Authors:
Farhang Sahba
and
Anastasios Venetsanopoulos
Affiliation:
Ryerson University, Canada
Keyword(s):
Mammography images, Mass detection, Mass segmentation, Bilateral filter, Mean shift, Computer-aided detection.
Related
Ontology
Subjects/Areas/Topics:
Biometrics and Pattern Recognition
;
Image and Video Processing, Compression and Segmentation
;
Multimedia
;
Multimedia Signal Processing
;
Telecommunications
Abstract:
This paper presents a new method for mass detection and segmentation in mammography images. The extraction of the breast border is the first step. A bilateral filter is then applied to the breast area to smooth the image while preserving the edges. Image pixels are subsequently clustered using an adaptive mean shift scheme that employs intensity information to extract a set of high density points in the feature space. Due to its non-parametric nature, adaptive mean shift algorithm can work effectively with non-convex regions resulting in suitable candidates for a reliable segmentation. The clustering is then followed by further stages involving mode fusion. An artificial neural network is also used to remove the false detected regions and recognize the real masses. The proposed method has been validated on standard database. The results show that this method detects and segments masses in mammography images effectively, making it useful for breast cancer detection
systems.