Diagnosing prostate cancer using backpropagation neural network and greedy decision procedure

Abstract

A novel procedure for diagnosing prostate cancer (PC) based on Back propagation Neural Network (BPNN) is proposed. Elderly men with symptoms such as urinary retention, urinary hesitancy, urinary dribbling, burning urination, hematuria, etc. are considered as primary attributes. Prostate-specific antigen (PSA) level and Gleason score are the secondary attributes. Initial dataset is generated based on the clinical database. The BPNN assigns symptom levels of a set of patients based on their primary attributes. A greedy decision procedure predicts tumor stages of patients based on their strong symptom levels and secondary attributes. The simulation shows that the proposed procedure is an effective way for diagnosing prostate tumor stages.