A Technique for Lung Nodule Candidate Detection in CT Using Global Minimization Methods

The first stage in computer aided pulmonary nodule detection schemes is a candidate detection step designed to provide a simplified representation of the lung anatomy, such that features like the lung wall, and large airways are removed leaving only data which has greater potential to be a nodule. Nodules which are connected to blood vessels tend to be characterized by irregular geometrical features which can result in their remaining undetected by rule-based classifiers relying only local image metrics. In the current paper a novel approach for lung nodule candidate detection is proposed based on the application of global segmentation methods combined with mean curvature minimization and simple rule-based filtering. Experimental results indicate that the proposed method can accurately detect nodules displaying a diverse range of geometrical features.