Raw Perfusion Research Articles

Abstract TP132: Optimal CT Perfusion Post-Processing Method in Posterior Circulation Stroke

Objective: CT-perfusion (CTP) has revolutionised stroke care by improving diagnostic accuracy and expanding eligibility to acute therapies. Software packages utilise various mathematical techniques to transform raw perfusion data into measures of infarct core and ischaemic penumbra. These techniques have been derived from studies of anterior-circulation stroke and not yet validated in posterior circulation stroke (PCS). We examined the optimal CTP thresholds for acute PCS. Methods: Data were analysed from 331-patients diagnosed with a PCS enrolled in the International-stroke-perfusion-registry (INSPIRE). Twenty-seven-patients with baseline multimodal-CT with occlusion of a large posterior-circulation (PC) artery and follow up diffusion-weighted-MRI at 24-48 hours were included. CTP parametric maps were generated using five different post-processing methods. Receiver-operating-curve analysis was used to determine the optimal perfusion parameter and thresholds. Results: Partial deconvolution was the optimal post-processing method for characterisation of ischaemic penumbra and infarct core. Mean transit time (MTT) at a threshold of >165% and >195% most accurately defined ischaemic core (AUC=0.73) and penumbra (AUC=0.78) respectively. Post-processing technique influenced accuracy of core (AUC range=0.54-0.73) and penumbra (AUC range=0.72-0.79) estimates. MTT was consistently the most accurate parameter at distinguishing core and penumbra across all post-processing methods. Conclusion: CTP has significant diagnostic utility in PCS. Accuracy of CTP varies considerably by post-processing method. The underlying post-processing technique employed by individual software packages should be considered when interpreting the accuracy of ischemic core and penumbra estimates.

Predicting Infarct Core From Computed Tomography Perfusion in Acute Ischemia With Machine Learning: Lessons From the ISLES Challenge.

The ISLES challenge (Ischemic Stroke Lesion Segmentation) enables globally diverse teams to compete to develop advanced tools for stroke lesion analysis with machine learning. Detection of irreversibly damaged tissue on computed tomography perfusion (CTP) is often necessary to determine eligibility for late-time-window thrombectomy. Therefore, the aim of ISLES-2018 was to segment infarcted tissue on CTP based on diffusion-weighted imaging as a reference standard. The data, from 4 centers, consisted of 103 cases of acute anterior circulation large artery occlusion stroke who underwent diffusion-weighted imaging rapidly after CTP. Diffusion-weighted imaging lesion segmentation was performed manually and acted as a reference standard. The data were separated into 63 cases for training and 40 for testing, upon which quality metrics (dice score coefficient, Hausdorff distance, absolute lesion volume difference, etc) were computed to rank methods based on their overall performance. Twenty-four different teams participated in the challenge. Median time to CTP was 185 minutes (interquartile range, 180-238), the time between CTP and magnetic resonance imaging was 36 minutes (interquartile range, 25-79), and the median infarct lesion size was 15.2 mL (interquartile range, 5.7-45). The best performance for Dice score coefficient and absolute volume difference were 0.51 and 10.1 mL, respectively, from different teams. Based on the ranking criteria, the top team's algorithm demonstrated for average Dice score coefficient and average absolute volume difference 0.51 and 10.2 mL, respectively, outperforming the conventional threshold-based method (dice score coefficient, 0.3; volume difference, 15.3). Diverse algorithms were used, almost all based on deep learning, with top-ranked approaches making use of the raw perfusion data as well as methods to synthetically generate complementary information to boost prediction performance. Machine learning methods may predict infarcted tissue from CTP with improved accuracy compared with threshold-based methods used in clinical routine. This dataset will remain public and can be used to test improvement in algorithms over time.

Prediction of Stroke Lesion at 90-Day Follow-Up by Fusing Raw DSC-MRI With Parametric Maps Using Deep Learning

Stroke is the second most common cause of death in developed countries. Rapid clinical assessment and intervention have a major impact on preventing infarct growth and consequently on patients’ quality of life. Clinical interventions aim to restore perfusion deficits via pharmaceutical or mechanical intervention. Regardless of which reperfusion procedure is used, clinicians need to consider the risks and benefits based on multi-modal neuroimaging studies, such as MRI scans, as well as their own clinical experience. This intricate decision-making process would benefit from an automatic prediction of the final infarct, which would provide a estimation of tissue that will probably infarct. This paper introduces a deep learning method to automatically predict ischemic stroke tissue outcome. The authors propose an end-to-end deep learning architecture that combines information from perfusion dynamic susceptibility MRI, alongside perfusion and diffusion parametric maps. We aim to automatically extract features from the raw perfusion DSC-MRI to further complement the information gleaned from standard parametric maps, and to overcome the loss of information that can occur during perfusion postprocessing. Combining both data types in a single architecture, with dedicated paths, we achieve competitive results when predicting the final stroke infarct core lesion in the publicly available ISLES 2017 dataset.

Computed Tomography Perfusion Imaging Quality Affected by Different Input Arteries in Patients of Internal Carotid Artery Stenosis.

BackgroundThe aim of this study was to explore the influence of different input arteries on the parameters of computed tomography (CT) perfusion imaging for patients with different degree of stenosis of internal carotid artery (ICA).Material/MethodsForty patients were enrolled in the present study and divided into mild, moderate, severe stenosis and occlusion groups respectively with each 10 patients in each group. In reconstruction of cerebral CT perfusion (CTP) images, each raw perfusion image was reconstructed 3 times based on different reference input artery, including bilateral middle cerebral artery (MCA) and basilar arteries (BA). Region of interest (ROI) was drawn in the central territories of bilateral anterior cerebral artery, middle cerebral artery and posterior cerebral artery. And regional cerebral blood flow (rCBF) regional cerebral blood volume (rCBV), mean transit time (MTT), time to peak (TTP) and delay time (DT) were obtained from those ROI corresponding perfusion images.ResultsIn patients with mild and moderate ICA stenosis, there was no significant difference of perfusion parameters based on different input arteries (P>0.05). However, in severe ICA stenosis and occlusion CBF, MTT, and DT were significant different in affect side of the MCA group compared to the others (P<0.05).ConclusionsLarge intracranial artery can be selected as the input artery for patients with mild to moderate ICA stenosis, while for patients with severe stenosis and occlusion of ICA, the contra lateral middle cerebral artery or basilar artery would be better choice.

Neural Network-derived Perfusion Maps for the Assessment of Lesions in Patients with Acute Ischemic Stroke.

To perform a proof-of-concept study to investigate the clinical utility of perfusion maps derived from convolutional neural networks (CNNs) for the workup of patients with acute ischemic stroke presenting with a large vessel occlusion. Data on endovascularly treated patients with acute ischemic stroke (n = 151; median age, 68 years [interquartile range, 59-75 years]; 82 of 151 [54.3%] women) were retrospectively extracted from a single-center institutional prospective registry (between January 2011 and December 2015). Dynamic susceptibility perfusion imaging data were processed by applying a commercially available reference method and in parallel by a recently proposed CNN method to automatically infer time to maximum of the tissue residue function (Tmax) perfusion maps. The outputs were compared by using quantitative markers of tissue at risk derived from manual segmentations of perfusion lesions from two expert raters. Strong correlations of lesion volumes (Tmax > 4 seconds, > 6 seconds, and > 8 seconds; R = 0.865-0.914; P < .001) and good spatial overlap of respective lesion segmentations (Dice coefficients, 0.70-0.85) between the CNN method and reference output were observed. Eligibility for late-window reperfusion treatment was feasible with use of the CNN method, with complete interrater agreement for the CNN method (Cohen κ = 1; P < .001), although slight discrepancies compared with the reference-based output were observed (Cohen κ = 0.609-0.64; P < .001). The CNN method tended to underestimate smaller lesion volumes, leading to a disagreement between the CNN and reference method in five of 45 patients (9%). Compared with standard deconvolution-based processing of raw perfusion data, automatic CNN-derived Tmax perfusion maps can be applied to patients who have acute ischemic large vessel occlusion stroke, with similar clinical utility.© RSNA, 2019Supplemental material is available for this article.

Perilesional perfusion in chronic stroke-induced aphasia before and after behavioral treatment interventions

Perilesional perfusion in chronic stroke-induced aphasia before and after behavioral treatment interventions

Local spatio-temporal encoding of raw perfusion MRI for the prediction of final lesion in stroke.

We address the medical image analysis issue of predicting the final lesion in stroke from early perfusion magnetic resonance imaging. The classical processing approach for the dynamical perfusion images consists in a temporal deconvolution to improve the temporal signals associated with each voxel before performing prediction. We demonstrate here the value of exploiting directly the raw perfusion data by encoding the local environment of each voxel as a spatio-temporal texture, with an observation scale larger than the voxel. As a first illustration for this approach, the textures are characterized with local binary patterns and the classification is performed using a standard support vector machine (SVM). This simple machine learning classification scheme demonstrates results with 95% accuracy on average while working only raw perfusion data. We discuss the influence of the observation scale and evaluate the interest of using post-processed perfusion data with this approach.

Abstract 125: Vitamin D3-Induced Arterial Calcification Decreases Functional Recovery and Limb Perfusion in a Murine Model of Hindlimb Ischemia

Objectives: Clinical studies demonstrate associations between arterial calcification and adverse outcomes in patients with peripheral arterial disease. The causal relationship between calcification and ischemia, however, remains unclear. This study aims to evaluate the effects of arterial calcification on functional recovery and limb perfusion following acute limb ischemia in a murine model. Methods: Arterial calcification was induced in 64 Sprague-Dawley rats using Vitamin D 3 . A control group of 32 rats received vehicle only. After 2 weeks, to allow for calcification, unilateral hindlimb ischemia was induced by ligation and excision of the superficial femoral artery. At weekly intervals, functional and ischemia status were evaluated in a blinded manner and laser Doppler perfusion imaging was performed. Results: Calcium assay demonstrated greater arterial calcification in the superficial femoral arteries (0.1 vs. 17.0, P &lt; .01) of Vitamin D 3 injected animals. Rats injected with Vitamin D3 showed significantly worse Tarlov, Ischemia, Modified Ischemia, and Total Ischemia scores at each post-operative time point (Week 1: 16.6 vs. 14.5 P &lt;.01, Week 2: 17.1 vs. 14.5 P &lt;.01, Week 3 17.5 vs. 15.0 P &lt; .01). The control group regained normal functional status by week 4; however Vitamin D3-injected rats did not return to baseline by the conclusion of the study (17.9 vs. 16.6, P &lt; .01). Laser Doppler perfusion revealed decreased raw perfusion values pre-operatively (616 vs. 428, P &lt; .01) and at weekly intervals beginning at week 2 in both the ischemic and non-ischemic limbs among animals with calcified arteries as compared to controls (Table 1). Conclusions: Vitamin D3 induced calcification is associated with decreased perfusion and worse functional recovery in a rat model of hindlimb ischemia. Future research aimed at understanding the relationship between vitamin D 3 , arterial calcification, and lower extremity ischemia may help to improve outcomes in patients with PAD.

Decomposing cerebral blood flow MRI into functional and structural components: A non-local approach based on prediction

We present RIPMMARC (Rotation Invariant Patch-based Multi-Modality Analysis aRChitecture), a flexible and widely applicable method for extracting information unique to a given modality from a multi-modal data set. We use RIPMMARC to improve the interpretation of arterial spin labeling (ASL) perfusion images by removing the component of perfusion that is predicted by the underlying anatomy. Using patch-based, rotation invariant descriptors derived from the anatomical image, we learn a predictive relationship between local neuroanatomical structure and the corresponding perfusion image. This relation allows us to produce an image of perfusion that would be predicted given only the underlying anatomy and a residual image that represents perfusion information that cannot be predicted by anatomical features. Our learned structural features are significantly better at predicting brain perfusion than tissue probability maps, which are the input to standard partial volume correction techniques. Studies in test–retest data show that both the anatomically predicted and residual perfusion signals are highly replicable for a given subject. In a pediatric population, both the raw perfusion and structurally predicted images are tightly linked to age throughout adolescence throughout the brain. Interestingly, the residual perfusion also shows a strong correlation with age in selected regions including the hippocampi (corr=0.38, p-value <10−6), precuneus (corr=−0.44, p<10−5), and combined default mode network regions (corr=−0.45, p<10−8) that is independent of global anatomy-perfusion trends. This finding suggests that there is a regionally heterogeneous pattern of functional specialization that is distinct from that of cortical structural development.

Abstract 5: Dynamic Perfusion Assessment of Collateral Blood Flow in MCA Occlusion as Indicator of Tissue Fate

Objective: M1 occlusions with concomitant large cortical infarcts can result in life long severe disabilities. Therapeutic decisions for iv thrombolysis only or bridging to endovascular treatment require sufficient information of collateral flow. Hence, we tried to evaluate the potential of dynamic perfusion in the assessment of collateral flow with respect to tissue fate and outcome. Method: Between 1/2009 until 1/2012 ninety patients with MCA infarcts due to M1 occlusions were examined in a 3 T MR scanner. Selection criteria were DWI, PI, MR angiography, NIHSS on admission; FLAIR, MR angiography day 6 and mRS day 90, which left 30 patients for further evaluation. For evaluating dynamic perfusion the second previous image was subtracted from each frame of the raw perfusion data and the final images were assessed according to Higashida′s scale (Stroke 2003;34: e109-37). Rapid collateral flow was defined as arrival of signal drop in the ischemic side before the last arterial phase signal in the unaffected side. The ischemic region was monitored for complete or incomplete filling and flow direction. Results: Collateral flow was inversely correlated with infarct size on day 1 (p=0.005) and day 6 (p=0.006) as well as infarct growth (p=0.025) indicating collateral flow can influence infarct development. Collateral grade was significantly better in proximal M1 occlusions than in distal ones (p&lt;0.001), however the capillary filling was significantly delayed (p&lt;0.001). Mismatch volume was found to directly correlate with the delay of capillary filling (p=0.015) but not with infarct growth (p=0.064) or infarct size on day 6 (p=0.15). Higher NIHSS was directly correlated with capillary delay (p=0.042) and inversely correlated with collateral grade (p=0.027). mRS at day 90 correlated inversely with collateral grade (p=0.027), directly with infarct size on day 1 (p=0.006) and day 6 (p=0.011) demonstrating good collateral flow beneficial for tissue survival. Conclusion: Dynamic perfusion can provide additional information of collateral flow and capillary delay, which showed strong associations with infarct size and infarct growth, as well as clinical scores, such as NIHSS and mRS at follow-up. It can add to the assessment of tissue fate before therapy decision.

Abstract W P32: Association of Low Cerebral Blood Volume With Negative FLAIR Hyperintensity in Acute Ischemic Stroke

Objective: FLAIR hyperintensity is being used in clinical trials as a surrogate imaging biomarker for stroke onset time to test the safety of thrombolysis. Studies have shown that patients with negative and positive FLAIR hyperintensity overlap at similar time points from stroke onset in the early phase of acute ischemic stroke (AIS). Hyperintensity on FLAIR MRI likely represents increased tissue water content. We sought to determine if cerebral blood volume (CBV) mediates FLAIR hyperintensity in the early phase of AIS. Methods: AIS patients seen in 2012 were included in the study if i) onset time was known, ii) an MRI with perfusion was performed within 12 hours of onset time, iii) had imaging confirmed vascular occlusion of ICA, M1, or M2. Following co-registration of raw perfusion images with FLAIR, CBV maps were generated using PMA ASIST™ software. Two raters blinded to clinical information separately evaluated the DWI, FLAIR and CBV maps and measured the signal intensity ratio (SIR) for the brightest region on FLAIR normalized by homologous contra-lateral tissue. The SIR was similarly measured for CBV in same region. FLAIR negative was defined as SIR&lt;1.15, “Low CBV” was defined as CBV SIR &lt;0.5. Results: One hundred eighty two patients were screened and 30 met all study criteria; 21 women, with mean age of 71 (± 16) years and median NIHSS 18 (IQR 9-22). Using linear regression analysis, CBV SIR was associated with FLAIR SIR (p &lt;0.049). In the 0-3hr time window, overall CBV was not associated with FLAIR hyperintensity. However, in the 3-7.5hr time window, patients with negative FLAIR were more likely to have low CBV and conversely, patients with positive FLAIR were more likely to have normal CBV. Conclusion: CBV likely mediates FLAIR hyperintensity in 3-7.5hr of stroke onset but it has less impact on FLAIR hyperintensity in the first 3 hours of AIS. Low CBV could be a potential surrogate imaging biomarker in addition to FLAIR hyperintensity in the early phase of AIS.

Abstract 3007: What Perfusion Parameters Best Define Ischemic Penumbra in Patients with TIA and Minor Stroke; Data from the VISION study

Background: Patients with TIA and minor ischemic (MIS) stroke are at high risk for clinical and radiographic deterioration. Acute Perfusion-Diffusion (PWI-DWI) mismatch have been used to identify those at risk for early deterioration. However, optimal perfusion thresholds for ischemic penumbra have not been defined in TIA/MIS patients. We aimed to determine the interrater reliability of different perfusion measurements and the optimal values that predict DWI lesion expansion in patients with TIA/MIS. Methods: Patients with minor stroke (NIH Stroke Scale ≤3) and TIA were prospectively enrolled and imaged within 24 hours of symptom onset. All patients had follow-up imaging at day 30. Raw perfusion images were used to generate maps of time to peak of the impulse response (Tmax), cerebral blood flow (CBF) and cerebral blood volume (CBV). DWI and PWI volumes were measured with planimetric and thresholding techniques by two independent groups of readers (1 group with a single experienced rater and a second group with 2 inexperienced raters) and 95% confidence intervals were calculated. Infarct progression (FLAIR 30 vol-DWI Acute vol) was a priori defined as growth of at least 2.5 ml. Results: At baseline 55.2% of the 116 patients included had DWI lesions and 42.2% had PWI deficits. The inter-rater reliability for assessment of hypoperfused regions was excellent for all Tmax thresholds (Tmax+2s (Intraclass Correlation Coefficient (ICC)=0.996 [0.994, 0.997], Tmax+4s(0.99 [0.985, 0.993], Tmax+6s (0.998 [0.998, 0.999] and Tmax+8s(0.996 [0.994, 0.997]). The interrater agreement was very good for assessment of regions with reduced CBF (ICC=0.84[0.78-0.89] and moderate for evaluation of regions with low CBV (ICC=0.43 [0.27, 0.57]). 18.1% of patients had evidence of radiographic progression on follow-up. Mismatch volumes (PWI-DWI) calculated using Tmax+4s predicted infarct expansion (R=0.8, [0.21-0.28], P&lt;0.001). In contrast, mismatch volumes calculated with CBF did not predict infarct expansion (R=0.04, [-0.31,0.50] P=0.66). A mismatch volume (Tmax+4s-DWI) of ≥10 ml predicted infarct progression with 81% sensitivity and 92% specificity (AUC =0.85, [0.74-0.97]). Conclusion: In patients with TIA/MIS, penumbral patterns are common. Application of perfusion thresholds produced Tmax maps with excellent intra-rater reliability. Mismatch volume calculated as Tmax+4s-DWI volume reliably predicted infarct expansion. Objective PWI-DWI mismatch definitions such as these may be useful for acute reperfusion treatment decision making in patients with minor ischemic symptoms

Refining the Perfusion–Diffusion Mismatch Hypothesis

The Echoplanar Imaging Thrombolysis Evaluation Trial (EPITHET) tests the hypothesis that perfusion-weighted imaging (PWI)-diffusion-weighted imaging (DWI) mismatch predicts the response to thrombolysis. There is no accepted standardized definition of PWI-DWI mismatch. We compared common mismatch definitions in the initial 40 EPITHET patients. Raw perfusion images were used to generate maps of time to peak (TTP), mean transit time (MTT), time to peak of the impulse response (Tmax) and first moment transit time (FMT). DWI, apparent diffusion coefficient (ADC), and PWI volumes were measured with planimetric and thresholding techniques. Correlations between mismatch volume (PWIvol-DWIvol) and DWI expansion (T2(Day 90-vol)-DWI(Acute-vol)) were also assessed. Mean age was 68+/-11, time to MRI 4.5+/-0.7 hours, and median National Institutes of Health Stroke Scale (NIHSS) score 11 (range 4 to 23). Tmax and MTT hypoperfusion volumes were significantly lower than those calculated with TTP and FMT maps (P<0.001). Mismatch > or =20% was observed in 89% (Tmax) to 92% (TTP/FMT/MTT) of patients. Application of a +4s (relative to the contralateral hemisphere) PWI threshold reduced the frequency of positive mismatch volumes (TTP 73%/FMT 68%/Tmax 54%/MTT 43%). Mismatch was not significantly different when assessed with ADC maps. Mismatch volume, calculated with all parameters and thresholds, was not significantly correlated with DWI expansion. In contrast, reperfusion was correlated inversely with infarct growth (R=-0.51; P=0.009). Deconvolution and application of PWI thresholds provide more conservative estimates of tissue at risk and decrease the frequency of mismatch accordingly. The precise definition may not be critical; however, because reperfusion alters tissue fate irrespective of mismatch.