Analysis of Interval Changes on Mammograms for Computer Aided Diagnosis
Abstract
A multistage regional registration technique (MRRT) was developed for identifying masses on temporal pairs of mammograms. It was investigated the use of the density-weighted contrast enhancement (DWCE) technique to improve the localization of the corresponding mass on the prior mammogram. 179 temporal pairs of mammograms containing biopsy-proven masses were used for evaluation. 84% of the estimated lesion locations resulted in an area overlap of at least 50% with the true lesion locations. The average distance between the estimated and the true centroid of the lesions on the prior mammogram was 4.816.9 mm. The registration accuracy was improved in comparison with the registration without DWCE. Regions of interest containing the corresponding masses were identified on the current and prior mammograms of the temporal pair. The masses were automatically segmented using a K-means clustering algorithm and active contour model. Texture, spiculation and morphological features were extracted from each mass. An additional difference features were obtained by subtracting the features of the prior mass from those of the current mass. The feature space for each temporal pair consisted of the texture, spiculation and morphological features from both the prior and the current mammograms and the difference features. MRRT can be useful for identification of corresponding lesions on temporal pairs of mammograms. The obtained features will be used for classification of malignant and benign temporal masses as well as detection of temporal change.
Document Details
- Document Type
- Technical Report
- Publication Date
- May 01, 2001
- Accession Number
- ADA392312
Entities
People
- Lubomir Hadjiiski
Organizations
- University of Michigan