Example-based assisting approach for pulmonary nodule classification in 3-D thoracic CT images

Abstract

This paper describes an example-based assisting approach for classifying pulmonary nodules in 3-D thoracic CT images. In this approach the internal and surrounding structures of the nodule are characterized by the distribution pattern of CT density and 3-D curvature indexes. Each nodule is represented by means of a joint histogram using the distance value from the nodule center. When given an indeterminate nodule image, the images of lesions with known diagnoses (e.g. malignant vs. benign) are retrieved from a 3-D nodule image database. The malignant likelihood of the indeterminate case is estimated by the difference between the representation patterns of the indeterminate case and the retrieved lesions. In the present study, we adopt the Mahalanobis distance as the difference measure and then, explore the feasibility of the classification based on patterns of similar lesion images.