Displaying One Session

e-Poster Discussion
Session Type
e-Poster Discussion
Room
Station 11 (E-Poster Area)
Date
07.11.2020, Saturday
Session Time
06:30 PM - 07:30 PM

IDENTIFICATION OF LARGE VESSEL OCCLUSION ON NON-CONTRAST CT USING A DEEP LEARNING SOFTWARE

Session Type
e-Poster Discussion
Date
07.11.2020, Saturday
Session Time
06:30 PM - 07:30 PM
Room
Station 11 (E-Poster Area)
Lecture Time
06:30 PM - 06:35 PM

Abstract

Background And Aims

Reliable recognition of large vessel occlusion (LVO) patterns on non-contrast CT (NCCT) may accelerate the identification of endovascular treatment candidates. We aim to validate a machine learning algorithm (Deepstroke) to identify the presence of LVO on NCCT in suspected acute ischemic stroke (AIS) patients.

Methods

Patients with suspected AIS who underwent NCCT+CT Angiography (CTA) from two comprehensive stroke centers were included, patients with intracranial haemorrhage were excluded. ident. Two experienced radiologists identified the presence of LVO on CTA (NR-CTA) and served as ground truth. A deep learning system was used to create an algorithm to identify acute ischemia and clot signs on NCCT. Software image output was used to train a binary classifier to determine LVO on NCCT. Cross-validation was performed in a stratified 5-fold of the data, including deep learning training. We also studied software accuracy when adding NIHSS and time from onset to the model.

Results

We analyzed 1453 patients, 823(56.6%) with LVO by NR-CTA. The area under the curve (AUC) for LVO identification with Deepstroke was 0.87 (sensitivity 0.83, specificity 0.71, PPV 0.79, NPV 0.76), and it improved to 0.91 when combining imaging and clinical data (sensitivity 0.83, specificity 0.85, PPV 0.88, NPV 0.79). Among patients identified as LVO by automated software, only 13% showed no findings on NR-CTA.

image.pngexample2.png

Conclusions

In patients with suspected AIS, Deepstroke identifies LVO on NCCT with a high correlation with NR-CTA. Deepstroke software might reduce the need of performing CTA and increase the efficiency of patient management in stroke care networks.

Trial Registration Number

Not applicable

Hide

POTENTIAL WHITE MATTER RECOVERY IN STROKE PATIENTS TREATED WITH AUTOLOGOUS MONONUCLEAR CELLS

Session Type
e-Poster Discussion
Date
07.11.2020, Saturday
Session Time
06:30 PM - 07:30 PM
Room
Station 11 (E-Poster Area)
Lecture Time
06:35 PM - 06:40 PM

Abstract

Background And Aims

Bone marrow mononuclear cells (MNCs) attenuate secondary degeneration after ischemic injury in animal stroke models. We compared the integrity of the white matter in ischemic stroke (IS) patients treated with autologous bone marrow MNCs versus patients who did not receive cells (controls).

Methods

We imaged 37 IS patients (17=treated, 20=control) at 1, 3, and 12 months after stroke on a 3T MRI. A maximum of 10 million cells/kg were administered IV within 72 hrs of stroke. Using DTI , white matter integrity of the cerebrospinal tracts (CST) was assessed by relative (ipsilesional/contralesional) rFA whereas by absolute FA in the genu and splenium of the corpus callosum. A voxel-by-voxel global tracts based spatial statistics (TBSS) was used to compare the two groups.

Results

2924 eso_figure_1.png
Infarct size of controls was 26.4 ±27 mL and the treated group was 64.7 ± 40.3 mL. Given the larger infarct size, the FA in each region of treated group was significantly (p < 0 .05) lower than the control group but increased in the CST and splenium over time while the corresponding FA in the control group remained unchanged or trended downward over time. In the TBSS analysis, the differences in white matter integrity between the groups diminished over time (fig1), demonstrating converage over time.

Conclusions

Despite larger infarct size, the cell treated patients showed an improvement in white matter integrity while the control group showed minimal or worsening changes. These findings support a possible treatment of MNCs in stroke patients.

Hide

DETECTION OF ISCHEMIC CHANGES ON BASELINE MULTIMODAL COMPUTED TOMOGRAPHY: EXPERT READING VS. BRAINOMIX AND RAPID SOFTWARE

Session Type
e-Poster Discussion
Date
07.11.2020, Saturday
Session Time
06:30 PM - 07:30 PM
Room
Station 11 (E-Poster Area)
Lecture Time
06:40 PM - 06:45 PM

Abstract

Background And Aims

The aim of the study was to compare the assessment of ischemic changes by expert reading and available automated software for non-contrast CT (e-ASPECTS) and CT perfusion (RAPID) on baseline multimodal imaging and demonstrate the accuracy for the final infarct prediction.

Methods

Early ischemic changes were measured by ASPECTS on the baseline neuroimaging of consecutive patients with anterior circulation ischemic stroke. The presence of early ischemic changes was assessed a) on NCCT by two experienced raters, b) on NCCT by e-ASPECTS, and c) visually on derived CT perfusion maps (CBF<30%, Tmax>10s). Accuracy was calculated by comparing presence of final ischemic changes on 24-hour follow-up for each ASPECTS region and expressed as sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). The subanalysis for patients with successful recanalization was conducted.

Results

Of 263 patients, 81 fulfilled inclusion criteria. Median baseline ASPECTS was 9 for all tested modalities. Accuracy was 0.76, for e-ASPECTS, 0.79 for consensus, 0.82 for CBF<30% and 0.80 for Tmax>10s. e-ASPECTS, consensus, CBF<30%, and Tmax>10s had sensitivity 0.41, 0.46, 0.49 and 0.57, respectively; specificity 0.91, 0.93, 0.95 and 0.91, respectively; PPV 0.66, 0.75, 0.82 and 0.73, respectively; NPV 0.78, 0.80, 0.82 and 0.83, respectively. Results did not differ in patients with and without successful recanalization.

Conclusions

This study demonstrated high accuracy for the assessment of ischemic changes by different CT modalities with the best accuracy for CBF<30% and Tmax>10s. The use of automated software has a potential to improve the detection of ischemic changes.

Trial Registration Number

Not applicable

Hide

ESTIMATING NOCTURNAL STROKE SYMPTOM ONSET TIMES BASED ON FLUID-ATTENUATED INVERSION RECOVERY (FLAIR) SIGNAL INTENSITIES IN THE WAKE-UP TRIAL

Session Type
e-Poster Discussion
Date
07.11.2020, Saturday
Session Time
06:30 PM - 07:30 PM
Room
Station 11 (E-Poster Area)
Lecture Time
06:45 PM - 06:50 PM

Abstract

Group Name

for the WAKE-UP investigators

Background And Aims

Relative signal intensities of acute stroke lesions in FLAIR (FLAIR-rSI) have been shown to be linearly associated with time elapsed since stroke onset. We estimated previously unknown symptom onset times in a subgroup of patients from the WAKE-UP trial based on data from PRE-FLAIR, a multicenter observational trial utilizing FLAIR as a surrogate parameter for time since stroke onset.

Methods

FLAIR-rSI were quantified in stroke lesions in PRE-FLAIR and WAKE-UP. Stroke onset times were recorded in PRE-FLAIR and used to fit a linear regression model in relation to FLAIR-rSI, adjusted for patient age and lesion volume. The model was subsequently applied to FLAIR-rSI of stroke lesions in those patients screened for WAKE-UP who had symptom onset during night-sleep

Results

FLAIR-rSI was quantified in 399 patients from PRE-FLAIR. Linear regression indicated significant effects of age (p=0.001), lesion volume (p=0.005) and FLAIR-rSI (p<0.001, adjusted R2 = 0.179) on time since symptom onset. In data from 813 patients available from WAKE-UP, distribution of recorded events and estimated stroke onset are illustrated below. Median times of “last seen well” were one hour before midnight (IQR 2.4 hours), symptom recognition 7 hours (IRQ 2.2 hours) and estimated stroke onset time 4.5 hours after midnight (IQR 5.2 hours).

time_curves.jpg

Conclusions

Based on estimations from FLAIR-rSI, stroke onsets predominantly occurred during the early morning hours in patients screened for WAKE-UP. Although associated with a relevant amount of uncertainty, our results are in line with evidence of characteristic circadian patterns of increased risk for ischemic stroke.

Trial Registration Number

Not applicable

Hide

PERFUSION IMAGING PREDICTS EARLY NEUROLOGICAL DETERIORATION IN PATIENTS WITH EMERGENT LARGE VESSEL OCCLUSION PRESENTING MILD SYMPTOMS

Session Type
e-Poster Discussion
Date
07.11.2020, Saturday
Session Time
06:30 PM - 07:30 PM
Room
Station 11 (E-Poster Area)
Lecture Time
06:50 PM - 06:55 PM

Abstract

Background And Aims

A significant proportion of the patients with emergent large vessel occlusion (ELVO) presenting mild symptoms experienced early neurological deterioration (END). However, little is known about parameters associated with END in this population. This study aims to evaluate whether perfusion imaging can predict END among those with ELVO and mild symptoms.

Methods

A total of 94 patients were identified who met the following criteria: onset to arrival within 24 hours, baseline National Institute of Health Stroke Scale score was ≤5, and anterior circulation ELVO identified on baseline imaging. Patients who underwent endovascular therapy before occurring END were excluded. Perfusion parameters including Tmax >6s and mismatch volume (Tmax >6s – infarct core volume) were measured and dichotomized (highest quartile vs all other quartiles), then their associations with END were analyzed using multivariable logistic regression model.

Results

Median Tmax >6s and mismatch volume were 29.97 mL (inter quartile range (IQR): 10.15-56.67) and 24.96 mL (IQR: 6.45-46.25), respectively. We observed more occurrence of END in those with the highest quartile of Tmax >6s volume compared to those with the lower quartiles (odds ratio (OR): 5.44, 95% confidence interval (CI): 1.96–15.15, p=0.001) and in those with highest quartile of mismatch volume compared with those who were not (OR: 5.44 (1.96-15.15), p=0.001). These results were statistically significant even after adjusting confounding variables (adjusted OR (aOR) for Tmax >6s: 5.78 (1.61-20.76), p=0.007 and aOR for mismatch volume: 4.60 (1.37-15.52), p=0.014).

Conclusions

Tmax >6s and mismatch volume might be useful parameters for predicting END in patients with ELVO presenting mild symptoms.

Trial Registration Number

Not applicable

Hide

FLAIR HYPERINTENSE VESSELS – COLLATERALS YES, BUT RELEVANT? NOT NECESSARILY

Session Type
e-Poster Discussion
Date
07.11.2020, Saturday
Session Time
06:30 PM - 07:30 PM
Room
Station 11 (E-Poster Area)
Lecture Time
06:55 PM - 07:00 PM

Abstract

Background And Aims

The FLAIR hyperintense vessel (FHV) sign on MRI is a radiological association of vessel occlusion and indirect sign of collateral circulation, but is of uncertain clinical relevance. We explored the association of FHV with outcome in a homogenous subgroup from a randomized controlled trial in unknown onset stroke (WAKE-UP).

Methods

We retrospectively evaluated 165 patients who presented with a confirmed unilateral vessel occlusion. Two raters blinded to clinical and radiological outcome independently assessed the presence and extent of FHV. Consensus was reached for discrepant cases. Patients were then separated into two groups (few vs extensive FHV) at the median of FHV extent. We investigated the differences in radiological and clinical outcome between these groups.

Results

72.7% of all patients (n=165) and 80.6% of MCA-occlusion patients (n=134) showed FHV on baseline imaging. Neither the entire nor the MCA-occlusion cohort were characterized by a difference in mRS (p=0.119; p=0.267), NIHSS (p=0.223; p=0.402) or absolute lesion growth (p=0.543; p=0.884) between patients with few and extensive FHV.

Regression analysis revealed recanalization, occlusion site, baseline NIHSS and baseline lesion volume, but not FHV or treatment group, as major predictors of outcome in both the entire and MCA-occlusion cohort. Overall, patients experienced a reduction in FHV between baseline and follow-up (median 50%, IQR 14.9-100.0%), which was significantly more pronounced in patients who recanalized (p<0.001).

Conclusions

In this subgroup of WAKE-UP trial patients, we did not find evidence that the extent of FHV modifies the evolution of stroke. However, the chance of a type II error must be considered.

Trial Registration Number

NCT01525290

Hide

PORTABLE, BEDSIDE, LOW-FIELD MAGNETIC RESONANCE IMAGING OF ISCHEMIC STROKE

Session Type
e-Poster Discussion
Date
07.11.2020, Saturday
Session Time
06:30 PM - 07:30 PM
Room
Station 11 (E-Poster Area)
Lecture Time
07:00 PM - 07:05 PM

Abstract

Background And Aims

Advances in low-field MRI have enabled acquisition of images at the point-of-care (POC). The appearance of ischemic stroke in low-field MRI is unknown. We aim to describe the characteristics of ischemic stroke in low-field, POC MRI.

Methods

We studied 12 ischemic stroke patients (7 females; ages 33-92 years). T2-weighted (T2W), fluid-attenuated inversion recovery (FLAIR), and diffusion-weighted imaging (DWI) exams were acquired on a 64mT, portable bedside MRI system. Two patients underwent serial imaging studies 1) 46 hours and 2) 18 and 25 hours apart. A total of 43 POC images (14 T2W; 15 FLAIR; 14 DWI) were obtained. Two DWI images were excluded due to motion degradation. Two raters analyzed the images using a standardized qualitative evaluation for lesion presence, location, and morphology.

Results

POC exams were obtained within 7 days of symptom onset (3.5 ± 2.2 days). High-field (3T) MRI exams (23 ± 21 hours from POC scans) demonstrated ischemic infarcts in 11 patients. Ischemic infarcts were detected on all POC exams of patients with infarcts. Serial imaging demonstrated evolution of infarcts in both patients. Similar to 3T MRI, low-field T2W, FLAIR, and DWI images showed infarcts as demarcated areas of hyperintensity (Figure 1). Lesions were visualized in both supratentorial and infratentorial slices. Infarcts as small as 7.4 mm in diameter (5.2 ± 3.4 cm) were identified.

poc ischemic stroke example.png

Conclusions

These preliminary data suggest that low-field, POC MRI may be clinically useful in the evaluation of ischemic stroke. Further work is needed to evaluate this approach in hyperacute and acute settings.

Trial Registration Number

Not applicable

Hide

RESTING-STATE FUNCTIONAL MAGNETIC RESONANCE IMAGING AND ALTERED GRANGER CAUSAL CONNECTIVITY OF RESTING-STATE NEURAL NETWORKS IN PATIENTS WITH LEUKOARAIOSIS ASSOCIATED WITH COGNITIVE IMPAIRMENT: A CROSS-SECTIONAL STUDY

Session Type
e-Poster Discussion
Date
07.11.2020, Saturday
Session Time
06:30 PM - 07:30 PM
Room
Station 11 (E-Poster Area)
Lecture Time
07:05 PM - 07:10 PM
Presenter

Abstract

Group Name

on behalf of the LA clinical study Investigators and the Imaging Collaborators

Background And Aims

The present study evaluated the difference in resting-state neural networks between difference degree of cognitive impairment. The purpose of the study was to provide an imaging reference for the measurement of leukoaraiosis (LA) with cognitive impairment, and to reveal the pathogenesis of LA.

Methods

Eighty-two subjects were divided into three groups: LA with vascular dementia (LA-VaD), LA without vascular dementia (LA-VCIND),and normal controls (NC), and then gave them the resting-state functional magnetic resonance imaging (rs-fMRI), finally we used independent component analysis (LA-VCIND) to study the components of individual resting-state networks (RSNs) , and used multivariate granger causality analysis (mGCA) to analyze changes in the effective connectivity of these functional networks.

Results

effective connectivity of rsns of the three groups.jpgthe results of multivariate granger causality analysis.jpgaverage z scores of active voxels for each selected rsn of patients with the three groups.jpgTen RSNs were identified: primary visual network, secondary visual network, auditory network; sensorimotor network; anterior default mode network (DMN); posterior default mode network; salience network (SN); dorsal attention network (DAN); left working memory network (L-MeN), right working memory network (R-MeN). Using ICA, significant average Z score were found in DMN, SN, DAN and R-MeN between LA-VAD and NC group. Using mGCA, we found different connections and connection strengths between the networks of the three groups.

Conclusions

Our data suggested components of the RSNs changed as the disease progressed. The FC strength increased at an early stage of LA and decreased when cognitive impairment became worsened. Furthermore, the direction and strength of the connections between these networks changed depending on the level of cognitive impairment. These findings suggest the human brain compensates for specific functional deficits at different cognitive impairment caused by LA.

Trial Registration Number

Not applicable

Hide

Infarction Pattern in Diffusion-Weighted Image and Atrial Cardiopathy in Patients with Embolic Strokes of Unknown Source

Session Type
e-Poster Discussion
Date
07.11.2020, Saturday
Session Time
06:30 PM - 07:30 PM
Room
Station 11 (E-Poster Area)
Lecture Time
07:10 PM - 07:15 PM

Abstract

Background And Aims

Atrial cardiopathy is an emerging potential embolic stroke cause which refers to atrial structural or functional abnormalities in the absence of atrial fibrillation The aim of the present study was to clarify the difference of the infarction topography in diffusion-weighted image (DWI) according to the presence of atrial cardiopathy in patients with embolic strokes of undetermined source (ESUS).

Methods

We retrospectively identified consecutive patients with acute ischemic stroke who met the previously established ESUS criteria. Atrial cardiopathy was defined as the presence of P-wave terminal force >5000 mV X ms in ECG lead V1 or serum NT-proBNP >250 pg/mL. Infarction image profiles in DWI were compared between patients with and without atrial cardiopathy.

Results

Of 204 participants diagnosed as ESUS (78 women; mean age, 69.4 ± 12.8 years), atrial cardiopathy was identified in 77 (37.7%). Patients with atrial cardiopathy had more common a large cortical (28.6% vs 7.9%, P<0.001) and multi-territorial infarct (32,3% vs 5,8%, P=0.001) pattern and larger infarct volume in DWI (P=0.043) than them without. In multinominal analysis, the atrial cardiopathy group was more likely to have a large cortical infarct [odds ratio (OR) 2.54, 95% confidence interval (CI) 1.10–5.86] and multi-territorial lesion (OR 4.42; 95% CI 1.89–10.3) pattern than non-atrial cardiopathy group.

Conclusions

We found atrial cardiopathy was associated with multi-territorial and large infarction pattern in ESUS patients. Our findings suggest that left atrial cardiopathy may be cause of left atrial embolism, as a different mechanism of thromboembolism in ESUS patients.

Trial Registration Number

N/A

Hide

THE IMPACT OF MRI TEXTURAL FEATURES ON STROKE OUTCOME PREDICTION

Session Type
e-Poster Discussion
Date
07.11.2020, Saturday
Session Time
06:30 PM - 07:30 PM
Room
Station 11 (E-Poster Area)
Lecture Time
07:15 PM - 07:20 PM

Abstract

Background And Aims

FLAIR positivity of ischemic stroke lesions has been associated with tissue viability. We aimed to assess the impact of T2-FLAIR textural features on stroke outcome prediction.

Methods

We included an institutional cohort of 460 patients with acute ischemic stroke and MRI acquired within 48hours of symptom onset. Infarct lesion masks manually segmented on DWI, were co-registered to T2-FLAIR images. Textural features were extracted from T2-FLAIR images within lesion masks. A machine learning algorithm (XGboost) was used to build predictive models of good functional outcome (mRS≤2 at 90-180days). Inputs were varying sets of features: Clinical features vs. radiological & textural features vs. combination of both. The entire dataset was repeatedly split into train and test sets (80%-20%). Recursive feature elimination was performed within a five-fold stratified cross-validation on the train set, the model was then evaluated on the test set. Prediction performances were assessed based on AUC ROC values. SHAP values were computed to explore features participations to the models.

Results

Area under the ROC curves of the clinical, radiological and radioclinical models were respectively 0.77 (95% confidence interval: 0.71 0.83), 0.74 (0.67 0.82) and 0.80 (0.76 0.84). Selected, and therefore most important features of the model combining clinical and radiological features were NIHSS, age, brain volume, sex, white matter hyperintensity volume, pre-stroke mRS, anti-hypertensive treatment, diabetes mellitus, and textural features related to size, shape and to high-intensity confluence.

figure1_v2png.png

Conclusions

In this proof-of-concept study, T2-FLAIR textural features of infarct lesions appeared to improve stroke outcome prediction and may yield future insights into underlying pathology.

Hide