A Computerized System for Detection of Spiculated Margins based on Mammography

 Qaisar Abbas¹, Irene Fondo´n ² and Emre Celebi³

¹College of Computer and Information Sciences, Al Imam Mohammad Ibn Saud Islamic University, Saudi Arabia

²Department of Signal Theory and Communications, School of Engineering Path of Discovery, Spain

³Department of Computer Science, Louisiana State University, USA

 Abstract: Spiculated margins indicate a high risk of malignancy for breast cancer. Detection accuracy of current computerized diagnostic systems Computer-Aided Detections (CADs) for spiculated margins is not high due to the existence of intensity heterogeneities, often subtle and varied in appearance. This paper presents an automatic system for Accurately Detection Of Spiculated Margins (ADSM) by measuring its physical properties. In proposed system, a pre-processing step is performed to suppress background noise and enhance contrast. Spiculated margins are then segmented by a Maximum Fuzzy Entropy Partitioning (MFEP) algorithm whose parameters are optimized using the Quantum Genetic Algorithm (QGA). Afterwards, the characterization of spicule regions is completed using morphological operators, Steerable-Ridge-Filtering (SRF) and quantification of physical properties. A data set of 220 mammogram masses was used to evaluate the proposed system. Experimental results indicate that the ADSM system achieves a high accuracy level of Area Under the receiver operating characteristics Curve (AUC): 0.875 compared to state-of-art systems. By integrating the ADSM system, the performance of CADs could potentially be improved.

 Keywords: CAD, spiculated mass segmentation, image enhancement, fuzzy entropy, QGA, SRF.

Received May 13, 2013; accepted July 21, 2013