Development of an automated method for detecting mammographic masses with a partial loss of region

Abstract

Abstract-Recently, we have been developing several automated algorithms for detecting masses on mammograms. For our algorithm, we devised an adaptive thresholding technique for detecting masses, but our system failed to detect masses with a partial loss of region that were located on the edge of the film. This is a common issue in all of the algorithms developed so far by other groups. In order to deal with this problem, we propose a new method in the present study. The partial loss masses are identified by their similarity to a sector-form model in the template matching process. To calculate the similarity, four features are applied: 1) average pixel value; 2) standard deviation of pixel values; 3) standard correlation coefficient defined by the sector-form model; and 4) concentration feature determined from the density gradient. After employing the new method to 335 digitized mammograms, the detection sensitivity for the partial loss masses jumped from 70% to 90% when the number of false positives was kept constant (0.2/image). Moreover, a combination of the existing method and the new method improved the true-positive rate up to 97%. Such results indicate that the new technique may improve the performance of our computeraided diagnosis system for mammographic masses effectively.Index Terms-Breast cancer, computer-aided diagnosis, mammography, mass detection, template matching.

Automatic Detection of Pectoral Muscle Using Average Gradient and Shape Based Feature

“…In the band division, algorithm tolerance δ 1 and δ 2 are introduced. Any small positive integer in the range [3][4][5][6][7][8][9][10][11][12] and [1][2][3][4][5] may be used for them. Small variations of these parameter values do not affect the performance.…”

“…Normally, in medio-lateral oblique (MLO) view of mammogram, pectoral muscle appears as a triangular, high-density region at the posterior corner of the image. The presence of pectoral muscle can affect the automatic detection of suspicious regions such as mass [3,4], or automatic identification of breast tissue density [5,6]; as the pectoral muscle approximately have the same density, so is the dense tissues of interest in the image.…”

Detection of Cancerous Masses in Mammograms by Template Matching: Optimization of Template Brightness Distribution by Means of Evolutionary Algorithm

“…For example, te Brake and Karssemeijer [15] extend the circular template to 1.3R, where R is the radius of the model of the lesion, and brightness of the ring surrounding the model is zero. A more general template using a sector-form model is considered by Hatanaka et al [16], who analyze detection of masses with a partial loss of region in the vicinity of film edge. Our basic assumption concerning the template is that it has a circular symmetry, so that the consideration of brightness distribution is reduced to the optimization along the radius.…”

“…The overall result is TPF=0.65 for FPI=2, which is approximately the same as obtained with the HT described below. The paper by Hatanaka et al [16] analyzes 335 mammograms from a proprietary database. The paper actually describes a system utilizing several methods of mass detection, one of which is template matching by means of the correlation coefficient.…”

Automatic Pectoral Muscle Segmentation on Mediolateral Oblique View Mammograms

“…Also the area overlying the pectoral muscle is a common area for cancers to develop and is particularly checked by radiologists to reduce false negatives. It is, therefore, necessary to segment out the pectoral muscle before lesion detection, as stated in [6]. Similarly, exclusion of the pectoral muscle is required for automatic breast tissue density quantification [7], [8].…”