Background: Intravenous thrombolysis (IVT) plays a prominent role in the treatment of acute ischemic stroke (AIS). The sooner IVT is administered, the higher the odds of a good outcome. Therefore, registering the in-hospital time to treatment with IVT, i.e. the door-to-needle time (DNT), is a powerful way to measure quality improvement. The aim of this study was to identify determinants that are associated with extended DNT. Methods: Patients receiving IVT in 2015 and 2016 registered in the Dutch Acute Stroke Audit were included. DNT and onset-to-door time (ODT) were dichotomized using the median (i.e. extended DNT) and the 90th percentile (i.e. severely extended DNT). Logistic regression was performed to identify determinants associated with (severely) extended DNT/ODT and its effect on in-hospital mortality. A linear model with natural spline was used to investigate the association between ODT and DNT. Results: Included were 9518 IVT treated patients from 75 hospitals. Median DNT was 26 min (IQR 20-37). Determinants associated with a higher likelihood of extended DNT were female sex (OR 1.17, 95% CI 1.05-1.31) and admission during off-hours (OR 1.12, 95% CI 1.01-1.25). Short ODT correlated with longer DNT, whereas longer ODT correlated with shorter DNT. Young age (OR 1.38, 95% CI 1.07-1.76) and admission to a comprehensive stroke center (OR 1.26, 1.10-1.45) were associated with severely extended DNT, which was associated with in-hospital mortality (OR 1.54, 95%CI 1.19-1.98). Conclusions: Even though DNT in the Netherlands is short compared to other countries, lowering the DNT may be achievable by focusing on specific subgroups.

Ventricular volume and its progression are known to be linked to several brain diseases such as dementia and schizophrenia. Therefore accurate measurement of ventricle volume is vital for longitudinal studies on these disorders, making automated ventricle segmentation algorithms desirable. In the past few years, deep neural networks have shown to outperform the classical models in many imaging domains. However, the success of deep networks is dependent on manually labeled data sets, which are expensive to acquire especially for higher dimensional data in the medical domain. In this work, we show that deep neural networks can be trained on muchcheaper-to-acquire pseudo-labels (e.g., generated by other automated less accurate methods) and still produce more accurate segmentations compared to the quality of the labels. To show this, we use noisy segmentation labels generated by a conventional region growing algorithm to train a deep network for lateral ventricle segmentation. Then on a large manually annotated test set, we show that the network significantly outperforms the conventional region growing algorithm which was used to produce the training labels for the network. Our experiments report a Dice Similarity Coefficient (DSC) of 0.874 for the trained network compared to 0.754 for the conventional region growing algorithm (p < 0.001).