Deep learning ensemble approach with explainable AI for lung and colon cancer classification using advanced...

Area of Science:

• Oncology
• Medical Imaging
• Artificial Intelligence

Background:

• Lung and colon cancers are leading causes of cancer mortality globally.
• Accurate histopathological classification via medical imaging is crucial for diagnosis and treatment.
• Existing diagnostic methods face challenges like overfitting and poor generalizability.

Purpose of the Study:

• To develop a novel deep learning framework for enhanced classification of lung and colon cancers.
• To improve feature extraction, model robustness, and reduce overfitting in cancer diagnostics.
• To integrate interpretability methods for clinical validation and personalized treatment.

Main Methods:

• A hybrid deep learning model was created by combining Xception and MobileNet architectures.
• The ensemble model was trained on a comprehensive dataset of histopathological images.
• Model performance was validated using a balanced test set, incorporating Gradient-weighted Class Activation Mapping (Grad-CAM) for interpretability.

Main Results:

• The hybrid model achieved an outstanding classification accuracy of 99.44%.
• The model demonstrated perfect precision and recall for specific cancerous and non-cancerous tissue identification.
• Grad-CAM visualization provided interpretable insights into the model's diagnostic reasoning.

Conclusions:

• The developed deep learning framework significantly improves the accuracy and robustness of lung and colon cancer classification from medical images.
• The model's interpretability through Grad-CAM enhances clinical trust and facilitates personalized treatment strategies.
• This research paves the way for applying similar advanced AI techniques to diagnose other cancer types.