Segmentation Techniques: A Comparison and Evaluation on MR Images for Brain Tumour Detection

Abstract

Brain tumour is inherently serious and life-threatening because of its character in the constrained space of the intracranial cavity (space formed inside the skull). If the tumour is detected at an advanced stage it turns to be a grave medical problem. Various techniques were developed for the detection of brain tumour. The image segmentation technique plays a pivotal role in early tumour detection. The segmentation is the process that partitions an image into regions. The widely used common image segmentation techniques are edge detection and clustering techniques. Edges cause significant local changes in the image intensity and have been an important feature for analysing images. It is the first step in receiving information from images. The techniques discussed here are Gradient-based methods such as Roberts, Sobel, Prewitt, Canny operators and Laplacian based edge detection method such as Laplacian of Gaussian operator(LOG). Clustering is the method of grouping a set of patterns into a number of clusters. The two important clustering algorithms namely centroid based K-Means and representative object based Fuzzy C-Means (FCM) clustering algorithms are compared. This paper presents the qualitative comparison of edge detection and clustering techniques for brain tumour MRI images based on image quality parameters like PSNR (Peak Signal to Noise Ratio), MSE (Mean Square Error), RMSE (Root Mean Square Error) and computing time.

Keywords: Gradient-based, Laplacian, K-Means, Fuzzy C-Means, PSNR, MSE, RMSE, Computing time.