Related Experiment Videos
A multiple circular path convolution neural network system for detection of mammographic masses.
Shih-Chung B Lo1, Huai Li, Yue Wang
1Center for Imaging Science and Information System, Radiology Department, Georgetown University Medical Center, Washington, DC 20007, USA. lo@isis.imac.georgetown.edu
IEEE Transactions on Medical Imaging
|April 4, 2002
Summary
A novel multiple circular path convolution neural network (MCPCNN) improved mammographic mass detection. This AI approach enhanced analysis of tumor structures, outperforming traditional methods for identifying breast cancer.
Area of Science:
- Artificial Intelligence in Medical Imaging
- Machine Learning for Breast Cancer Detection
- Deep Learning Architectures for Radiology
Background:
- Accurate detection of mammographic masses is crucial for early breast cancer diagnosis.
- Conventional feed-forward neural networks have limitations in analyzing complex spatial features of tumors.
- There is a need for advanced computational models to improve the sensitivity and specificity of mammogram analysis.
Purpose of the Study:
- To introduce and evaluate a novel Multiple Circular Path Convolution Neural Network (MCPCNN) architecture.
- To assess the performance of MCPCNN in detecting tumor and tumor-like structures in mammograms.
- To compare the efficacy of MCPCNN against conventional feed-forward neural networks for mammographic mass detection.
Main Methods:
- Development of an MCPCNN architecture tailored for analyzing segmented tumor regions.
- Division of suspected tumor areas into sectors, computing mass features for each sector independently.
- Processing mammograms using dual morphological enhancement, region growing for delineation, and calculating 144 Breast Imaging-Reporting and Data System-based features across 36 sectors.
Main Results:
- The MCPCNN achieved improved area under the receiver operating characteristic curve (Az) values ranging from 0.84 to 0.89.
- This represents a significant performance enhancement compared to conventional feed-forward neural networks, which yielded Az values of 0.78-0.80.
- The MCPCNN demonstrated a marked improvement in the detection accuracy of mammographic masses.
Conclusions:
- The MCPCNN architecture offers a potentially superior method for analyzing defined mass features in mammography.
- This novel neural network structure shows promise for enhancing the accuracy of computer-aided detection systems for breast cancer.
- Further research may explore the broader applicability of MCPCNN in medical image analysis.