Initialization of clustering algorithms for unsupervised segmentation of multi-echo MR images

Unsupervised segmentation is a key step towards the automatic analysis and understanding of magnetic resonance (MR) images. A number of techniques based on multi-dimensional data classification have been applied to this problem. Since most unsupervised classification approaches suffer from local traps, the segmentation often depends on the initialization of the classification algorithm used. In this paper, a method to deal with the initialization, especially of the class centres, is addressed. The method consists of two steps: firstly, finding class centre candidates through analyzing the 1D and multi-dimensional histograms of the MR images, and, secondly, selecting the required number of most possible class centres from these candidates under a certain criterion. Results obtained using actual dual-echo MR images (both the class centre candidates and segmentation of the images) have shown that the proposed method is able to find suitable class centres for classification algorithms, and hence consistent segmentation can be obtained.