Improved method of pectoral muscle detection based on two straight line fitting algorithm

Pectoral muscle in mediolateral mammogram view is one of the three landmarks for mammogram registration. Mammogram registration is an important step in detecting gradual changes that occur over time and may be missed by radiologists. Detection of pectoral muscle is a challenging process due to factors such as overlapping fibroglandular tissue especially for the lower part of the pectoral muscle. In our research, an algorithm that detects pectoral muscle is presented. First, two straight line fitting procedure is utilized to detect the initial location of pectoral muscle. Then, a curve fitting algorithm is used to curve the lines through the detection of cliff locations. Cliff location represents the edge of the pectoral muscle and it is detected using sigmoid operation. Our algorithm is compared with an algorithm that detects the pectoral muscle using one line only. The accuracy of the two straight line fitting algorithm using subjective evaluation is 63.7%, which is a notable increase over the performance of one straight line fitting algorithm of 36.3% only.