Three-Class Mammogram Classification Based on Descriptive CNN Features.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Biomedical Engineering

Background:

• Mammography is crucial for breast cancer screening.
• Accurate classification of mammograms is essential for timely diagnosis and treatment.
• Existing methods face challenges in handling large datasets and subtle image variations.

Purpose of the Study:

• To propose novel deep learning techniques for classifying large mammogram datasets.
• To compare the effectiveness of two distinct feature extraction and classification approaches.
• To improve the accuracy of identifying normal, benign, and malignant breast conditions.

Main Methods:

• Developed two deep learning models: Convolutional Neural Network-Discrete Wavelet (CNN-DW) and Convolutional Neural Network-Curvelet Transform (CNN-CT).
• Utilized Contrast Limited Adaptive Histogram Equalization (CLAHE) for image enhancement.
• Employed Dense Scale Invariant Feature (DSIFT) extraction on image subbands generated by 2D-DWT and Discrete Curvelet Transform (DCT).
• Trained Convolutional Neural Networks (CNNs) using Softmax and Support Vector Machine (SVM) layers.

Main Results:

• The CNN-CT method achieved an accuracy rate of 83.74%.
• The CNN-DW method achieved an accuracy rate of 81.83%.
• Both proposed methods demonstrated significant improvements over existing techniques in validation measures.

Conclusions:

• The proposed CNN-DW and CNN-CT models offer effective solutions for large-scale mammogram classification.
• The CNN-CT approach shows superior performance in distinguishing between normal, benign, and malignant cases.
• These deep learning techniques hold promise for enhancing the accuracy and efficiency of breast cancer diagnosis.