An efficient implementation of Fuzzy C-Means and watershed algorithms for MRI segmentation

Abstract

Image segmentation is one of the most common steps in digital image processing. It classifies a digital image into different segments. There are many algorithms for image segmentation such as thresholding, edge detection, and region growing, which finding a suitable algorithm for medical image segmentation is a challenging task. This is due to noise, low contrast, and steep light variations of medical images. The main goal of this paper is improving the performance of fuzzy c-means clustering. Improving is achieved using parallel implementation of this algorithm. Fuzzy c-means clustering is an important iterative clustering algorithm, but it is computationally intensive and it uses the same data between the iterations. The center of the clusters changes in each iteration, which requires considerable amount of time for large data sets. The parallel fuzzy c-means clustering is implemented by using task pipeline concept in CUDA technology. The experimental results show that the performance is improved up to 23.35×. After that watershed algorithm is applied for the final segmentation. The implementation results show that the accuracy of diagnosis in magnetic resonance imaging 97/33% is improved. This improvement is achieved using enhancing edges and reducing noises in images.