Title:	SEGMENTATION AND DISCRIMINATION OF BREAST TUMORS IN ULTRASONIC IMAGES USING AN ENSEMBLE CLASSIFIER AND APPLICATION TO A DIAGNOSTIC SUPPORT SYSTEM
Author(s):	Atsushi Takemura
ISBN:	978-972-8939-48-9
Editors:	Yingcai Xiao
Year:	2011
Edition:	Single
Keywords:	Ultrasonic image, breast tumor, Nakagami distribution, AdaBoost, Marcov random field, geodesic active contour
Type:	Short Paper
First Page:	207
Last Page:	211
Language:	English
Cover:
Full Contents:	click to dowload
Paper Abstract:	This study proposes a novel diagnostic support system for automated diagnosis of breast tumors in ultrasonic images. The proposed system performs breast tumor segmentation and discrimination using the AdaBoost machine learning algorithm. In this study, novel features for segmentation and discrimination are defined on the basis of an estimated log-compressed Nakagami distribution parameter, which corresponds to physical characteristics of ultrasonic echoes from tumors. The segmentation process is performed by using an ensemble classifier trained by the AdaBoost algorithm with a Marcov random field and is followed by a geodesic active contour to increase the sophistication of the extracted tumor boundary. The process for breast tumor discrimination is determined using the multi-class AbaBoost algorithm. The performance of the proposed diagnostic system was evaluated by 10-fold cross validation tests, where 300 carcinomas, 60 fibroadenomas, and 50 cysts were used.
	Go Back