Processing And Reasearch On ICU Patients Data

The existing ICU data analysis mainly focuses on mortality risk prediction and phenotypic analysis.However,in a dynamic clinical environment,they provide limited support for decision-making.This paper presents a new method to predict the severity of ICU patients by analyzing different organ systems at the same time,which can directly reflect the patientundefineds condition.Specifically,we propose a new deep learning model based on multi-view learning recurrent neural network,that is,MV-RNN.In the characterization of time series,the physiological characteristics of each organ system are learned by a single long-term short-term memory unit(LSTM)as a specific task.In order to make use of the relationship between organ systems,we use a shared LSTM unit to mine the correlation between different domains,so as to further improve the performance.We conducted extensive experiments on a real-world clinical data set(MIMIC-Ⅲ)to compare our methods with many of the most advanced methods.The experimental results show that this method performs well in predicting the severity of disease.Laboratory tests in critically ill patients are interpreted as normal or abnormal based on comparison to a reference interval that is generated by sampling healthy outpatient volunteers.Whether this represents a useful or valid comparison has never been demonstrated.Cross-sectional study of a large critical care database,the Medical Information Mart for Intensive Care database(MIMIC).Common laboratory measurements over the time window of interest were sampled.We calculated the overlapping coefficient(OVL)and Cohen’s standardized mean difference(SMD)between distributions,and created kernel density estimate visualizations for the relationship between laboratory values and the probability of death or quartile of ICU length-of-stay.All laboratory values for the best outcome group differed significantly from those in the worst outcome group.Both the best and worst outcome group curves revealed little overlap with and marked divergence from the reference range.