Signal variance-based collateral index in DSC perfusion: A novel method to assess leptomeningeal collateralization in acute ischaemic stroke

Abstract

As a determinant of the progression rate of the ischaemic process in acute large-vessel stroke, the degree of collateralization is a strong predictor of the clinical outcome after reperfusion therapy and may influence clinical decision-making. Therefore, the assessment of leptomeningeal collateralization is of major importance. The purpose of this study was to develop and evaluate a quantitative and observer-independent method for assessing leptomeningeal collateralization in acute large-vessel stroke based on signal variance characteristics in T2*-weighted dynamic susceptibility contrast (DSC) perfusion-weighted MR imaging (PWI). Voxels representing leptomeningeal collateral vessels were extracted according to the magnitude of signal variance in the PWI raw data time series in 55 patients with proximal large-artery occlusion and an intra-individual collateral vessel index (CVI_PWI) was calculated. CVI_PWI correlated significantly with the initial ischaemic core volume (rho = -0.459, p = 0.0001) and the PWI/DWI mismatch ratio (rho = 0.494, p = 0.0001) as an indicator of the amount of salvageable tissue. Furthermore, CVI_PWI was significantly negatively correlated with NIHSS and mRS at discharge (rho = -0.341, p = 0.015 and rho = -0.305, p = 0.023). In multivariate logistic regression, CVI_PWI was an independent predictor of favourable functional outcome (mRS 0-2) (OR = 16.39, 95% CI 1.42-188.7, p = 0.025). CVI_PWI provides useful rater-independent information on the leptomeningeal collateral supply in acute stroke.

Keywords: Acute ischaemic stroke; collaterals; functional outcome; magnetic resonance imaging; perfusion-weighted imaging.

Figure 1.

Illustration of calculation of signal variance in PWI. The top left image exemplarily shows the first volume of the PWI time series. Coefficient of variation (CV) maps were calculated by dividing the standard deviation of every voxel by its mean across time. Feeding vessels of the pial compartment are characterized by a large standard deviation and a relatively low mean signal intensity across time. CV maps (lower left image) were thresholded below the upper 50% of the robust range (lower middle image) and corrected for voxels representing the ventricles and outer CSF spaces. Colour bars represent robust intensity ranges. Standard deviations and mean values of signal intensity across time for the PWI raw data are given in arbitrary units, CV for each voxel is given as dimensionless number. Note the increased image contrast between larger vascular structures and brain parenchyma on the CV map (lower left image) compared to the standard deviation map (upper middle image). The corrected CV map is shown as a binary mask (lower right image). PWI: perfusion-weighted imaging; σ: standard deviation; µ: mean value; a.u.: arbitrary units; CV: coefficient of variance.

Figure 2.

(a) Ischaemic core (violet) and different tissue compartments based on the severity of the TTP-delay overlaid on the first diffusion-weighted image (b = 0 s/mm²) of a representative patient (same patient as in Figure 1). The infarct core was defined on ADC maps applying a threshold of < 600 × 10⁻⁶ mm²/s. The scale bar shows the colour coding for different severities of TTP-delay defined as follows. Green: benign oligemia, yellow: tissue at risk, red: severely hypoperfused tissue. (b) Cortex mask used for measurement of the cortical infarct volume. (c) mean-signal time courses for each of the regions with TTP-delay shown in (a) and mean signal time course for adjacent pial collateral vessels. Major characteristics of the signal-time courses including standard deviation, mean signal intensity across time and coefficient of variation are provided in text form. It becomes clear that collateral vessels can be extracted from PWI raw data based on the magnitude of signal variance. TTP: time-to-peak; s: seconds; a.u.: arbitrary units; σ: standard deviation; µ: mean value; CV: coefficient of variation.

Figure 3.

Images of two representative patients with good (a) and poor collaterals (b) with M1 occlusion and similar time from symptom onset to MRI (3.4 and 3.2 h). Both patients received intravenous thrombolysis but did not undergo endovascular treatment. (a) Female patient (73 y) with left M1 occlusion of unknown source. ADC map at admission shows small scattered infarcts in the left MCA territory (hypointense areas) involving the basal ganglia, parts of the insular ribbon, frontal operculum and the frontoparietal cortex. Some of these infarcts are not clearly visible in the b = 1000 s/mm² diffusion-weighted image. FLAIR at admission shows hyperintense vessels over the left lateral convexity without clear demarcation of infarcts. Baseline ischaemic core defined by automatic thresholding and areas of TTP-delay are overlaid on the ADC map (lower left images). In this patient, no severe hypoperfusion (TTP-delay ≥ 9.5 s) is present, leading to a hypoperfusion intensity ratio of 0. Areas with a TTP-delay of ≥ 4.5 s outside the ischaemic core had progressed to infarction as visible on DWI and FLAIR at follow-up. NIHSS at discharge was 2 (improvement of four points compared to admission) and mRS was 1. (b) Female patient (81 y) with right M1 occlusion of cardioembolic origine. ADC and DWI maps at admission show extensive infarction involving large parts of the right MCA territory. FLAIR at admission was waived in this patient due to severe clinical conditions. Baseline ischaemic core defined by automatic thresholding and areas of TTP-delay are overlaid on the ADC map (lower left images). Areas with TTP-delay outside the ischaemic core showed progression to infarction at follow-up. NIHSS at discharge was 16 (deterioration of two points compared to admission) and mRS was 5. CVI_PWI: PWI-based collateral vessel index; h: hours; ADC: apparent diffusion coefficient; DWI: diffusion-weighted imaging; FLAIR: fluid-attenuated inversion recovery; cm³: cubic centimeters; HIR: hypoperfusion intensity ratio; TTP: time-to-peak; s: seconds; NCCT: non-contrast computed tomography.