Related Experiment Video
Updated: Jun 17, 2025
13:44
Detection of Architectural Distortion in Prior Mammograms via Analysis of Oriented Patterns
Published on: August 30, 2013
42.8K
Hybrid Feature Mammogram Analysis: Detecting and Localizing Microcalcifications Combining Gabor, Prewitt, GLCM
Miguel Alejandro Hernández-Vázquez1, Yazmín Mariela Hernández-Rodríguez1, Fausto David Cortes-Rojas2
1Departamento de Tecnologías Avanzadas, UPIITA-Instituto Politécnico Nacional, Av. Instituto Politécnico Nacional 2580, Ciudad de México 07340, Mexico.
Diagnostics (Basel, Switzerland)
|August 10, 2024
Summary
This study introduces a hybrid approach using feature extraction and Convolutional Neural Networks (CNNs) to improve microcalcification detection in mammograms. The method enhances early breast cancer diagnosis by achieving high accuracy and sensitivity.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Oncology
Background:
- Breast cancer is a leading cause of mortality, with early detection crucial for improved patient outcomes.
- Microcalcifications on mammograms are key indicators of early-stage breast cancer, but their identification can be challenging.
- Current detection methods face limitations in accurately identifying subtle microcalcifications.
Purpose of the Study:
- To develop and evaluate a hybrid feature extraction and Convolutional Neural Network (CNN) model for enhanced detection and localization of microcalcifications in mammograms.
- To improve the accuracy and sensitivity of microcalcification identification compared to existing methods.
Main Methods:
- A hybrid feature extraction technique combining Gabor, Prewitt, and Gray Level Co-occurrence Matrix (GLCM) kernels was employed.
- A CNN architecture with convolutional, ReLU, and maxpooling layers was designed for classification.
- Preprocessing involved contrast enhancement (VOI LUT) and region segmentation; Top Hat filter used for localization.
Main Results:
- The proposed method achieved an accuracy of 89.56%, sensitivity of 82.14%, and specificity of 91.47%.
- Performance surpassed related works, which typically report accuracies around 85-87% and sensitivities of 76-81%.
- The system demonstrated effective detection and precise localization of microcalcifications.
Conclusions:
- Combining traditional feature extraction with deep learning (CNNs) significantly enhances microcalcification detection and localization.
- This AI-powered system shows promise as an auxiliary tool for radiologists, improving early breast cancer detection and reducing diagnostic errors.
