Contributions from computational intelligence to healthcare data processing

Abstract

The increasing ability to gather, store and process health care information, through the electronic health records and improved communication methods opens the door for new applications intended to improve health care in many different ways. Crucial to this evolution is the development of new computational intelligence tools, related to machine learning and statistics. In this thesis we have dealt with two case studies involving health data. The first is the monitoring of children with respiratory diseases in the pediatric intensive care unit of a hospital. The alarm detection is stated as a classification problem predicting the triage selected by the nurse or medical doctor. The second is the prediction of readmissions leading to hospitalization in an emergency department of a hospital. Both problems have great impact in economic and personal well being. We have tackled them with a rigorous methodological approach, obtaining results that may lead to a real life implementation. We have taken special care in the treatment of the data imbalance. Finally we make propositions to bring these techniques to the clinical environment.