Introduction: Perfusion imaging (PI) could guide decision-making for endovascular treatment (EVT) of acute ischemic stroke (AIS). However, PI was underused even in the US before the pivotal EVT trials proved its usefulness in 2018. This study aimed to describe the secular trends of PI utilization and investigate the effectiveness of PI-based EVT in real-world practice. Methods: Using a prospective multicenter (n=17) stroke registry in South Korea, we identified patients with AIS who presented within 24 hours from onset between 2011 and 2021. The study period was divided into 3 epochs: 2011-2014, 2015-2017, and 2018-2021. The study population was divided into the early (arrival within 6 hours) and late window (6-24 hours) groups. Results: A total of 51,650 patients (15,654 patients in 2011-2014, 14,432 patients in 2015-2017, and 21,564 patients in 2018-2021) were analyzed. Utilization of PI decreased in the overall population and early window group ( P trend <0.001); 43.3% and 54.1% in 2011-2014, 40.1% and 44.1% in 2015-2017, and 38.4% and 40.2% in 2018-2021, respectively; but increased in the late window group ( P trend <0.001); 31.3% in 2011-2014, 35.7% in 2015-2017, and 36.5% in 2018-2021. Of 10,872 patients with anterior large-vessel occlusion (aLVO), the EVT rate was not different between patients with and without PI (48.7% vs. 46.6%, P =0.08) in the early window but higher in those with PI than without PI in the late window (29.8% vs. 18.7%, P <0.001). The EVT outcome (3-month mRS 0-2) was not different between patients with and without PI in the early window (44.1% vs. 41.8%, P =0.21) and late window (38.4% vs. 39.2%, P =0.81). Propensity score analysis and instrumental variable analysis with PI rate per center as an instrument will be performed to adjust imbalances between patients with and without PI. Conclusion: Between 2011 and 2021 in South Korea, PI utilization has decreased in patients arriving within 6 hours from onset but has increased in those arriving between 6 and 24 hours. Among patients with aLVO, PI likely increased the EVT rate in the late window but did not in the early window. PI utilization did not seem to affect the EVT outcomes, but in-depth analysis is required.

Background: Deep learning-based artificial intelligence techniques have been developed for automatic segmentation of diffusion-weighted magnetic resonance imaging (DWI) lesions, but currently mostly using single-site training data with modest sample sizes. Objective: To explore the effects of 1) various sample sizes of multi-site vs. single-site training data, 2) domain adaptation, the utilization of target domain data to overcome the domain shift problem, where a model that performs well in the source domain proceeds to perform poorly in the target domain, and 3) data sources and features on the performance and generalizability of deep learning algorithms for the segmentation of infarct on DW images. Methods: In this nationwide multicenter study, 10,820 DWI datasets from 10 hospitals (Internal dataset) were used for the training-and-validation (Training-and-validation dataset with six progressively larger subsamples: n=217, 433, 866, 1,732, 4,330, and 8,661 sets, yielding six algorithms) and internal test (Internal test dataset: 2,159 sets without overlapping sample) of 3D U-net algorithms for automatic DWI lesion segmentation. In addition, 476 DW images from one of the 10 hospitals (Single-site dataset) were used for training-and-validation (n=382) and internal test (n=94) of another algorithm. Then, 2,777 DW images from a different hospital (External dataset) and two ancillary test datasets (I, n=50 from three different hospitals; II, n=250 from Ischemic Stroke Lesion Segmentation Challenge 2022) were used for external validation of the seven algorithms, testing each algorithm performance vs. manual segmentation gold standard using DICE scores as a figure of merit. Additional tests of the six algorithms were performed after stratification by infarct volume, infarct location, and stroke onset-to-imaging time. Domain Adaptation was performed to fine-tune the algorithms with subsamples (50, 100, 200, 500, and 1000) of the 2,777 External dataset, and its effect was tested using a) 1,777 DW images (from the External dataset, without overlapping sample) and b) 2,159 DW images from the Internal test dataset. Results: Mean age of the 8,661 patients in the Training-and-validation dataset was 67.9 years (standard deviation 12.9), and 58.9% (n = 4,431) were male. As the subsample size of the multi-site dataset was increased from 217 to 1,732, algorithm performance increased sharply, with DSC scores rising from 0.58 to 0.65. When the sample size was further increased to 4,330 and 8,661, DSC increased only slightly (to 0.68 and 0.70, respectively). Similar results were seen in external tests. Although a deep learning algorithm that was developed using the Single-site dataset achieved DSC of 0.70 (standard deviation 0.23) in internal test, it showed substantially lower performance in the three external tests, with DSC values of 0.50, 0.51, and 0.33, respectively (all p < 0.001). Stratification of the Internal test dataset and the External dataset into small (< 1.7 ml; n = 994 and 1,046, respectively), medium (1.7-14.0 ml; n = 587 and 904, respectively), and large (> 14.0; n = 446 and 825, respectively) infarct size groups, showed the best performance (DSCs up to ~0.8) in the large infarct group, lower (up to ~0.7) in the medium infarct group, and the lowest (up to ~0.6) in the small infarct group. Deep learning algorithms performed relatively poorly on brainstem infarcts or hyperacute (< 3h) infarcts. Domain adaptation, the use of a small subsample of external data to re-train the algorithm, was successful at improving algorithm performance. The algorithm trained with the 217 DW images from the Internal dataset and fine-tuned with an additional 50 DW images from the External dataset, had equivalent performance to the algorithm trained using a four-fold higher number (n=866) of DW images using the Internal dataset only (without domain adaptation). Conclusion: This study using the largest DWI data to date demonstrates that: a) multi-site data with ~1,000 DW images are required for developing a reliable infarct segmentation algorithm, b) domain adaptation could contribute to generalizability of the algorithm, and c) further investigation is required to improve the performance for segmentation of small or brainstem infarcts or hyperacute infarcts.

BackgroundRegional eloquence of brainstem structures may contribute to neurological status in basilar artery occlusion (BAO) stroke. The corticospinal tract (CST) which is vulnerable to BAO is important for motor activity. This study investigated the impact of CST salvage on outcomes and its associated factors in patients with BAO treated with thrombectomy.MethodsWe retrospectively investigated 88 patients with BAO admitted ≤24 h after onset and presented with motor deficits and who underwent thrombectomy. Patients with a pre-stroke modified Rankin Scale (mRS) score of 4–5 who did not undergo baseline brain computed tomography angiography were excluded. CST salvage was evaluated using follow-up imaging (magnetic resonance imaging [MRI] or computed tomography when MRI was not available) after thrombectomy. A good outcome was defined as a 3-month mRS score of ≤2 or 3 if a patient's pre-stroke mRS score was 3. The associations between CST salvage and outcomes and clinical parameters were analyzed using logistic regression analyses.ResultsThirty-nine (44.3%) patients had CST salvage and the same number of patients had good outcomes. CST salvage was independently associated with a good outcome [adjusted odds ratio (aOR): 18.52, 95% confidence interval (CI): 4.31–79.67, p &lt; 0.001]. After adjusting for confounders, atrial fibrillation (aOR: 3.92, 95% CI: 1.18–13.00, p = 0.026), location of occlusion (mid-BAO; aOR: 0.21, 95% CI: 0.06–0.72, p = 0.013), length of occlusion (involved segment of BAO &lt;2; aOR: 4.77, 95% CI: 1.30–17.59, p = 0.019), and onset-to-puncture-time ≤180 min (aOR: 4.84, 95% CI: 1.13–20.75, p = 0.034) were significantly associated with CST salvage.ConclusionCST salvage was associated with good functional outcomes in patients with BAO treated with thrombectomy. The presence of atrial fibrillation, location and length of BAO may predict CST salvage after thrombectomy, and rapid treatment with thrombectomy may protect this eloquent tract in these patients.

Introduction: There is lack of knowledge on whether symptomatic steno-occlusion (SYSO), common in acute ischemic stroke (AIS) patients with atrial fibrillation (AF), could increase the long-term risk of stroke recurrence in these patients. Methods: From a prospective cohort of patients with AIS and AF enrolled in 14 centers between Oct 2017 and Dec 2018, we identified patients who underwent MR angiography during hospitalization and completed 3-year follow-up including death during follow-up. SYSO was defined as (1) ≥ 50% stenosis or occlusion of cerebral arteries relevant to acute infarction or (2) any residual stenosis after endovascular treatment. Using cause-specific hazard models with non-stroke death as a competing risk, the risk of any recurrent stroke and recurrent ischemic stroke was estimated according to SYSO, respectively. Results: A total of 889 patients (mean age, 74.4 years; men, 54.6 %; median NIHSS, 6) were analyzed for this study. During the median 1096 days of follow-up, 152 any recurrent strokes, 142 recurrent ischemic strokes, and 208 deaths were observed. Patients with SYSO, compared to those without, were more likely to be older, be female, have hypertension, diabetes and history of stroke/TIA, and be on antiplatelets at discharge and were less likely to be on anticoagulants at discharge ( p <.05). The cumulative incidence of recurrent stroke in patients with and without SYSO was 25.2% and 8.3% at 1 month, 33.1% and 9.9% at 1 year, and 41.8% and 13.1% at 3 years, respectively ( p <.001). With adjusting age, sex, hypertension, diabetes, history of stroke/TIA, discharge antiplatelets, and discharge anticoagulants, SYSO increased the risk of any stroke recurrence (adjusted hazard ratio [95% confidence interval]; 3.02 [2.18-4.20]; p <.001) and ischemic stroke recurrence (3.20 [2.28-4.51]; p <.001). Conclusions: SYSO in AIS patients with AF substantially increased the risk of recurrent stroke by a 3-fold or more. Accordingly, SYSO should be considered in stratifying the risk of recurrence in AIS patients with AF.