Aims: Early aggressive fluid resuscitation in acute pancreatitis is frequently recommended but its benefits remain unproven. The aim of this study was to determine the outcomes associated with early fluid volume administration in the emergency room (FVER) in patients with acute pancreatitis. Methods: A four-center retrospective cohort study of 1010 patients with acute pancreatitis was conducted. FVER was defined as any fluid administered from the time of arrival to the emergency room to 4 h after diagnosis of acute pancreatitis, and was divided into tertiles: nonaggressive (<500 ml), moderate (500 to 1000 ml), and aggressive (>1000 ml). Results: Two hundred sixty-nine (26.6%), 427 (42.3%), and 314 (31.1%) patients received nonaggressive, moderate, and aggressive FVER respectively. Compared with the nonaggressive fluid group, the moderate group was associated with lower rates of local complications in univariable analysis, and interventions, both in univariable and multivariable analysis (adjusted odds ratio (95% confidence interval): 0.37 (0.14-0.98)). The aggressive resuscitation group was associated with a significantly lower need for interventions, both in univariable and multivariable analysis (adjusted odds ratio 0.21 (0.05-0.84)). Increasing fluid administration categories were associated with decreasing hospital stay in univariable analysis. Conclusions: Early moderate to aggressive FVER was associated with lower need for invasive interventions.
We have developed a method of segmenting DCIS lesions in WSIs using a U-Net architecture. The purpose of this study was to evaluate several different architectures and to determine the optimal resolution to field of view ratio for patches. The architecture trained at lowest resolution (5x) achieved the best test results (DSC=0.771, F1=0.601), implying that the U-Net benefits from having wider contextual information. A custom U-Net based architecture was trained to incorporate patches from all available resolutions. It achieved test results of DSC=0.759, F1=0.682, showing improvement in the object (duct) detecting capabilities of the model. Both architectures show comparable performance to a second expert annotator on the test set. This work will be used as the preliminary part of a pipeline targeted at predicting recurrence risk in DCIS patients.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.