Related Experiment Video
Updated: Jul 12, 2026
13:44
Detection of Architectural Distortion in Prior Mammograms via Analysis of Oriented Patterns
Published on: August 30, 2013
Fractional Active Contour Model with Adaptive Hessian Weighting for Mammogram Segmentation.
Ruhollah Motamedi1, Nasser Aghazadeh1,2,3, Mahdi Hashemzadeh4,5
1Department of Mathematics, Azarbaijan Shahid Madani University, Tabriz, Iran.
Journal of Medical Signals and Sensors
|July 11, 2026
Summary
This study introduces a new active contour model for segmenting breast masses in mammograms. The method enhances image details and uses curvature information to accurately detect lesions, improving computer-aided diagnosis.
Area of Science:
- Medical Imaging
- Computer-Aided Diagnosis
- Image Segmentation
Background:
- Accurate segmentation of breast masses in mammograms is vital for computer-aided diagnosis (CAD).
- Mammograms present segmentation challenges due to noise, low contrast, weak boundaries, and intensity inhomogeneity.
- These factors significantly impede reliable mass segmentation in clinical practice.
Purpose of the Study:
- To propose a novel variational active contour model for robust mammographic mass segmentation.
- To integrate spatially adaptive fractional-order enhancement with Hessian-based curvature weighting for improved accuracy.
- To address intensity inhomogeneity using a local region-fitting energy term within a level-set framework.
Main Methods:
- A spatially adaptive fractional-order map enhances image texture and weak edges based on image gradients.
- Hessian matrix analysis extracts second-order structural information for an adaptive edge-stopping function sensitive to local curvature.
- A local region-fitting energy term is incorporated to handle intensity inhomogeneity, implemented within a numerically stable level-set framework.
Main Results:
- The proposed model was evaluated on the INbreast and CBIS-DDSM mammogram datasets.
- Quantitative analysis using Dice Similarity Coefficient (DSC) and Hausdorff Distance showed superior performance over classical active contour models.
- The method achieved competitive results against learning-based approaches, especially for low-contrast masses and irregular boundaries.
Conclusions:
- The developed fractional active contour model offers a robust, unsupervised, and interpretable solution for mammographic mass segmentation.
- The model effectively overcomes common challenges in mammograms by combining adaptive fractional enhancement, Hessian-based curvature, and local intensity fitting.
- This approach shows promise as a valuable tool for enhancing breast cancer detection in CAD systems.
