Magnetic Resonant Image segmentation using trained K-means clustering

Magnetic Resonant Image segmentation is an indispensable process in the visualization of human tissues, particularly during clinical analysis. In this paper, we describe a method for segmentation of White matter and Gray matter from real MR images using a LM-k-means technique. After preprocessing, a simple unsupervised clustering system like k-means is taken and made into a supervised system by using Levenberg-Marquardt optimization technique. It was inferred that a k-means system does not arrive on its own at the means which will give a good segmentation. Hence the LM algorithm trains it for that purpose. The results are compared with that of a k-means system and they show a considerable improvement with a much higher precision.