Segmentation of Ground Glass Opacity Pulmonary Nodules Based on a Modified Random Walker Approach

摘要

The segmentation of pulmonary nodules in computed tomography (CT) images has become a significant research focus in medical image processing domains. Especially in case of ground glass opacity (GGO) nodules, this type has a high risk of being malignancy rate. The GGO nodule segmentation remains challenging due to irregular shapes, inhomogeneous intensities and blurry boundaries. Due to its simplicity and efficiency, the random walker algorithm has been progressively popular for interactive image segmentation. However, it may fail when the object has the fuzzy boundary and inhomogeneous intensity. In this paper, we put forward a modified random walker approach for the segmentation of GGO pulmonary nodules, which contains a novel weight function and an energy function of the random walker. Experiments performed on the Lung Images Dataset Consortium (LIDC) demonstrate the effectiveness of the proposed approach. The proposed method achieves the overlap measure of 0.88 +/- 0.07. The quantitative analysis also demonstrates the benefit of introducing a novel weight function for the difficult cases of GGO pulmonary nodules.