Research On Improvement Of Immune Fuzzy Clustering Algorithm For Medical Image Segmentation

Medical image segmentation has been the hot issue in medical image analysis in recent years. Due to the interference of external factors like imaging equipment, the medical images present fuzzy boundaries and intensity inhomogeneity. So, it affects the doctor's diagnosis of the disease. Accordingly, it is important for us to find a method to do a quick, efficient and accurate segmentation of images in the follow-up clinic analysis.Due to the fuzzy uncertainty of the images, some scholars introduced the fuzzy theory into image processing, and used fuzzy clustering to do image segmentation. Fuzzy c-means algorithm (FCM) is used most widely, but we should first determine the initial clustering center and clustering number. Besides, FCM is sensitive to the noise and running into "prematurity" easily. Artificial immune algorithm (AIS) inherits the well characteristics of biological immune system and has the characteristics of distributed parallel processing, self-organization, self-learning, having been successfully applied in many fields. To improve FCM, this paper combines artificial immune algorithm and fuzzy clustering algorithm, providing a new immune clustering algorithm. The main achievements of this paper can be summarized as follows:1. The noise seriously affects the efficiency of the algorithm and segmentation result. Before segmenting medical images, we add improved extremum median filtering in the controllable link which can identify noise points and data points. Through the test, improved algorithm can both remove noise and retain image details.2. In order to find the number of the initial clustering center and clustering number in advance, we smooth gray histogram with the interpolation method. It can effectively filter pseudo peak point and get well initial clustering center and clustering number, which reduces iterative process of the algorithm and improves the efficiency of the algorithm. It is using the multi center combination that improves the accuracy and avoids being trapped in local extremum. The experimental results show that the improved FCM algorithm is more stable and accuracy is higher.3. We introduce concentration adjustment mechanism into the basic clonal selection algorithm, to ensure that the population not only continues to develop into benign characteristics, but also avoids excessive simplification. Besides, combing the Gaussian variation with Cauchy variation, we propose a hybrid self-adaptive mutation which is called'Gauss-Cauchy'hybrid self-adaptive mutation., which can dynamically adjust mutation step to effectively prevent algorithm from trapping in local optimal solution and improve the global optimization ability. Finally, In order to make the algorithm turn to the healthy development, we should make full use of immune memory mechanism so that excellent antibodies are kept and poor antibodies are replaced. The experimental results show that the global optimization ability and the convergence speed improve a lot.4. With improved filtering algorithm, we use the improved clonal selection algorithm to optimize the improved FCM algorithm. Through the test, compared with the classical algorithm, the new algorithm in this paper is more effective. The segmentation precision is improved significantly in anti-noise ability, convergence speed and the global optimization.