Automated Lung Parenchyma Segmentation in the Presence of High Attenuation Patterns Using Modified Robust Spatial Kernel FCM

Employing an accurate lung segmentation procedure is an essential stride in any Computer Aided Diagnosis (CAD) system designed for lung pathology evaluation, which significantly affects the performance of the system. In this paper a fully automated Modified Robust Spatial Kernel Fuzzy C Means (MRSKFCM) algorithm is used for extracting the lung parenchyma affected with Diffuse Lung Disease (DLD). The algorithm comprises of two steps, the Robust Spatial Kernel Fuzzy C Means (RSKFCM) is applied to acquire the coarse lung segmentation in the first step and in the second step the results of coarse segmentation is improved using Convex Hull (CH) algorithm and morphological operations. The algorithm used demonstrates high segmentation results with Dice index of 0.9057 ± 0.025, Jaccard Index of 0.8561 ± 0.029 and Accuracy 0.9864 ± 0.002. The execution time for each slice is approximately 2.5 s. Based on segmentation results and the evaluation time, the algorithm used can be considered as one of the desirable method for the extracting lung parenchyma in the CAD system.