Related Experiment Video
Updated: Jun 27, 2026
13:44
Detection of Architectural Distortion in Prior Mammograms via Analysis of Oriented Patterns
Published on: August 30, 2013
B-spline active rays segmentation of microcalcifications in mammography
Nikolaos S Arikidis1, Spyros Skiadopoulos, Anna Karahaliou
1Department of Medical Physics, School of Medicine, University of Patras, 265 00 Patras, Greece.
Medical Physics
|December 17, 2008
Summary
A new active ray segmentation method accurately identifies microcalcifications in mammograms, improving computer-aided diagnosis. This novel approach enhances the classification of benign versus malignant clusters by better characterizing microcalcification morphology.
Area of Science:
- Medical Imaging
- Computer-Aided Diagnosis
- Biomedical Engineering
Background:
- Accurate microcalcification segmentation in mammography is vital for quantifying morphologic properties.
- Existing computer-aided diagnosis (CAD) schemes require precise segmentation for feature extraction.
- Pleomorphic microcalcifications pose segmentation challenges in mammographic analysis.
Purpose of the Study:
- To propose and evaluate a novel active ray segmentation method for microcalcifications in mammography.
- To compare the proposed method's accuracy against a radial gradient-based approach.
- To assess the impact of segmentation accuracy on the classification of benign versus malignant microcalcification clusters.
Main Methods:
- Implementation of active rays (polar-transformed active contours) on B-spline wavelet representation for coarse-to-fine segmentation.
- Utilizing an iterative region growing method constrained by contour point estimates for final delineation.
- Comparative analysis with a radial gradient-based segmentation method on a DDSM database dataset.
- Evaluation of segmentation accuracy by three radiologists and assessment of classification performance using area under ROC curve (Az).
Main Results:
- The proposed active ray method achieved significantly higher radiologist accuracy ratings (average 3.83-3.97) compared to the radial gradient method (average 2.10-2.91).
- Segmentation by the proposed method led to statistically significant improvements in patient-based classification performance (Az: 0.82-0.86) versus the radial gradient method (Az: 0.71-0.75).
- Morphologic features (area, length) extracted using the proposed method enhanced the characterization of microcalcification clusters.
Conclusions:
- The novel active ray segmentation method demonstrates superior accuracy, meeting human visual criteria.
- Improved segmentation accuracy directly enhances the ability of morphologic features to differentiate between benign and malignant microcalcification clusters.
- This method holds promise for advancing computer-aided diagnosis in mammography by providing more reliable microcalcification characterization.
