Background: Image segmentation is an important part of computer-aided diagnosis\n(CAD), the segmentation of small ground glass opacity (GGO) pulmonary nodules is\nbeneficial for the early detection of lung cancer. For the segmentation of small GGO\npulmonary nodules, an integrated active contour model based on Markov random\nfield energy and Bayesian probability difference (IACM-MRFEBPD) is proposed in this\npaper.\nMethods: First, the Markov random field (MRF) is constructed on the computed\ntomography (CT) images, then the MRF energy is calculated. The MRF energy is used\nto construct the region term. It can not only enhance the contrast between pulmonary\nnodule and the background region, but also solve the problem of intensity inhomogeneity\nusing local spatial correlation information between neighboring pixels in the\nimage. Second, the Gaussian mixture model is used to establish the probability model\nof the image, and the model parameters are estimated by the expectation maximization\n(EM) algorithm. So the Bayesian posterior probability difference of each pixel can\nbe calculated. The probability difference is used to construct the boundary detection\nterm, which is 0 at the boundary. Therefore, the blurred boundary problem can be\nsolved. Finally, under the framework of the level set, the integrated active contour\nmodel is constructed.\nResults: To verify the effectiveness of the proposed method, the public data of the\nlung image database consortium and image database resource initiative (LIDC-IDRI)\nand the clinical data of the Affiliated Jiangmen Hospital of Sun Yat-sen University are\nused to perform experiments, and the intersection over union (IOU) score is used to\nevaluate the segmentation methods. Compared with other methods, the proposed\nmethod achieves the best results with the highest average IOU of 0.7444, 0.7503, and\n0.7450 for LIDC-IDRI test set, clinical test set, and all test sets, respectively.\nConclusions: The experiment results show that the proposed method can segment\nvarious small GGO pulmonary nodules more accurately and robustly, which is helpful\nfor the accurate evaluation of medical imaging.
Loading....