Explainable AI-enabled hybrid deep learning architecture for breast cancer detection

Area of Science:

• Medical Imaging Analysis
• Artificial Intelligence in Healthcare
• Oncology Diagnostics

Background:

• Breast cancer is a leading cause of mortality in women, necessitating advanced diagnostic tools.
• Deep Learning (DL) shows promise in medical image analysis but lacks clinical interpretability due to its black-box nature.
• Current diagnostic systems require enhanced reliability and transparency for effective clinical adoption.

Purpose of the Study:

• To develop a hybrid Deep Learning (DL) framework for breast cancer diagnosis.
• To enhance classification performance and address the interpretability limitations of DL models.
• To integrate explainable artificial intelligence (XAI) for clinical validation.

Main Methods:

• A hybrid DL framework was developed, fusing three pre-trained Convolutional Neural Network (CNN) architectures: DENSENET121, Xception, and VGG16.
• An Explainable Artificial Intelligence (XAI) component using GradCAM++ was incorporated to provide model interpretability.
• The framework was evaluated on benchmark breast cancer datasets.

Main Results:

• The fused DL model achieved a classification accuracy of 97%, an improvement of approximately 13% over individual models.
• The hybrid approach demonstrated robust feature extraction and enhanced classification performance.
• GradCAM++ effectively highlighted lesions with finer edges in ultrasound images, aiding diagnostic validation.

Conclusions:

• The fused DL model significantly improves breast cancer classification accuracy and robustness.
• The integration of XAI provides interpretable results, making the system suitable for clinical applications.
• This interpretable AI system can aid medical professionals in validating diagnoses and improving patient outcomes.