Authors : Norhene Gargouri Ben Ayed, Alima Damak Masmoudi and Dorra Sellami Masmoudi
Abstract: This study presents a completely automated Computer-Aided Diagnostic (CAD) System for mass detection, segmentation and classification. This system performs mass detection followed by the classification as benign-malignant on the detected and segmented masses. In order to make mass detection more effective, a sequence of preprocessing steps are designed for contrast enhancement and noise effects removal as well as the effectiveness of the stage of detection. The location of suspicious masses using a new approach named Improved Against Noise Gray Level and Local Difference (IANGLLD) is developed for mass texture extraction. As the shapes of masses are fundamental in the classification between benignancy and malignancy, two shape features are used and joined with the texture features applied in mass detection to be the input of the ANN for mass classification. For the evaluation of the proposed system the Digital Database for Screening Mammography (DDSM) was applied to evaluate the performance. The obtained results are encouraging and have revealed promise of the proposed system.
Norhene Gargouri Ben Ayed, Alima Damak Masmoudi and Dorra Sellami Masmoudi, 2014. A New Automated CAD System for Classification of Malignant and Benign Lesions. Asian Journal of Information Technology, 13: 477-484.