Integrating anisotropic heat flow and transformer encoders in convolutional neural network for skin cancer classification
-
By
-
Sanad Aburass
-
Osama Dorgham
-
Ibrahim Aljarah
-
June 9, 2026
-
Objective:
To integrate an Advanced Heat Flow Layer and transformer encoding within deep learning architectures for skin cancer classification.
Key Findings:
- The proposed model outperformed benchmark models across key metrics including accuracy, precision, recall, F1 score, and AUC.
- The Advanced Heat Flow Layer effectively preserved critical edge details while smoothing images.
- Ensemble Learning enhanced the robustness and reliability of skin cancer classification.
Interpretation:
Limitations:
- The study does not claim novelty in anisotropic diffusion or CNN-Transformer architectures individually.
- The effectiveness of the ensemble approach may vary depending on the diversity of the models used.
Conclusion:
The study presents advanced image processing techniques within a deep learning framework for skin cancer classification.
