•An efficient early fusion structure is designed to fuse the features of native CTP, CBF, CBV, MTT and Tmax for accurate ischemic stroke prediction.•The attention mechanism is firstly employed into ...multi-scale convolutional extractor to capture the context information.•Model was evaluated on a real clinical dataset which was collected from multiple stroke centers.•The dataset uses two kinds of follow-up images: Non-contrast CT and MRI DWI.
Display omitted
Background: Acute ischemic stroke is one of the leading death causes. Delineating stoke infarct core in medical images plays a critical role in optimal stroke treatment selection. However, accurate estimation of infarct core still remains challenging because of 1) the large shape and location variation of infarct cores; 2) the complex relationships between perfusion parameters and final tissue outcome. Methods: We develop an encoder-decoder based semantic model, i.e., Ischemic Stroke Prediction Network (ISP-Net), to predict infarct core after thrombolysis treatment on CT perfusion (CTP) maps. Features of native CTP, CBF (Cerebral Blood Flow), CBV (Cerebral Blood Volume), MTT (Mean Transit Time), Tmax are generated and fused with five-path convolutions for comprehensive analysis. A multi-scale atrous convolution (MSAC) block is firstly put forward as the enriched high-level feature extractor in ISP-Net to improve prediction accuracy. A retrospective dataset which is collected from multiple stroke centers is used to evaluate the performance of ISP-Net. The gold standard infarct cores are delineated on the follow-up scans, i.e., non-contrast CT (NCCT) or MRI diffusion-weighted image (DWI). Results: In clinical dataset cross-validation, we achieve mean Dice Similarity Coefficient (DSC) of 0.801, precision of 81.3%, sensitivity of 79.5%, specificity of 99.5%, Area Under Curve (AUC) of 0.721. Our approach yields better outcomes than several advanced deep learning methods, i.e., Deeplab V3, U-Net++, CE-Net, X-Net and Non-local U-Net, demonstrating the promising performance in infarct core prediction. No significant difference of the prediction error is shown for the patients with follow-up NCCT and follow-up DWI (P >0.05). Conclusion: This study provides an approach for fast and accurate stroke infarct core estimation. We anticipate the prediction results of ISP-Net could offer assistance to the physicians in the thrombolysis or thrombectomy therapy selection.
Neuroimaging provides detailed information regarding the hemodynamic, metabolic and cellular parameters of cerebral ischemia (CI). Although providing just a snapshot in time, it assists in ...delineating the boundaries and extent of this continually evolving process, from the irreversibly damaged infarct core to the penumbral tissue, where salvage via reperfusion has been the focus of acute stroke therapies. Beyond the extent of the ischemic lesion, neuroimaging elucidates the topography and underlying mechanism of CI. Finally, based on the pathophysiological information, neuroimaging assists in the selection of optimal therapeutic strategies, from hyperacute to chronic phases of CI. Here we review different neuroimaging techniques by which the pathophysiology of cerebral ischemia can be delineated.
This article is part of the Special Issue entitled ‘Cerebral Ischemia’.
•CI is the result of a constant interaction of low CBF with duration and severity of CI.•Cytotoxic edema (DWI lesion with ADC≤ 620 μm2/s) is used to outline the ischemic core.•The early ischemic changes manifest as hypoattenuation on CT due to ionic edema.•Tmax ≥ 6sec on perfusion scans is associated with the least penumbral overestimation.•Collateral imaging assists in delineating the pathophysiology, mechanism, and fate of CI.
The size and location of infarct and penumbra are key to decision-making for acute ischemic stroke (AIS) management. CT perfusion (CTP) software estimate infarct and penumbra volume using ...contralateral hemisphere relative thresholding. This approach is not robust and widely contested by the scientific community. In this study, we investigate the use of deep learning-based algorithms to efficiently locate infarct and penumbra tissue on CTP hemodynamic maps.
CTP scans were retrospectively collected for 60 and 59 patients in the infarct only and infarct + penumbra substudies respectively. Commercial CTP software was used to generate cerebral blood flow, cerebral blood volume, mean transit time, time to peak, and delay time maps. U-Net-shaped architectures were trained to segment infarct or infarct + penumbra. Test-time-augmentation, ensembling, and watershed segmentation were used as postprocessing techniques. Segmentation performance was evaluated using Dice coefficients (DC) and mean absolute volume errors (MAVE).
The algorithm segmented infarct tissue resulted in DC of
(0.63, 0.65), and MAVE of
(4.5, 5.32) mL. In comparison, the commercial software predicted infarct with a DC of
(0.26, 0.36) and MAVE of
(7.12, 12.42) mL. The algorithm was able to segment infarct + penumbra with a DC of
(0.6, 0.63), and MAVE of
(5.91, 7.11) mL. In comparison, the commercial software predicted infarct + penumbra with a DC of
(0.25, 0.35) and MAVE of
(7.25, 11.11) mL.
Use of deep learning algorithms to assess severity of AIS in terms of infarct and penumbra volume is precise and outperforms current relative thresholding methods. Such an algorithm would enhance the reliability of CTP in guiding treatment decisions.
Neurological impairment is associated with collateral status in acute ischaemic stroke (AIS). We aimed to validate the association between admission National Institutes of Health Stroke Scale ...(aNIHSS) score and infarct core volume (ICV) and target infarct core/penumbra volume mismatch (TMM) on CT perfusion (CTP) in AIS patients.
Patients with acute middle cerebral artery or internal carotid artery occlusion from 2011 to 2020 were included. All patients underwent pretreatment CTP at admission. ICV and TMM were analyzed with MIStar software on CTP maps. aNIHSS scores and clinical characteristics of patients were obtained from our prospectively recorded stroke database.
We recruited 182 patients with a median age of 69.5 years; 85 (63.7%) were male, and the median aNIHSS score was 14. Of those, 149 (81.8%) had an ICV < 70 mL, and 139 (76.3%) had TMM. Lower aNIHSS was associated with an ICV < 70 mL, with an area under the curve (AUC) of 0.74, and TMM with an AUC of 0.76. Among all 15 items of the aNIHSS, the gaze score was the only item independently associated with an ICV < 70 mL (adjusted odds ratio OR = 0.42, 95% confidence interval CI: 0.22-0.79, p = 0.008) and TMM (adjusted OR = 0.5, 95% CI: 0.28-0.9, p = 0.021). One or both aNIHSS ≤ 16 and gaze score = 0 predicted TMM with a sensitivity of 0.79 and a specificity of 0.62.
aNIHSS may be a useful tool to predict an ICV < 70 mL and TMM on CTP in AIS patients.
Although computed tomography perfusion (CTP) is used to select and guide decision-making processes in patients with acute ischemic stroke, there is no clear standardization of the optimal threshold ...to predict ischemic core volume accurately. The infarct core volume with a relative cerebral blood flow(rCBF) threshold of < 30% is commonly used. We aimed to assess the volumetric agreement of the infarct core volume with different CTP parameters and thresholds using CTP software (RAPID, VITREA) and the infarct volume on diffusion-weighted imaging (DWI), with a short interval time (within 60 min) between CTP and follow-up DWI.
This retrospective study included 42 acute ischemic stroke patients with occlusion of the large artery in the anterior circulation between April 2017-November 2020. RAPID identified infarct core as tissue rCBF < 20-38%. VITREA defined the infarct core as cerebral blood volume (CBV) < 26-56%. Olea Sphere was used to measure infarct core volume on DWI. The CTP-infarct core volume with different thresholds of perfusion parameters (CBF threshold vs CBV threshold) were compared with DWI-infarct core volumes.
The median time between CTP and DWI was 37.5min. The commonly used threshold of CBV< 41% (4.3 mL) resulted in lower median infarct core volume difference compared to the commonly used thresholds of rCBF < 30% (8.2mL). On the other hand, the optimal thresholds of CBV < 26% (-1.0mL; 95% CI, -53.9 to 58.1 mL; 0.945) resulted in the lowest median infarct core volume difference, narrowest limits of agreement, and largest interclass correlation coefficient compared with the optimal thresholds of rCBF < 38% (4.9 mL; 95% CI, -36.4 to 62.9 mL; 0.939).
Our study found that the both optimal and commonly used thresholds of CBV provided a more accurate prediction of the infarct core volume in patients with AIS than rCBF.
Large clinical trials have helped establish the benefit of endovascular treatment (EVT) in patients with acute ischemic stroke with large vessel occlusion and small infarct core volume as determined ...by scores ≥6 on the Alberta Stroke Program Early CT Scores. Several small studies have suggested that patients with large infarct core volume (LICV) may also benefit from EVT. Currently, at least 6 randomized clinical trials are examining the benefit of extending EVT to this population of patients with acute ischemic stroke. These trials were independently conceived and have significant differences in their inclusion criteria. Understanding these inclusion criteria and other differences in trial design is pivotal for the field to interpret the upcoming results of these trials. In this review, the designs of the 6 trials are summarized and compared. Specific differences are described, including (1) the rationale for EVT treatment in patients with LICV, (2) how to define LICV and the imaging modality used to identify LICV, (3) inclusion of an Alberta Stroke Program Early CT Score 0 to 5 versus 3 to 5, (4) use of the mismatch between blood flow and the size of infarct as an inclusion criterion, and (5) inclusion of early window and/or late window patients. The potential impact of these trial results on current guidelines for acute ischemic stroke is discussed. Differences in trial design as well as inclusion and exclusion criteria may influence trial outcomes. The implications of these trial results will likely be enhanced by a pooled analysis.
In the setting of an extended time window for endovascular treatment (EVT) for acute stroke patients, computed tomography perfusion (CTP) has become a major tool in patient selection. However, there ...are some data suggesting that the initial ischemic core may be overestimated by CTP depending on stroke onset time. This study aims to evaluate possible predictors of overestimation of infarct core by CTP.
We studied all consecutive stroke patients undergoing EVT during 1 year who underwent CTP at admission and had a successful recanalization. Admission infarct core was measured on cerebral blood volume maps generated using the Intellispace Portal (Philips Healthcare, Best, the Netherlands) and final infarct was measured on noncontrast follow-up computed tomography at 24 hours. We defined overestimation of the infarct core as initial core minus final infarct >10 mL.
Out of 107 patients undergoing EVT in the study period, 60 were anterior circulation and had CTP done at our institute, and of them 31 were compatible with the inclusion criteria (known time of onset, no hemorrhagic conversion, and good recanalization). Median National Institute of Health Stroke Scale on admission was 13. Median time from symptoms to CTP was 148 minutes. Seventeen patients were found to have overestimation of the infarct core. Logistic regression analyses showed time from symptom onset to CTP to be inversely related to overestimation with a cutoff of 170 minutes (sensitivity 94% and specificity 43%).
Over estimation of the infarct core by CTP in patients undergoing EVT is time dependent and so CTP results among early arrivers should be interpreted cautiously.
Advanced stroke imaging has generated much excitement for the early diagnosis of acute ischemic stroke (AIS) and facilitation of intervention. However, its therapeutic impact has not matched its ...diagnostic utility; most notably, lacking significant contributions to recent major AIS clinical trials. It is time to reexamine the fundamental hypotheses from the enormous body of imaging research on which clinical practices are based and reassess the current standard clinical and imaging strategies, or golden rules, established over decades for AIS. In this article, we will investigate a possible new window of opportunity in managing AIS through a better understanding of the following: first, the potential limitations of the golden rules; second, the significance of imaging-based parenchymal hypoperfusion (i.e., lower-than-normal relative cerebral blood flow rCBF may not be indicative of ischemia); third, the other critical factors (e.g., rCBF, collateral circulation, variable therapeutic window, chronicity of occlusion) that reflect more individual ischemic injury for optimal treatment selection; and, fourth, the need for penumbra validation in successfully reperfused patients (not in untreated patients).
Individual variations in the therapeutic window, ischemic injury (rCBF), and chronicity of vascular lesion development have not been comprehensively incorporated in the standard algorithms used to manage AIS. The current established imaging parameters have not been consistently validated with successfully reperfused patients and rCBF to quantitatively distinguish between oligemia and ischemia and between penumbra and infarct core within ischemic tissue. A novel paradigm incorporating rCBF values or indirectly incorporating relative rCBF values with higher statistically powered imaging studies to more reliably assess the severity of ischemic injury and differentiate reversibility from viability within the area of imaging-based parenchymal hypoperfusion may provide a more personalized approach to treatment, including no treatment of infarction core, to further enhance outcomes.
To determine whether perfusion computed tomography (PCT) adds value to noncontrast head CT (NCT), CT angiogram (CTA), and clinical assessment in patients suspected of acute ischemic stroke.
We ...retrospectively reviewed 165 patients with acute ischemic stroke. PCT was used to calculate the volumes of infarct core and ischemic penumbra on admission. Other imaging data included Alberta Score Program Early CT Score, site of occlusion, and collateral flow. Clinical data included age, time, National Institutes of Health Stroke Scale at baseline, treatment type, and modified Rankin score (mRS) at 90 days. Recanalization status was assessed on follow-up imaging. In a first multivariate regression analysis, we assessed whether volumes of PCT penumbra and infarct core could be predicted from clinical variables, NCT, or CTA, or whether they represented independent information. In a second multivariate regression analysis, we used mRS at 90 days as outcome and determined which variables predicted it best.
Of 165 patients identified, 76 had a mRS score of 0 to 2 at 90 days, 89 had a mRS score >2. PCT infarct could be predicted by clinical data, NCT, CTA, and combinations of this data (P<0.05). PCT penumbra could not be predicted by clinical data, NCT, and CTA. All of the variables but NCT and CTA were significantly associated with 90-day mRS outcome. The single most important predictor was recanalization status (P<0.001). PCT penumbra volume (P=0.001) was also a predictor of clinical outcome, especially when considered in conjunction with recanalization through an interaction term (P<0.001).
PCT penumbra represents independent information, which cannot be predicted by clinical, NCT, and CTA data. PCT penumbra is an important determinant of clinical outcome and adds relevant clinical information compared with a stroke CT workup, including NCT and CTA.
The aim of this study was to investigate diffusion tensor (DT) imaging-derived properties of benign oligemia, true "at risk" penumbra (TP), and the infarct core (IC) during the first 3 hours of ...stroke onset.
The study was approved by the local animal care and use committee. DT imaging data were obtained from 14 rats after permanent middle cerebral artery occlusion (pMCAO) using a 7T magnetic resonance scanner (Bruker) in room air. Relative cerebral blood flow and apparent diffusion coefficient (ADC) maps were generated to define oligemia, TP, IC, and normal tissue (NT) every 30 minutes up to 3 hours. Relative fractional anisotropy (rFA), pure anisotropy (rq), diffusion magnitude (rL), ADC (rADC), axial diffusivity (rAD), and radial diffusivity (rRD) values were derived by comparison with the contralateral normal brain.
The mean volume of oligemia was 24.7 ± 14.1 mm
, that of TP was 81.3 ± 62.6 mm
, and that of IC was 123.0 ± 85.2 mm
at 30 minutes after pMCAO. rFA showed an initial paradoxical 10% increase in IC and TP, and declined afterward. The rq, rL, rADC, rAD, and rRD showed an initial discrepant decrease in IC (from -24% to -36%) as compared with TP (from -7% to -13%). Significant differences (
< 0.05) in metrics, except rFA, were found between tissue subtypes in the first 2.5 hours. The rq demonstrated the best overall performance in discriminating TP from IC (accuracy = 92.6%, area under curve = 0.93) and the optimal cutoff value was -33.90%. The metric values for oligemia and NT remained similar at all time points.
Benign oligemia is small and remains microstructurally normal under pMCAO. TP and IC show a distinct evolution of DT-derived properties within the first 3 hours of stroke onset, and are thus potentially useful in predicting the fate of ischemic brain.