Title:	ESTIMATION OF A LOG-COMPRESSED K-DISTRIBUTION AND APPLICATION TO BREAST TUMOR SEGMENTATION AND DISCRIMINATION IN ULTRASONIC IMAGES
Author(s):	Atsushi Takemura
ISBN:	978-972-8924-84-3
Editors:	Yingcai Xiao, Tomaz Amon and Piet Kommers
Year:	2009
Edition:	Single
Keywords:	Ultrasonic image, breast tumor, log-compressed K-distribution, AdaBoost, active contour model
Type:	Short Paper
First Page:	266
Last Page:	271
Language:	English
Cover:
Full Contents:	click to dowload
Paper Abstract:	This paper proposes a new method for segmentation and discrimination of breast tumor in ultrasonic images. To achieve accurate segmentation of breast tumors, the segmentation process uses an adaptive speckle suppression filter and a modified active contour model based on estimation of a log-compressed K-distribution. This paper also describes effective features and classifiers for accurate discrimination of breast tumors. I defined 208 features corresponding to diagnostic observations used by medical doctors, such as internal echo, shape, and boundary echo. These features included novel features based on the estimated parameter of the log-compressed K-distribution. Furthermore, this paper proposes a method for discrimination of breast tumors using an ensemble classifier trained by the multi-class AdaBoost learning algorithm (AdaBoost.M2) combined with a sequential feature selection process. Validation testing using 200 carcinomas, 50 fibroadenomas, and 50 cysts showed the high performance of the proposed method of segmentation and discrimination.
	Go Back