No abstract

No abstract

We evaluated the trend of admission of patients with acute cerebrovascular accidents (CVAs) during social distancing measures implemented during the coronavirus disease 2019 (COVID-19) era. The data of patients admitted with transient ischemic attack, ischemic stroke, intracerebral hemorrhage (ICH), and subarachnoid hemorrhage (SAH) to the emergency department of the Hanyang University Seoul Hospital were retrospectively analyzed. The data were compared between the pre-COVID-19 and COVID-19 periods. Poisson regression analysis was performed to evaluate changes in admission rates as a function of the year, social distancing level, and the interaction between the year and social distancing level. The number of admissions for CVAs dropped from 674 in the pre-COVID-19 period to 582 in the COVID-19 period. The decline in the number of admissions for ICH during social distancing measures was statistically significant, while the declines in SAH and ischemic stroke admissions were not. When the social distancing level was raised, admissions for CVAs dropped by 19.8%. The correlation between social distancing and decreased admissions for CVAs is a paradoxical relationship that may be of interest to the field of public health.

Intracerebral hemorrhage (ICH) is life-threatening. The intraventricular extension of ICH (IVH) also frequently occurs, thus increasing the risk of disability or death. The site and amount of bleeding are important predictors of prognosis. This study aimed to predict Glasgow Outcome Scores (GOSs) by automatically segmenting hemorrhage sites from brain computed tomography data combined with clinical information. Data of 192 patients from Hanyang University Hospital with IVH and ICH from March 2016 to September 2020 were analyzed. To perform automatic segmentation through deep learning, two neurosurgeons manually generated correct answer values. Preprocessing was performed to capture more pathological tissue. Hemorrhagic sites were robustly predicted by feeding high-quality generated data to the HTransUNet that hierarchically combined a Convolution Neural Network and transformer; the predicted mask was combined with clinical information to predict GOSs. This model showed high performance and competitive performance in segmenting IVH and ICH compared with other segmentation models. Additionally, when predicting GOS, it performed better than the ICH score system. The predicted mask with clinical information performed better than clinical information alone. Using the cerebral hemorrhage segmentation and GOS prediction models in clinical settings, as auxiliary indicators for rapid decision-making, can contribute significantly to patient management.