Computer-aided diagnosis for breast cancer classification using deep neural networks and transfer learning

Area of Science:

• Medical Image Analysis
• Artificial Intelligence in Healthcare
• Oncology

Background:

• Breast cancer is a leading cause of death in women globally, often due to late diagnosis.
• Early detection and accurate classification are crucial for effective breast cancer treatment.
• Computer-aided diagnosis (CAD) systems, particularly those using deep learning, can enhance the speed and accuracy of cancer detection.

Purpose of the Study:

• To develop and evaluate a novel computer-aided diagnosis (CAD) method for breast cancer classification.
• To assess the performance of deep neural networks (ResNet 18, ShuffleNet, Inception-V3Net) combined with transfer learning for breast cancer detection.
• To achieve high accuracy in both binary (benign vs. malignant) and multi-class breast cancer classification.

Main Methods:

• Utilized the publicly available BreakHis dataset for breast cancer image analysis.
• Employed a combination of deep neural network architectures: ResNet 18, ShuffleNet, and Inception-V3Net.
• Applied transfer learning techniques to train the deep neural networks for classification tasks.

Main Results:

• Achieved high average accuracy for binary classification: 99.7% (ResNet), 97.66% (InceptionV3Net), and 96.94% (ShuffleNet).
• Demonstrated strong average accuracy for multi-class classification: 97.81% (ResNet), 96.07% (InceptionV3Net), and 95.79% (ShuffleNet).
• The proposed method shows significant potential for accurate and automated breast cancer classification.

Conclusions:

• The developed CAD system effectively classifies breast cancer using deep learning models.
• The ResNet 18 architecture demonstrated superior performance in both binary and multi-class classification tasks.
• This approach offers a promising tool for aiding physicians in the early and accurate diagnosis of breast cancer.