Research Article

Attention-Guided Hybrid CNN–Transformer Architecture for Multi-Scale Skin Lesion Classification

Number: 2026 November 20, 2026
EN TR

Attention-Guided Hybrid CNN–Transformer Architecture for Multi-Scale Skin Lesion Classification

Abstract

The automatic classification of skin lesions is a critical requirement in dermatology for the development of early diagnosis and clinical decision support systems. In this context, a comprehensive evaluation of the skin lesion classification problem was performed using the HAM10000 dermoscopic image dataset. The study examined 14 different transfer learning-based convolutional neural network (CNN) models, 3 different Vision Transformer (ViT) architectures, and a novel CNN+ViT hybrid model that combines the strengths of both approaches. The proposed hybrid architecture enhances the CNN-based local feature extraction process at the channel and spatial levels using the Convolutional Block Attention Module (CBAM), while effectively modeling global context information through integrated self-attention layers. This creates a multi-level attention structure focused on discriminative lesion regions. All models were trained using a five-fold cross-validation strategy and evaluated across six complementary performance metrics: accuracy, precision, recall, specificity, F-score, and Matthews correlation coefficient. Experimental results demonstrate that the proposed CNN+ViT hybrid approach delivers consistent and superior performance across different input patch sizes. Specifically, the hybrid model, achieving 99.0% accuracy, 98.8% F-score, and 97.9% MCC, provided lower variance and higher stability compared to both pure CNN and pure ViT-based architectures. These findings demonstrate that CBAM and self-attention-supported hybrid architectures offer an effective, reliable, and clinically promising solution for the automatic classification of skin lesions.

Keywords

References

  1. Abohashish, S. M., Amin, H. H., & Elsedimy, E. I. (2025). Enhanced melanoma and non-melanoma skin cancer classification using a hybrid LSTM-CNN model. Scientific Reports, 15(1), 24994.
  2. Akilandasowmya, G., Nirmaladevi, G., Suganthi, S. U., & Aishwariya, A. (2024). Skin cancer diagnosis: Leveraging deep hidden features and ensemble classifiers for early detection and classification. Biomedical Signal Processing and Control, 88, 105306.
  3. Attallah, O. (2024). Skin-CAD: Explainable deep learning classification of skin cancer from dermoscopic images by feature selection of dual high-level CNNs features and transfer learning. Computers in Biology and Medicine, 178, 108798.
  4. Bazgir, E., Haque, E., Maniruzzaman, M., & Hoque, R. (2024). Skin cancer classification using Inception Network. World Journal of Advanced Research and Reviews, 21(02), 839-849.
  5. Campanella, G., Hanna, M. G., Geneslaw, L., Miraflor, A., Werneck Krauss Silva, V., Busam, K. J., ... & Fuchs, T. J. (2019). Clinical-grade computational pathology using weakly supervised deep learning on whole slide images. Nature medicine, 25(8), 1301-1309.
  6. Chanda, D., Onim, M. S. H., Nyeem, H., Ovi, T. B., & Naba, S. S. (2024). DCENSnet: A new deep convolutional ensemble network for skin cancer classification. Biomedical Signal Processing and Control, 89, 105757.
  7. Chollet, F. (2017). Xception: Deep learning with depthwise separable convolutions. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 1251-1258).
  8. Dosovitskiy, A. (2020). An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929.

Details

Primary Language

English

Subjects

Deep Learning, Machine Vision

Journal Section

Research Article

Publication Date

November 20, 2026

Submission Date

September 21, 2026

Acceptance Date

November 20, 2026

Published in Issue

Year 2026 Number: 2026

APA
Barin Özkan, S. (2026). Attention-Guided Hybrid CNN–Transformer Architecture for Multi-Scale Skin Lesion Classification. Computer Science, 2026. https://doi.org/10.53070/bbd.2044616
AMA
1.Barin Özkan S. Attention-Guided Hybrid CNN–Transformer Architecture for Multi-Scale Skin Lesion Classification. JCS. 2026;(2026). doi:10.53070/bbd.2044616
Chicago
Barin Özkan, Sibel. 2026. “Attention-Guided Hybrid CNN–Transformer Architecture for Multi-Scale Skin Lesion Classification”. Computer Science, nos. 2026. https://doi.org/10.53070/bbd.2044616.
EndNote
Barin Özkan S (November 20, 2026) Attention-Guided Hybrid CNN–Transformer Architecture for Multi-Scale Skin Lesion Classification. Computer Science 2026
IEEE
[1]S. Barin Özkan, “Attention-Guided Hybrid CNN–Transformer Architecture for Multi-Scale Skin Lesion Classification”, JCS, no. 2026, Nov. 2026, doi: 10.53070/bbd.2044616.
ISNAD
Barin Özkan, Sibel. “Attention-Guided Hybrid CNN–Transformer Architecture for Multi-Scale Skin Lesion Classification”. Computer Science. 2026 (November 20, 2026). https://doi.org/10.53070/bbd.2044616.
JAMA
1.Barin Özkan S. Attention-Guided Hybrid CNN–Transformer Architecture for Multi-Scale Skin Lesion Classification. JCS. 2026. doi:10.53070/bbd.2044616.
MLA
Barin Özkan, Sibel. “Attention-Guided Hybrid CNN–Transformer Architecture for Multi-Scale Skin Lesion Classification”. Computer Science, no. 2026, Nov. 2026, doi:10.53070/bbd.2044616.
Vancouver
1.Sibel Barin Özkan. Attention-Guided Hybrid CNN–Transformer Architecture for Multi-Scale Skin Lesion Classification. JCS. 2026 Nov. 20;(2026). doi:10.53070/bbd.2044616

The Creative Commons Attribution 4.0 International License 88x31.png is applied to all research papers published by JCS and

A Digital Object Identifier (DOI) Logo_TM.png is assigned for each published paper.