Vector quantization-based automatic detection of pulmonary nodules in thoracic CT images

Abstract

Computer-aided detection (CADe) of pulmonary nodules from computer tomography (CT) scans is critical for assisting radiologists to identify lung lesions at an early stage. In this paper, we propose a novel CADe system for lung nodule detection based on a vector quantization (VQ) approach. Compared to existing CADe systems, the extraction of lungs from the chest CT image is fully automatic, and the detection and segmentation of initial nodule candidates (INCs) within the lung volume is fast and accurate due to the self-adaptive nature of VQ algorithm. False positives in the detected INCs are reduced by rule-based pruning in combination with a feature-based support vector machine classifier. We validate the proposed approach on 60 CT scans from a publicly available database. Preliminary results show that our CADe system is effective to detect nodules with a sensitivity of 90.53 % at a specificity level of 86.00%.