The Effect of Mammogram Databases on Algorithm Performance
CBMS '04 Proceedings of the 17th IEEE Symposium on Computer-Based Medical Systems
Digital Image Processing (3rd Edition)
Digital Image Processing (3rd Edition)
The effectiveness of self-organizing maps in mammographic image analysis
The effectiveness of self-organizing maps in mammographic image analysis
A region-based active contour method for extraction of breast skin-line in mammograms
ISBI'10 Proceedings of the 2010 IEEE international conference on Biomedical imaging: from nano to Macro
| Hi-index | 0.00 |
Identification of the skin-air interface is a challenging, yet essential task in mammographic image analysis. Two strategies for determining this boundary are presented in this study. The stroma edge is initially located using a K-clustered self-organizing map and morphological operations. The first procedure then applies a thresholding technique near the stroma edge to identify the skin line, while the second supposes that the distance between the stroma edge and the skin line is approximately equal throughout the image. Applied to 500 mammograms from the Digital Database for Screening Mammography (DDSM), qualitative findings indicate that the mammographic projection determined which approach was most effective. Specifically, medio-lateral oblique (MLO) view mammograms favored the first approach and the cranio-caudal (CC) view favored the second, regardless of mammographic density. To confirm the viability of this method as a pre-processing step for computer-aided diagnosis (CAD), the resulting segmentation was compared to the location of abnormalities when present. In this data set, 218 mammograms contained abnormalities, annotated by a radiologist. Four of these were technically suboptimal, while 214 (98.17%) were completely encompassed by the segmentation.