TY  - JOUR
T1  - Denoising and Automatic Detection of Breast Tumor in Ultrasound Images
AU - Prabhakar, Telagarapu AU - Poonguzhali, S. 
JO  - Asian Journal of Information Technology
VL  - 15
IS  - 18
SP  - 3506
EP  - 3512
PY  - 2016
DA  - 2001/08/19
SN  - 1682-3915
DO  - ajit.2016.3506.3512
UR  - https://makhillpublications.co/view-article.php?doi=ajit.2016.3506.3512
KW  - Breast ultrasound image
KW  -speckle noise
KW  -texture analysis
KW  -image segmentation
KW  -active contour
KW  -tetrolet transform
AB  - Over the past three decades, Breast cancer has been the leading cause of death. Detection of breast
cancer at the early stage is a critical procedure. So far Mammography is used for screening and detection, but
this method is actually found to be uncomfortable among young woman. Whereas ultrasound can be the best
replacement for mammography as imaging of human organs and soft tissue can be done much more easily in
ultrasound without much pain and it is cost effective as well. In the ultrasound, the only drawback is its poor
quality which is affected by speckle noise which in turn makes the segmentation and classification of interested
lesion problematic. Usually, active contour segmentation technique is used which is proved to be ineffective
when we go for automatic detection and more over it usually causes improper segmentation and classification.
So in order to escape improper segmentation and classification we have developed a scheme which capable to
locate region of lesions automatically. This method involves Tetrolet Transform speckle reduction method
followed by statistical features of the lesion region and K-Nearest Neighbor (KNN) classifier. This technique
is tested over 110 lesion images of breast. The accuracy of this method is around 91.51% and Sensitivity is
around 94.42%. The Dice similarity which is found to be 91.27% is obtained between segmented ROIs and
ground truth images. Hence, the automatic segmentation of lesion region is made possible. This method will
help the radiologist to detect the lesion boundary automatically.
ER  -