Computer-Aided Prognosis

Publisher Summary This chapter presents a comparison of several ways to predict survival following diagnosis and treatment of a disease. It discusses alternative estimators of the five-year survival rate and compares their performance on a large group of breast cancer cases. The chapter presents a density estimation method and a standard method for the estimation of the five-year survival rate and then compares several other regression methods for the same problem. The density estimation method has two advantages over the regression methods: (1) incomplete data can be used, and (2) rough estimates of the accuracy of the probability assessments are available. The regression method also has two advantages over the density estimation method: (1) regression of survival time on indicants retains detailed information about survival times, which density estimation methods lose by dichotomizing survival time, and (2) a regression method can furnish a probability distribution of survival time, given an array of indicants.