A Computer Aided Detection System for Mammograms based on Asymmetry and Feature Extraction Techniques

M. Jirari (USA)


Computer Aided Detection, Mammograms, Feature Ex traction Techniques, Registration, Linear Subtraction.


An intelligent Computer Aided Detection system (CAD) was developed based on linear subtraction and fea ture extraction techniques to identify asymmetries between left and right breast mammograms. This system is based on the idea that a deviation from the normal architectural symmetry of the right and left breasts could indicate a can cerous mass. The mammograms were first normalized, reg istered, and then linear subtraction was followed by various feature extraction techniques that were used in order to re duce the number of false positive detections. These tech niques include preprocessing by morphological filtering, a size test and a border test. The system was tested on 102 pairs of mammograms, yielding a true positive rate of 93% with an average number of 1.26 false positive per image. Computerized detection was evaluated by using Receiver Operating Characteristic analysis (ROC), which measured the overall sensitivity of the presented system by the area under the curve Az to be 0.95. The results show that the system could be helpful to the radiologist by serving as a second reader in mammography screening.

Important Links:

Go Back