Breast Cancer Detection and Analytics Using Hybrid CNN and Extreme Learning Machine

Area of Science:

• Medical Imaging
• Artificial Intelligence in Healthcare
• Biomedical Engineering

Background:

• Early breast cancer detection significantly improves survival rates globally.
• Mammography is a key diagnostic tool, but challenges remain in image analysis algorithms.
• Computer-aided diagnosis (CAD) systems are crucial for enhancing diagnostic accuracy.

Purpose of the Study:

• To develop and evaluate a hybrid computer-aided diagnosis (CAD) model for enhanced breast cancer detection, segmentation, and classification.
• To improve the accuracy and efficiency of breast cancer diagnosis using advanced machine learning techniques.
• To address the research challenges in selecting appropriate algorithms for mammogram analysis.

Main Methods:

• A hybrid model combining Convolutional Neural Networks (CNN) with a pruned ensembled Extreme Learning Machine (HCPELM) was developed.
• The model utilizes the rectified linear unit (ReLU) activation function for enhanced data analytics and artifact removal.
• Transfer learning techniques were employed by freezing specific layers and modifying the architecture to reduce parameters for efficient cancer detection.

Main Results:

• The hybrid HCPELM model achieved an 86% breast image recognition accuracy on the MIAS database.
• The proposed model demonstrated superior performance compared to benchmark deep learning models.
• The system effectively performed image enhancement, segmentation, feature extraction, and classification.

Conclusions:

• The HCPELM hybrid classifier shows superior performance in early breast cancer detection and diagnosis.
• This advanced CAD system can significantly aid healthcare practitioners in diagnosing breast cancer.
• The developed model offers a promising solution for improving mammogram analysis and patient outcomes.