Modified fuzzy C-means clustering algorithm application in medical image segmentation

Abstract EN

Developing effective algorithm for segmenting image is very important in pattern recognition, medical magnetic resonance image, X-ray images analysis, and in computer vision.

Fuzzy c-means (FCM) is one of the mostly used methodologies in clustering image for segmentation.

However, the results of the standard and the modified version FCM are not always satisfactory.

This paper introduces a spatial FCM that considers the weighted fuzzy effect of neighboring pixels on the cluster center depending on the location and intensity (kernel metric).

The objective function in FCM algorithm is modified to minimize the intensity in homogeneities, by implicating the spatial neighborhood information and modifying the membership weighting of each cluster.

The advantages of the new FCM algorithm are: (a) Produces homogeneous regions more than FCM algorithm, (b) handles noisy spots, and (c) it is relatively less sensitive to noise.

The integrated clustering techniques have produced tremendous output images with the minimal process to separate the objects from the background.

Experimental results on real images show that the algorithm is effective, efficient, and is relatively independent of the type of noise.

Especially, it can process non-noisy and noisy images without knowing the type of the noise.