ABSTRACT
In this chapter, the authors discuss the efficacy of a novel development of the genetic algorithm neural network method using Bayes’ theorem to predict patient outcome after resection of Nonsmall Cell Lung Cancer (NSCLC). NSCLC accounts for approximately three quarters of all lung cancer histologies. Between 1980–1992, 886 pulmonary resections from NSCLC patients were received by the Department of Histopathology, Broadgreen Hospital, Liverpool, UK. Positive predictive value and odds ratios were greater than prevalence and prior odds for all time points for the Genetic Algorithm Neural Network (GANN), and for all time points except 9 months for logistic regression. Significant differences in classification between GANN and logistic regression were found at the 9-, 12-, and 15-month time points. Sensitivity for classification by GANN was less than logistic regression at all time points. The factors that affect outcome of patients with NSCLC are unclear.
