core and the modified Rankin Scale (mRS) at 90 days post-stroke. Noteworthy, they excluded patients who underwent endovascular treatment.

Consequently, prior research predominantly focused on pre-treatment measurements, with inconsistent findings regarding IVIM parameters in the penumbra compared to healthy tissue. Strikingly, the relationship between IVIM characteristics and long-term clinical outcomes in ischemic stroke remains uncertain.

This exploratory study addresses two primary research questions. (i) The study examines the association between IVIM measures in the ischemic core and recanalized penumbra 24 h post-treatment and clinical outcomes at 90 days post-stroke. Chen et al. studied IVIM parameters’ relationship with long-term clinical outcomes but focused solely on the infarct core and used the mRS score⁵. Recognizing the importance of penumbral reperfusion¹⁰ and the NIHSS’s heightened sensitivity to detect minor changes¹¹, this study integrates penumbral segmentation and prioritizes NIHSS scores. Of note, the penumbra in our study was segmented based on pre-treatment PWI scans, but IVIM was assessed in this area only post-treatment, thereby reflecting the recanalized penumbra region. (ii) The present study evaluates the differences in IVIM parameters post-treatment between the core, recanalized penumbra, and healthy contralateral control tissue, with a particular focus on comparing the recanalized penumbra to unaffected contralateral tissue. While Zhu et al. observed reduced f and fD* values in the penumbra compared to healthy tissue before endovascular intervention⁹, Federau et al. did not report such differences⁶. Moreover, in both studies, the perfusion-related IVIM parameters showed no significant difference between the core and penumbra. However, no study has yet assessed IVIM in the recanalized penumbra region after endovascular intervention.

Methods

Study population

Participants from the IMPreST (Interplay of Microcirculation and Plasticity after Ischemic Stroke; registered at clinicaltrials.gov, No. NCT04035746) prospective longitudinal observational cohort study were enrolled in this study. The primary objective of IMPreST was to investigate the association between various imaging modalities for microcirculation and their impact on clinical outcomes in individuals diagnosed with acute ischemic unilateral large vessel occlusion stroke. Eligible participants consisted of individuals who presented with symptoms indicative of their first-ever acute hemispheric large vessel occlusion stroke and were undergoing triage for endovascular acute stroke treatment. General exclusion criteria encompassed individuals below the age of 18, contraindications to MRI, and the presence of major neurological, psychiatric, or medical comorbidities. Data collection and assessments occurred at the University Hospital Zurich, Switzerland, between October 2019 and March 2022. The imaging protocol entailed the acquisition of standard-of-care pre-interventional computed tomography (CT) or MRI images (performed within 24 h of symptom onset), followed by repeated post-interventional multimodal MRI measurements at 1 d (max. 2 d) and 90 d (± 14 d) post-stroke (i.e., posthospitalization). Neurological evaluations were conducted at 3 d and 90 d post-stroke using the NIHSS. All participants received standard-of-care post-stroke rehabilitation. Ethical approval for the study was obtained from the Research Ethics Committee of the Canton Zurich, Switzerland (Kantonale Ethikkommission Zürich, KEK-ZH-NR. 2019 − 00750). The study adhered to the 1964 Declaration of Helsinki, and written informed consent was obtained from participants before their involvement.

Image acquisition

Pre-interventional CT or MRI contrast-enhanced perfusion imaging was performed at least within 24 h of stroke-onset. CT perfusion was conducted following the standard-of-care protocol at University Hospital Zurich or at other acute stroke units.

IVIM, DWI, and contrast-enhanced perfusion MR imaging were acquired post-treatment. MRI acquisition was performed on a 3-Tesla Skyra MRI scanner with a 32-channel head matrix coil (Siemens Healthineers, Forchheim; Germany). The imaging protocol included 3D T1 magnetization-prepared rapid gradient echo (MPRAGE) imaging [repetition and echo time (TR/TE) = 2200/5.14 ms, flip angle 8°, slice thickness = 1 mm, voxel size = 1 × 1 × 1 mm³, field of view (FOV) = 230 mm]; diffusion-weighted imaging (DWI) [2D EPI sequence, TR/TE = 2500/75 ms, flip angle = 90°, slice thickness = 4.5 mm, voxel size = 1.5 × 1.5 × 4.5 mm³, b-values = 0 and 1000 s/mm²]; MR PWI [TR/TE = 2040/36 ms, flip angle of 90°, slice thickness = 4 mm, voxel size = 1.7 × 1.7 × 4.0 mm³, FOV = 220 mm]; and IVIM [2D EPI sequence, TR/TE = 2500/97 ms, slice thickness = 4 mm, voxel size = 2 × 2 × 4 mm³, FOV = 260 mm, diffusion directions = 12, bvalues = 0, 50, 150, 200, 500, 1000 s/mm² with 3 averages each]. Additional MRI sequences were acquired as part of the imaging protocol but were not included in the analysis for the present study.

Contrast-enhanced perfusion imaging utilized Dotarem Gadoteric acid (Gadoterate meglumine) from Guerbet, Villepinte, France, administered intravenously at a dosage of 0.2 ml/kg body weight at 5 ml/s. A saline flush of 25–30 ml was followed to ensure full circulation of the contrast agent.

Image processing

Conventional contrast-enhanced PWI maps, comprising CBF, cerebral blood volume (CBV), and time-to-maximum of the residue function (Tmax), were generated using Syngo.via Client 8.7 (Siemens Healthineers GmbH, Erlangen, Germany) for CT-PWI or Olea Sphere 3.0.28 (Olea Medical SA, La Ciotat, France) for MR-PWI. ADC maps were computed via Bayesian parameter estimation using Olea Sphere 3.0.28 using a mono-exponential fitting model based on bvalues of 0 and 1000 s/mm². IVIM parameters were estimated using a bi-exponential model including the b-values 0, 50, 150, 200, 500, and 1000 s/mm². The IVIM phenomenon can be described by the following formula:

where S(b) is the signal intensity at a given b-value and S0 the signal intensity at b = 0. IVIM parameter f reflects the perfusion fraction (fraction of MRI signal attributed to perfusion-related effects), D* the pseudo-diffusion coefficient (perfusion-related pseudo-diffusion), and D the diffusion coefficient (diffusion of water molecules in tissue). The IVIM model was fitted using a two-stage least squares approach as implemented in the Python package dipy¹². Initially, a linear least squares fit was performed on the high b-value data (b > 400) to estimate the diffusion coefficient D and the initial signal intensity S0 based on mono-exponential decay. A second linear fit was applied to the low b-value data (b < 200) to obtain an estimate of S0’, and the perfusion fraction f was calculated as f = 1 - S0’/S0. Non-linear least squares fitting was then used to determine D* and f. Finally, these initial estimates were refined using a non-linear least squares fit to optimize all parameters. Perfusion fraction f is considered the most reliable and interpretable IVIM parameter and was therefore used as the primary IVIM outcome measure⁶.

Brain extraction of CT and MRI images was performed using Brain Extraction Tool (BET) as implemented in FSL¹³. For CT images, the multi-step BET procedure outlined in¹⁴ was utilized to enhance accuracy. Subsequently, all DWI, PWI, and IVIM images, along with the derived region-of-interest (ROI) masks (details provided below), were registered to each patient’s T1 anatomical reference image using rigid body transformation implemented in FSL. All registered images were resampled to 1 mm isotropic resolution using trilinear interpolation or nearest-neighbor interpolation for ROI masks as implemented in Advanced Normalisation Tools¹⁵.

Segmentation of core and penumbra

The infarct core, penumbra region, and corresponding control regions were delineated as follows: [1] The ischemic core was identified by applying a threshold of ADC < 600 × 10^− 6 mm²/s¹⁶ on the post-treatment DWI scan (1d post-stroke). [2] The penumbra region was delineated on the pre-treatment PWI scan by identifying delayed contrast medium arrival (Tmax > 6 s), and the infarct core was subsequently subtracted¹⁷. It is important to note that the penumbra was segmented using pre-treatment PWI scans, but in this study, IVIM and PWI values were extracted from this region only after endovascular treatment. Consequently, these values represent the recanalized penumbra. The term “recanalized penumbra” is used throughout the manuscript, although five patients in this study did not receive endovascular treatment. [3] Control regions were defined as the contralateral core and penumbra regions, with the ROIs mirrored to the contralateral hemisphere. All ROI masks underwent meticulous quality control and manual adjustments, when necessary, utilizing segmentation tools available in Slicer Version 5.2.1. To mitigate potential interference from cerebrospinal fluid (CSF) artefacts within the ROIs, CSF areas were segmented within the patients’ T1 anatomical reference image using FSL FAST¹³, and derived CSF areas were removed from all ROIs.

Statistical analysis

For each ROI, the median for each parameter of interest (IVIM f, D*, D, and PWI CBF, CBV) was computed per patient across all-voxels on non-normalized parameter maps. Voxels with parameter estimates exceeding 0.3 for IVIM f and 0.14 for IVIM D* were excluded from the analysis, as per Federau et al.‘s criteria³, since they are likely associated with noise or CSF. Extreme outlier data points, specifically individual IVIM D* measurements from two patients, were excluded from the analysis. All other measurements from these patient were retained in the analysis. Extreme outlier data points were defined following the approach recommended by Jones¹⁸, based on the robust scale estimator proposed by Rousseeuw & Croux¹⁹. To explore the relationship between the clinical NIHSS score at 3 and 90 days poststroke and infarct core volume, IVIM f as well as PWI CBF and CBV within core and recanalized penumbra at 24 h post-stroke, non-parametric Spearman rank-order correlations were computed as the NIHSS reflects an ordinal scale. Additionally, partial Spearman correlations were computed to determine whether the correlation between IVIM f in the core and recanalized penumbra and NIHSS at 90 days post-stroke were independent of the infarct volume. Subsequently, to further evaluate whether IVIM f adds prognostic value beyond the well-known prognostic factors core volume and patient age, we compared two linear regression models. The baseline model included core volume and age as predictors of NIHSS at 90 days, while the extended model included additionally the IVIM f measure of the core or recanalized penumbra. We used an F-test to compare the baseline model with the extended model to determine if adding IVIM f significantly improved the explained variance of the NIHSS score at 90 days post-stroke.

To assess differences regarding parameters of interest (IVIM f, D*, D and PWI CBF, CBV) between the ischemic core, recanalized penumbra, and corresponding control tissue at 24 h post-stroke, paired t-tests were employed. P-values were corrected for multiple comparisons using Bonferroni correction. Effect sizes of these differences were evaluated by computing Cohens’ d for paired samples. All statistical tests were two-sided and a significance level of 5% (alpha level) was applied. Statistical analysis and data visualization were conducted using R version 4.2.2.

Results

Characteristics of the study population

Imaging and clinical data were available for n = 37 acute ischemic stroke patients (Fig. 1). In total n = 13 participants were excluded from analysis due to missing PWI or IVIM data. The final analysis included n = 24 ischemic stroke patients (female/male = 14/10, age [mean ± sd] = 67.1 ± 15.8 years; Table 1). The mean post-treatment core volume was 27.8 ± 27.8 mL and mean penumbra volume 62.9 ± 40.9 mL. The mean NIHSS at 3 days post-stroke was 7.3 ± 6.6 and at 90 days post-stroke 1.2 ± 1.2. The majority of patients analyzed in this study underwent acute stroke treatment (n = 1 thrombolysis alone, n = 10 thrombectomy alone, n = 8 both combined), and n = 5 patients received no acute treatment. Representative imaging data of two patients are displayed in Fig. 2.

Fig. 1.

STROBE flow chart. Number of participants included in the present study and reasons for exclusion. PWI = contrast-enhanced perfusion weighted imaging, IVIM = intravoxel incoherent motion imaging.

Table 1.

Characteristics of patients included in the analysis.

N patients total	24
N patients per visit (Pre-treatment / 1 d / 90 d post-stroke)	24 / 24 / 16
Sex (female/male)	14 / 10
Age (years)	67.1 ± 15.8
Core volume (mL)	27.8 ± 27.8
Penumbra volume (mL)	62.9 ± 40.9
N thrombolysis / thrombectomy / both combined / none	1 / 10 / 8 / 5
Reperfusion (complete [mTICI = 3] / no or partial)	7 / 17
NIHSS 3 d / 90 d post-stroke	7.3 ± 6.6 / 1.2 ± 1.2

Values reflect total count or mean ± standard deviation.

mTICI = modified treatment in cerebral infarction score

Fig. 2.

Representative images of patients included in this study. The penumbra was defined using pre-treatment contrast-enhanced perfusion imaging, with time-to-maximum of the residue function (Tmax) > 6 s (outlined in pink). The core was identified on the post-treatment apparent diffusion coefficient (ADC) image with ADC < 600 × 10^-6 mm²/s (outlined in light blue). Post-treatment IVIM parameters f, D*, and D are shown with the core and penumbra region outlines overlayed.

Relationship between post-treatment IVIM parameters and long-term clinical outcome

IVIM perfusion fraction f at 1 d post-stroke (i.e., post-endovascular treatment) within the infarct core as well as within the recanalized penumbra displayed a strong negative association with the NIHSS score at 90 days post-stroke (core: ρ[11]=-0.64, p = 0.020; recanalized penumbra: ρ[13]=-0.69, p = 0.005; Fig. 3). The infarct core volume was also correlated with NIHSS at 90 days post-stroke (ρ[13] = 0.74, p = 0.002). Partial correlation analysis, controlling for core volume, revealed a sustained significant correlation of IVIM f in the core but not in the recanalized penumbra with NIHSS (core: ρ[11]=-0.61, p = 0.034; recanalized penumbra: ρ[13]=-0.37 p = 0.24). Notably, the incorporation of IVIM f as an additional predictor into a linear regression model, which already included core volume and patient age as predictors, significantly increased the explained variance of NIHSS at 90 days post-stroke (core: explained variance 83.4% vs. 42.3%; F = 22.32, p = 0.001; recanalized penumbra: explained variance 70.1% vs. 42.3%; F = 8.37, p = 0.018). In contrast, post-treatment PWI CBF and CBV were not associated with long-term NIHSS score (CBF core: ρ[11]=-0.01, p = 0.99, CBF recanalized penumbra: ρ[13]=-0.07, p = 0.80; CBV core: ρ[11] = 0.17, p = 0.58, CBV recanalized penumbra: ρ[13] = 0.19, p = 0.49). None of the IVIM as well as PWI measures were significantly correlated to the NIHSS score at 3 days post-stroke (IVIM f core ρ[16]=-0.13, p = 0.62; IVIM f recanalized penumbra: ρ[18] = 0.09, p = 0.68; CBF core: ρ[14] = 0.31, p = 0.24; CBF recanalized penumbra: ρ[18] = 0.32, p = 0.17; CBV core: ρ[14] = 0.06, p = 0.83; CBV recanalized penumbra: ρ[18] = 0.28, p = 0.24).

Fig. 4.

Post-treatment intravoxel incoherent motion (IVIM) imaging in stroke infarct core, recanalized penumbra region, and their mirrored contralateral healthy control region. (a and b) Differences in IVIM parameters f and D between region of interests. Post-treatment IVIM f and D were significantly reduced in both the infarct core and recanalized penumbra compared to their respective control region with large effect sizes (Cohens’ d). IVIM f and D were also significantly reduced in the core when compared to the recanalized penumbra. (c) IVIM D* exhibited a trend toward higher values in both the core and recanalized penumbra compared to their respective control regions, and also higher values in the core compared to the recanalized penumbra. However, these effects did not remain significant after correction for multiple comparisons.

Fig. 3.

Spearman rank-order correlations of post-treatment intravoxel incoherent motion (IVIM) perfusion fraction within the infarct core and recanalized penumbra region with long-term clinical outcome. (a and b) IVIM perfusion fraction f at 1 d post-stroke (i.e., post-treatment) displayed a strong correlation with clinical outcome at 90 d post-stroke measured by the National Institutes of Health Stroke Scale (NIHSS) in the stroke infarct core as well as in the recanalized penumbra region. ρ = Spearman rank-order correlation coefficient, p = p-value.

Post-treatment difference between core, recanalized penumbra, and healthy tissue

Post-treatment IVIM parameters f (perfusion fraction) and D (tissue diffusion) were significantly reduced with large effect sizes in the infarct core and recanalized penumbra compared to their corresponding contralateral (mirrored) control region (IVIM f core vs. control: t[17] = 7.65, p < 0.0001, Cohens’ d = 1.80; IVIM f recanalized penumbra vs. control : t[21] = 4.92, p < 0.001, Cohens’ d = 1.04; IVIM D core vs. control: t[17] = 10.68, p < 0.0001, Cohens’ d = 2.51; IVIM D recanalized penumbra vs. control: t[17] = 3.52, p = 0.006, Cohens’ d = 0.75). IVIM f and D were also significantly reduced in the core when compared to the recanalized penumbra (IVIM f core vs. recanalized penumbra: t[17] = 7.61, p < 0.0001, Cohens’ d = 1.79; IVIM D core vs. recanalized penumbra: t[17] = 13.28, p < 0.0001, Cohens’ d = 3.12). IVIM D* showed a trend toward higher values in the core and recanalized penumbra compared to their respective control regions, but these effects did not survive correction for multiple comparisons (IVIM D* core vs. control: t[15] = 2.41, p = 0.088, Cohens’ d = 0.60; IVIM D* recanalized penumbra vs. control : t[20] = 2.47, p = 0.068, Cohens’ d = 0.54; IVIM D* core vs. recanalized penumbra : t[5] = 2.10, p = 0.16). Results are displayed in Fig. 4. All p-values here are reported after the Bonferroni correction was applied. Of note, the penumbra region was segmented based on pre-treatment PWI (Tmax > 6 s). Since the values used in the analysis were extracted from post-treatment IVIM and PWI, they generally reflect the recanalized penumbra, even though five patients did not receive endovascular treatment.

In contrast, post-treatment PWI CBF was significantly increased with large effect size in the infarct core compared to the control region and recanalized penumbra (CBF core vs. control: t[15] = 3.32, p = 0.012, Cohens’ d = 0.83; CBF core vs. recanalized penumbra: t[15] = 4.66, p < 0.001, Cohens’ d = 1.16), while CBF in the recanalized penumbra was not different from the contralateral (mirrored) penumbra area (t[19] = 0.20, p = 1.00, Cohens’ d = 0.04). No significant differences between core, recanalized penumbra and contralateral control tissue were found for PWI CBV (CBV core vs. control t[15] = 2.28, p = 0.11, Cohens’ d = 0.57; CBV recanalized penumbra vs. control t[19] = 1.74, p = 0.30, Cohens’ d = 0.38; CBV core vs. recanalized penumbra t[15] = 1.59, p = 0.39, Cohens’ d = 0.40).

Discussion

The present exploratory study addressed two key questions of IVIM imaging in ischemic stroke: (i) How do IVIM measures within the ischemic core and recanalized penumbra post-treatment relate to long-term clinical outcomes? (ii) Are IVIM parameters different between the core, recanalized penumbra, and healthy contralateral control tissue after treatment?

• (i)A strong correlation was found between post-treatment IVIM perfusion fraction f and NIHSS scores at 90 days post-stroke, particularly notable within the infarct core and even more pronounced in the recanalized penumbra region. This correlation was absent when considering contrast-enhanced PWI parameters. Noteworthy, including IVIM perfusion fraction f significantly increased the explained variance of NIHSS scores compared to established prognostic factors like core volume and age.
• (ii)Post-treatment IVIM parameters f and D were reduced in the infarct core and recanalized penumbra compared to their contralateral control region, with the core also showing a significant decrease compared to the recanalized penumbra. Conversely, IVIM D* tended to be higher in both the core and recanalized penumbra compared to the control regions. Of note, among conventional PWI measures, only CBF showed increased values in the core, while no significant differences were observed between ROIs for other conventional PWI parameters.

Intriguingly, the strong association between reduced IVIM perfusion fraction f and poor long-term clinical outcome underlines the importance of successful penumbral reperfusion for brain tissue recovery and functional rehabilitation. This finding aligns with prior research, consistently indicating that the failure of penumbral reperfusion is linked to more adverse long-term outcomes¹⁰. Moreover, the high correlation between the IVIM perfusion fraction f in the core and long-term NIHSS score emphasizes the additional critical relevance of microvascular perfusion within the core for long-term stroke-associated impairments, although the core generally represents non-salvageable necrotic tissue in most cases. Although only IVIM f in the core was independently associated with long-term NIHSS after accounting for core volume, the significantly greater explained variance of NIHSS when including IVIM f – compared to established prognostic factors such as core volume and patient age – emphasizes the substantial prognostic value of IVIM measurements in acute stroke imaging. The prognostic value of IVIM f appears to be particularly related to long-term stroke outcomes, as it did not correlate with NIHSS at 3 days post-stroke.

Our findings on the relationship between IVIM and long-term clinical outcome diverge from the study conducted by Chen et al. which did not observe a significant association⁵. However, in their study, they assessed only the infarct core and did not include the penumbra, which is of utmost significance in acute stroke imaging. Moreover, it is important to note that we used the NIHSS as our primary outcome measure, which is more sensitive to subtle changes in stroke recovery compared to the mRS applied in their study¹¹. Accordingly, the mRS could have lacked the granularity needed to capture changes in functional impairments associated with IVIM measures. Moreover, it is essential to consider differences in patient populations between the two studies. Chen et al. excluded patients who underwent endovascular treatment⁵, whereas the majority of patients in our cohort received such treatment. These variations in patient characteristics may contribute to the divergent findings between the studies.

Of note, in contrast to IVIM measures pointing to decreased microvascular perfusion within the infarct core after recanalization, PWI CBF indicated heightened blood flow, a common finding after endovascular treatment. This heightened blood flow may be related to increased blood-brain barrier permeability and a breakdown of vascular autoregulation, both of which have been linked to CBF hyperperfusion after recanalization²⁰. In the recanalized penumbra, IVIM parameters and PWI CBF also yielded opposing results, with IVIM f suggesting decreased microvascular perfusion while CBF did not indicate altered blood flow after recanalization. These contrasting outcomes emphasize the capability of IVIM to offer valuable insights into microvascular properties within the stroke lesion after recanalization that are not evident in conventional PWI measures. A prior study by Zhu et al. had already highlighted the reliability of IVIM in characterizing stroke-related perfusion abnormalities before treatment, suggesting that IVIM alone might provide the required information for identifying critical regions essential for the triage of stroke patients⁹. It is worth noting that they reported reduced microvascular perfusion in the penumbra before treatment, consistent with our findings post-treatment, potentially reflecting that recanalization in our study was not successful in all cases. Thus, our research findings confirm the potential of IVIM in evaluating tissue characteristics within the area identified as the penumbra before treatment. Interestingly, IVIM D (equivalent to ADC and thus related to microstructural changes rather than microvascular perfusion) was reduced in the recanalized penumbra, consistent with previous studies assessing ADC²¹. This reduction indicates restricted diffusion, suggesting that some degree of cellular swelling persists in the recanalized penumbra after endovascular treatment. IVIM D* showed a trend toward higher values in both the core and recanalized penumbra compared to their respective control regions. However, accurately estimating D* is challenging, especially when IVIM D is small, as in the core, which increases sensitivity to noise. Previous studies have reported implausible increases in D* within the stroke core in some cases, likely due to these estimation difficulties²².Therefore, IVIM D* values, particularly in the infarct core, should be interpreted with caution.

Taken together, our findings underline the utility of IVIM in assessing microvascular perfusion within essential stroke-affected regions. Although our study is limited by the relatively small patient sample size, the remarkably strong correlation with clinical outcomes and the significant differences between core, penumbra and contralateral control region with high effect sizes add confidence to our findings. Nevertheless, it is important to note that the patients included in our analysis had mild degrees of disability (NIHSS < 5) at the 90-day follow-up, which may limit the generalizability of our results to patients with more severe long-term impairments. Of note, accurately separating diffusion and perfusion effects in IVIM measurements is challenging, as varying b-value distributions and noise can cause parameter instability⁴. In clinical contexts like stroke, microstructural changes such as axonal beading can significantly affect the diffusion signal²³. Consequently, it remains unclear to what extent IVIM parameters reflect microvascular perfusion versus microstructural alterations². Interpretations of IVIM effects as changes in microvascular perfusion should be approached with caution. Moreover, the available imaging data did not allow for the evaluation of diffusion lesion reversal (as ADC images were only available post-treatment for the majority of patients), and the infarct core derived from ADC might not accurately reflect the “true” core in some cases.

In summary, our study emphasizes the potential of post-interventional IVIM characteristics in both the infarct core and recanalized penumbra as prognostic indicators for long-term clinical outcomes in ischemic stroke patients. Additionally, IVIM measures may offer valuable information for characterizing microvascular perfusion in the core and recanalized penumbral tissue after endovascular treatment at 1 d post-stroke.

Author contributions

J.Z. analysed data. J.Z., L.M., and T.S. wrote the main manuscript. A.R.L and S.W acquired funding, A.R.L, S.W., M.P, L.M., T.S., J.F., Z.K., and C.S. planned study. B.R, B.N., D.G., M.B., J.B., B.N., M.S., V.S., J.F., and M.P. conducted study. All authors reviewed the manuscript and provided valuable inputs.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the UZH Clinical Research Priority Program (CRPP) Stroke.

Data availability

The anonymized data used for analysis (extracted median values for all parameters of interest within each region of interest, clinical scores, and demographics) as well as the code used for the complete analysis, will be made available upon publication on https://github.com/jziNeuro/IVIM_stroke_publication.

Declarations

Competing interests

The authors declare no competing interests.