Tumour Demarcation by Using Vector Quantization and Clubbing Clusters of Ultrasound Image of Breast

In most of the computer aided diagnosis, segmentation is used as the preliminary stage and further can be helpful in quantitative analysis. Ultrasound imaging (US) helps medical experts to understand clinical problem efficiently with low cost as compared to its counterparts. In this paper, vector quantization based clustering technique has been proposed to detect the tumour (malignant or benign) of the breast Ultrasound Images. Presence of artefacts like speckle, shadow, attenuation and signal dropout, makes image understanding and segmentation difficult for an expert. Here, we dealt with images having these artefacts and proposed fully automatic segmentation technique using clustering. Firstly well known Vector Quantization based LBG technique is used for clustering and eight clusters are obtained, sequential clubbing of these cluster are suggested to obtain segmentation results. Improvement is suggested using two new techniques over LBG to form clusters, known as KPE (Kekre’s Proportionate Error), and KEVR (Kekre’s Error Vector Rotation), further same method of sequential clubbing of clusters is followed here as that of LBG and their results are compared.