Segmentation of Pulmonary Nodules Based on Statistic Features of Wavelet Coefficients and Dual Level Sets

A major problem of pulmonary nodules segmentation can´t be solved well by conventional methods, which is other tissue in chest CT image slices, such as blood vessels and bronchi, often overlap with the nodules and they also have the same gray scale intensity approximately, for big size (>40pixels) nodules especially. This paper presents a novel approach to solve above problem, which works in two main steps: (1) transition region (TR), which is defined as the ambiguous region between nodule and background, is ascertained depending on statistic features of wavelet coefficients. (2) Precise boundaries of the nodule is segmented based on an improvement of dual level sets method. The validity of the proposed approach is demonstrated in the chest CT images. Experiments with real chest CT images confirm the high accuracy of our approach.