Mammogram Screening by Automated Followup: A Feasibility Study
Abstract
This paper describes part of a study aimed at developing a computer-based aid for mammogram screening that makes a detailed comparison between mammograms of the same patient acquired at different screenings and detects changes indicative of cancer. The focus is on determining control points in two mammograms; these points are used to put two mammograms into correspondence. The proposed procedure starts by extracting candidate points in each mammogram at the locations where elongated anatomical structures, such as ducts and blood vessels, intersect. This is achieved by using two multi-scale directional operators and logical operations applied to their combined outputs. The correspondence between points in a mammogram pair is established using an accumulator matrix and similarity criterion applied to each candidate pair. Two similarity criteria are considered. The first measures similarity between the geometry of intersecting elongated structures. The second measures similarity between texture patterns around candidate points. The methods were evaluated on synthetic images and archived film. Evaluation on the synthetic images is summarized in the form of ROC curves. Performance on real mammograms was evaluated by visual inspection involving two untrained observers and an experienced radiologist. A high degree of agreement was found between the observers and the algorithm, and in particular between the radiologist and the algorithm (the agreement was 91%).
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 1997
- Accession Number
- ADA329000
Entities
People
- Dragana Brzakovic
Organizations
- Lehigh University