Pulmonary CT image classification with evolutionary programming

Rationale and Objectives. It is often difficult to classify information in medical images from derived features. The purpose of this research was to i nvestigate the use of evolutionary programming as a tool for selecting impo rtant features and generating algorithms to classify computed tomographic ( CT) images of the lung. Materials and Methods. Training and test sets consisting of 11 features der ived from multiple lung CT images were generated, along with an indicator o f the target area from which features originated. The images included five parameters based on histogram analysis, Il parameters based on run length a nd co-occurrence matrix measures, and the fractal dimension. Two classifica tion experiments were performed. In the first, the classification task was to distinguish between the subtle but known differences between anterior an d posterior portions of transverse lung CT sections. The second classificat ion task was to distinguish normal lung CT images from emphysematous images . The performance of the evolutionary programming approach was compared wit h that of three statistical classifiers that used the same training and tes t sets. Results. Evolutionary programming produced solutions that compared favorabl y with those of the statistical classifiers. In separating the anterior fro m the posterior lung sections. the evolutionary programming results were be tter than two of the three statistical approaches. The evolutionary program ming approach correctly identified all the normal and abnormal lung images and accomplished this by using less features than the best statistical meth od. Conclusion. The results of this study demonstrate the utility of evolutiona ry programming as a tool for developing classification algorithms.