Automatic lung nodule detection from chest CT data using geometrical features: initial results

In this paper, a complete system for automatic lung nodule detection from Chest CT data is proposed. The proposed system includes the methods of lung segmentation and nodule detection from CT data. The algorithm for lung segmentation consists ofsurrounding air voxel removal, body fat/tissue identification, trachea detection, and pulmonary vessels segmentation. The nodule detection algorithm comprises of candidate surface generation, geometrical feature generation and classification. The proposed system shows 88.2% sensitivity for nodule >=3mm with 8.91 false positive per dataset