An Improved SKFCM-CV Whole Heart MR Image Segmentation Algorithm

Cardiac magnetic resonance imaging (MRI) is the only effective method of inspection for some serious heart disease, such as the observation of atrial fibrillation, the evaluation of cardiac iron deposition, the diagnosis and detection of congenital heart and so on. Therefore, the whole heart segmentation of MR image is very important for medicine. In this paper: (1) A new FCM algorithm (SKFCM) is proposed based on spatial neighborhood correlation information of pixel and kernel function. Two correlation factors R ij S and R ij G are introduced to make up for the lack of original algorithm for spatial information. (2) A kernel constrained CV algorithm based on entropy and edge guidance function is proposed, which solves the problem that the fixed energy weight coefficient has poor universality for different images. And propose a simple kernel function instead of Gaussian kernel function to improve the efficiency. (3) In order to improve the accuracy and efficiency of the entropy value, the entropy weight coefficient is proposed to solve the problem that the image evolution speed is too slow due to the low entropy value of the image after SKFCM. Experiments on open data sets show that the proposed algorithm is suitable for most medical images with blur, less contrast or more noise, and the segmentation accuracy and efficiency are greatly improved.