DARNet: Deep Attention Module and Residual Block-Based Lung and Colon Cancer Diagnosis Network.

Area of Science:

• Oncology
• Computer Science
• Artificial Intelligence

Background:

• Accurate lung and colon cancer classification is crucial for timely treatment and improved patient outcomes.
• Traditional classification methods are labor-intensive and require specialized expertise.
• Deep learning models offer potential but face challenges like generalization and overfitting.

Purpose of the Study:

• To propose an efficient deep learning network, DARNet, for lung and colon cancer classification.
• To address challenges in deep learning models, including generalization, overfitting, and hyperparameter tuning.
• To enhance feature extraction and learning capabilities for improved cancer classification accuracy.

Main Methods:

• Developed DARNet, incorporating residual blocks (RBs) for enhanced learning and attention modules (AM) for feature extraction.
• Utilized Bayesian Optimization (BO) for hyperparameter tuning to improve model generalization.
• Evaluated DARNet on benchmark lung and colon cancer datasets.

Main Results:

• DARNet demonstrated superior performance compared to existing models on lung and colon cancer datasets.
• The proposed model achieved a high median accuracy of 98.86%.
• DARNet exhibited lower variance, indicating robust and consistent classification performance.

Conclusions:

• The proposed DARNet effectively classifies lung and colon cancer with high accuracy and efficiency.
• Integrating residual blocks, attention modules, and Bayesian Optimization enhances deep learning model performance.
• DARNet presents a promising advancement for automated cancer classification in clinical settings.