Abstract
The authors propose a method for the classification between malignant pulmonary nodules and benign ones in 3D chest X-ray CT images. The method extracts nine image features and applies a SVM for the classification. For extracting the image features, the proposed method firstly segments the region of a nodule in a given X-ray CT image by applying a boundary detector, which is constructed by means of the AdaBoost, followed by a graph-cut. After the nodule region segmentation, geometrical image features are extracted for computing the nine image features. The experimental results show the quantitative evaluation of the contribution of each image feature to the classification.
