Research Of Mortality Prediction Of ICU Patients Based On FAM-CART

Intensive care unit(ICU)is an important unit in modern hospital to rescue patients with critical illness,the mortality rate of ICU patients is an important index to measure the level of ICU treatment and the quality of service.At present,there are a variety of scoring systems for the assessment of the patient’s illness condition and the mortality prediction,but these scoring systems require a lot of manpower and financial resources.Therefore,in the context of rapid development in artificial intelligence,many researchers try to use data mining and machine learning methods to study the project of predicting mortality for ICU patients,and some progress has been made.However,it is limited to the academic research in the laboratory,and there is still a distance between the academic research and clinical application.At the same time,the predicting result of machine learning methods is hard to be explained,so it is difficult to be accepted by clinical doctors and nurses.Therefore,this paper presents a new methods of predicting mortality for ICU patients based on a hybrid FAM-CART model.This paper mainly introduces the FAM-CART model for mortality prediction of the ICU patients.Based on the analysis of the characteristics of the ICU patients’ illness condition assessment and mortality prediction methods,the monitoring data of ICU patients was sorted and analyzed,the data were pretreated by three filling methods respectively,named the normal value,mean value and binary value filling methods.After that,according to the clinical characteristics of physiological data,some features are extracted for prediction,and then the Fuzzy ARTMAP neural network was used to predicting mortality for ICU patients.The predicting results of the three methods are compared,and the best data filling method is adopted.At last,the FAM-CART model was used to predicting mortality for ICU patients,the results of FAM-CART model are compared and analyzed with the results of logistic regression,artificial neural network support vector machine and Adaboost algorithm.The research work of this paper are shown as follow:(1)Summarize and analyze the research status and deficiency of the mortality prediction methods for clinical ICU patients,and then propose a new method based on FAM-CART model for predicting mortality of ICU patients.(2)Propose an ICU patients’ mortality prediction method based on a hybrid FAM-CART model.Use the data to train Fuzzy ARTMAP neural network and calculate the centroids and confidence factors of the prototype nodes.And then combine Fuzzy ARTMAP with CART using the centroids and confidence factors.At last,build the FAM-CART model to predict the mortality of ICU patients.(3)By analyzing the characteristics and the degree of lack of ICU patients’ data sets,put forward three kinds of data filling methods,and then use the Fuzzy ARTMAP neural network to validate the data preprocessing methods.And select the data preprocessing method of the best prediction result.(4)Use the FAM-CART model to realize the ICU patients’ mortality prediction.And then compare the result of FAM-CART model with other machine learning methods to validate the prediction effect of the method proposed in this paper.This study aims to design an ICU patients’ mortality prediction method with not only a good prediction performance,but also can be understood and accepted by clinical doctors and nurses according to the clinical monitoring data of ICU.The results show that the method proposed in the paper has better prediction performance,can be used to provide a theoretical reference for clinical application.