Adaptive Solitary Pulmonary Nodule Segmentation for Digital Radiography Images Based on Random Walks and Sequential Filter

Abstract

Solitary pulmonary nodules (SPN) in digital radiography (DR) images often have unclear contours and infiltration, which make it a challenging task for traditional segmentation models to get satisfactory segmentation results. To overcome this challenge, this paper has proposed an adaptive SPN segmentation model for DR images based on random walks segmentation and sequential filter. First, the SPN image is decomposed to get the cartoon component, which is used to acquire a set of seeds. Second, the seeds selection tactic is employed to optimize the scope of walking pixels and reduce the number of seeds, which could reduce the computational cost. Finally, we incorporate the sequential filter and construct the new representation of the weight and the probability matrices. In this paper, by using a data set of 724 SPN cases, the proposed method was tested and compared with four different models, and five kinds of evaluation indicators were given to evaluate the effect of segmentation. Experimental results indicate that the proposed method performs well on the blurred edge, as it could get relatively accurate results.