Pulmonary emphysema is a kind of lung disease and doctors diagnose it referring to the lung CT images findings in. Therefore computer aided diagnosis by image processing is very useful from the quantitative and objective points of view. In this study, we focused on shape and distribution of the area of lesions in the lung CT images, and propose a method for the classification of three kinds of pulmonary emphysema. We calculated five features to classify the emphysema types. Finally, we will classify emphysema by using those features and neural network in the feature work.