Intuitionistic Fuzzy C Means Clustering for Lung Segmentation in Diffuse Lung Diseases

Lung segmentation is a challenging task when increased attenuation appears at the periphery of the lungs. In case of increased attenuation, the intensity of the lung tissue closely matches with surrounding non-lung tissues. This paper presents intuitionistic fuzzy c means clustering to segment the lungs with increased attenuation appearing at the border in diffuse lung diseases. In the proposed approach intuitionistic spatial kernel fuzzy c means is employed to obtain the initial lung segment and it is further refined using quick hull and morphological operations. The proposed approach is evaluated on ’TALISMAN’ diffuse lung diseases benchmark dataset. The dataset has 108 high resolution computed tomography scans with different types of diseases such as reticulation, fibrosis, ground glass opacity, consolidation, emphysema, micro and macro nodules. The evaluation results clearly indicate that the proposed approach achieves better segmentation accuracy on lung scans with increased attenuation appearing at the lung border.