Pulmonary CT image classification using evolutionary programming
Mark T. Madsen, Renuka Uppaluri, Eric A. Hoffman, Geoffrey McLennan · 1997 IEEE Nuclear Science Symposium Conference Record · 2002
The authors report on the use of evolutionary programming for classifying lung CT images. Evolutionary programming uses a genetic algorithm to generate a complete, compilable program that optimizes a solution to set of training data, In this case, the training set consisted of 17 features derived from multiple lung CT images along with an indicator of the target area from which the features originated. The image features included 5 parameters based on histogram analysis, 11 parameters based on run length and co-occurrence matrix measures, and the fractal dimension. Evolutionary programming produced solutions that compared favorably with more complicated and sophisticated Bayesian classifiers. The results of this study suggest that evolutionary programming is a powerful tool for developing classification algorithms.