Transfer learning based histopathologic image classification for breast cancer detection.

Area of Science:

• Medical Imaging
• Computer Vision
• Machine Learning

Background:

• Breast cancer is a leading cause of death in women globally, often due to late diagnosis.
• Early detection systems using medical imagery are crucial for improving patient outcomes.
• Deep learning, particularly Convolutional Neural Networks (CNNs), shows promise in image analysis but faces challenges with parameter tuning and weight initialization.

Purpose of the Study:

• To investigate the effectiveness of transfer learning and deep feature extraction for early breast cancer detection.
• To adapt pre-trained CNN models (AlexNet, Vgg16) for breast cancer image analysis.
• To compare the performance of transfer learning against traditional deep feature extraction methods.

Main Methods:

• Utilized AlexNet and Vgg16 models for deep feature extraction from histopathologic breast cancer images.
• Applied transfer learning by fine-tuning the AlexNet model.
• Classified extracted features using Support Vector Machines (SVM).
• Conducted experiments on a public histopathologic breast cancer dataset.

Main Results:

• Transfer learning demonstrated superior performance compared to deep feature extraction.
• The combination of transfer learning with SVM classification yielded high accuracy scores.
• The study successfully adapted pre-trained models for the breast cancer detection task.

Conclusions:

• Transfer learning is a viable and effective approach for enhancing early breast cancer detection systems.
• Deep learning models, when adapted through transfer learning, can significantly improve diagnostic accuracy.
• Further research in transfer learning can lead to more robust and efficient breast cancer screening tools.