Algorithms, Architecture, Validation of an Open Source Toolkit for segmenting CT Lung lesions

We implented a lesion segmentation toolkit with a focus on segmenting solid and part-solid lesions from lung CT. The algorithm constructs four three-dimensional features to detect the lung wall, vessels, soft and diffuse tissue and edges. These features form boundaries and propagation zones that guide the evolution of a subsequent shape detection level set. User input is used to determine an initial seed point for the level set and users may also set a region of interest around the lesion. The methods are validated against 18 nodules on an anthropomorphic thorax phantom simulating a realistic anatomy, acquired under various scanner parameters. We also validated repeatability using 6 coffee-break (zero volume change) clinical cases. The methods, clinical and testing datasets have been made available to encourage reproducibility and foster scientific exchange. We integrated the methods into a user-friendly visualization application. We will also describe how one may obtain the toolkit and generate results for the VOLCANO workshop.