A Novel Fuzzy Topological Approach to the Detection of Mammographic Lesion and Quantification of Parenchymal Density
Abstract
This research focuses on mammographic image processing for the purpose of density quantification, lesion detection and classification. The approaches proposed are different from those taken in the literature in two respects: (1) They emphasize on identifying the dense regions and analyzing their parenchymal architecture. (2) They use a novel fuzzy connectedness method of object definition and image segmentation. During this report period, the following have been accomplished. (1)120 mammograms from our hospital database have been digitized, converted to the 3DVIEWNIX format, and stored on a medium. More data are being gathered in this fashion. (2) A scale-based fuzzy affinity relation has been devised that is suitable for mammographic image segmentation within the fuzzy connectedness framework. (3) An automatic method has been developed for the segmentation and quantification of parenchymal density. It shows excellent correlation for the measures obtained for the same patient from CC and MLO images. (4) A method of characterizing abnormal parenchymal architecture has been developed. Its utility in detecting and classifying lesions is being investigated.
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 01, 1998
- Accession Number
- ADB241926
Entities
People
- Jayaram K. Udupa
Organizations
- University of Pennsylvania