Detection of Breast Cancer Using Histopathological Image Classification Dataset with Deep Learning Techniques

Area of Science:

• Medical Imaging
• Computational Pathology
• Artificial Intelligence in Healthcare

Background:

• Breast cancer is a leading cause of mortality globally, particularly among women.
• Accurate diagnosis relies on imaging modalities and histopathology, but observer variability can impact results.
• Computer-aided diagnosis (CAD) systems aim to improve accuracy and reduce diagnostic discrepancies.

Purpose of the Study:

• To develop enhanced strategies for computer-aided diagnosis (CAD) phases in breast cancer detection.
• To minimize observer variability in image analysis for breast cancer diagnosis.
• To improve the performance of CAD systems through advanced segmentation and classification techniques.

Main Methods:

• Developed an automatic Fourier Transform-based segmentation approach with self-driven post-processing.
• Incorporated spatial information and eliminated the need for pre-set parameters in segmentation.
• Employed a classification strategy using weighted feature selection, an upgraded Genetic Algorithm, and a Convolutional Neural Network Classifier.

Main Results:

• The proposed segmentation technique is fast, magnification-independent, and automatically determines inputs for morphological operations.
• Extensive tests identified optimal feature extraction techniques (textural, morphological, graph characteristics) for classification accuracy.
• The integrated CAD framework demonstrated potential to reduce misdiagnoses and improve classification accuracy.

Conclusions:

• The enhanced segmentation and classification algorithms offer a significant improvement for CAD systems in breast cancer detection.
• The developed CAD framework can serve as a valuable second opinion tool for pathologists.
• This approach aids in the early and more accurate detection of breast cancer, improving patient outcomes.