Computer-aided diagnosis for early-stage breast cancer by using Wavelet Transform

Area of Science:

• Medical Imaging
• Computational Biology
• Biomedical Engineering

Background:

• Early detection of breast cancer is crucial for improving patient outcomes.
• Micro-calcifications are key indicators of early-stage breast cancer.
• Existing diagnostic methods can be limited in sensitivity and specificity.

Purpose of the Study:

• To develop a high-sensitivity computer-aided diagnosis (CAD) algorithm.
• To accurately detect and quantify micro-calcifications for early breast cancer diagnosis.
• To enhance the diagnostic performance compared to traditional classifiers.

Main Methods:

• The algorithm employs a two-phase approach: image reconstruction and micro-calcification region recognition.
• Phase I utilizes wavelet layers and Renyi's information theory for region separation, followed by Morphology-Dilation and Majority Voting Rule for reconstruction.
• Phase II uses 49 descriptors, including shape and texture features (e.g., Grey-Level Co-occurrence Matrix), reduced via Principal Component Analysis (PCA).

Main Results:

• The algorithm effectively separates and reconstructs suspicious micro-calcification regions.
• Principal Component Analysis (PCA) significantly reduces computational load while retaining descriptor efficiency.
• The Back-propagation Neural Network classifier demonstrated superior performance with a high true positive rate and low false positive rate.

Conclusions:

• The proposed computer-aided diagnosis algorithm shows significant promise for early breast cancer detection.
• The integration of advanced image processing and machine learning techniques enhances diagnostic accuracy.
• The algorithm's performance, validated on clinical data, suggests its potential for clinical application.