Machine Learning Techniques in Diagnosis of Pulmonary Embolism

Aim: Pulmonary embolism (PE), is a high mortality disease which clinical suspicion and a variety of diagnostic laboratory and imagingresults have a high importance in diagnose. Anticoagulation and fybrinolytic treatments are hard to decide in some cases therefore early diagnose is important in emergency medicine.Material and Method: The study was designed retrospectively based on the records of the 201 patients who were presenting to Emergency Department with pulmonary complaints including dyspnea and chest pain between January 2010 and October 2013. Results: Machine learning techniques were used for calculating the success in diagnosing PE. The success rate of the classification tree method for detection of PE was 95%, which was higher than that of KNN classification (75%) and Naive Bayes Classification (88.5%). Discussion: Classification tree and Bayesian method can be selected ones to diagnose or define possibility of pulmonary embolism in emergency centers with limited study tests and for the patients difficultly diagnosed