Related Experiment Video
Updated: Apr 1, 2026
13:44
Detection of Architectural Distortion in Prior Mammograms via Analysis of Oriented Patterns
Published on: August 30, 2013
43.9K
A two-stage method for microcalcification cluster segmentation in mammography by deformable models.
N Arikidis1, K Vassiou2, A Kazantzi1
1Department of Medical Physics, School of Medicine, University of Patras, Patras 26504, Greece.
Medical Physics
|October 3, 2015
Summary
This study presents a reliable semiautomated method for segmenting microcalcification clusters in mammography, improving accuracy for computer-aided diagnosis. The new approach significantly outperforms existing methods, aiding radiologists in quantitative image analysis.
Area of Science:
- Medical Imaging
- Computer-Aided Diagnosis
- Image Segmentation
Background:
- Accurate segmentation of microcalcification (MC) clusters in mammography is crucial for quantitative analysis and computer-aided diagnosis.
- Current segmentation methods face challenges in precision and efficiency for radiologists.
Purpose of the Study:
- To investigate a two-stage semiautomated segmentation method for microcalcification (MC) clusters in mammography.
- To evaluate the reliability and accuracy of the proposed segmentation method compared to existing techniques.
Main Methods:
- A two-stage semiautomated approach combining level set and active contour models within a wavelet transform scale-space.
- Evaluation of segmentation reliability using inter/intraobserver agreements and quantitative metrics (Hausdorff distance, average minimum distance, area overlap measure).
- Assessment of the method's impact on MC cluster characterization accuracy using feature extraction and support vector machine classification.
Main Results:
- Substantial interobserver and intraobserver agreements were achieved for distance-based metrics, with moderate agreement for area overlap.
- The proposed semiautomated method demonstrated statistically significant superior performance (Az=0.80 ± 0.04) compared to the B-spline active rays method (Az=0.69 ± 0.04).
- The method shows promise for improving MC cluster characterization accuracy in computer-aided diagnosis.
Conclusions:
- Deformable models offer a reliable semiautomated segmentation method for microcalcification (MC) clusters.
- The proposed method can be effectively utilized for quantitative image analysis of MC clusters in mammography.
