Monkeypox Skin-Lesion Detection with a Modified VGG16 and a Lightweight Custom CNN.
DOI:
https://doi.org/10.63682/jns.v14i32S.10122Keywords:
Monkeypox, skin lesion, deep learning, VGG16, transfer learning, custom CNN, medical image classification, Grad-CAM, focal loss, tele-dermatologyAbstract
Rapid, accurate recognition of monkeypox skin lesions from clinical photographs can support triage when laboratory testing is delayed or unavailable. We propose a hybrid deep-learning pipeline that couples a modified VGG16 backbone (initialized with ImageNet weights and fine-tuned on lesion images) with a custom lightweight CNN head for domain-specific feature refinement. The workflow (Fig. 1) comprises dataset curation, stratified train/test splitting, color/illumination normalization, lesion-aware augmentations (random crop, flip, scale, blur, JPEG artifacts), and class-imbalance handling via focal loss and mixup. The modified VGG16 replaces fully connected layers with global average pooling and dropout, integrates depthwise-separable bottlenecks to reduce parameters, and uses label-smoothing at the softmax output. A compact custom CNN branch in parallel learns texture-scale patterns typical of vesicular/pustular lesions; both streams are fused by attention-based feature concatenation before the final classifier. Evaluation uses accuracy, precision, recall, F1, ROC–AUC, Cohen’s κ, and calibration (ECE), with patient-level grouping to avoid leak. Grad-CAM visualizations provide clinical interpretability by highlighting lesion regions influencing predictions. The approach is designed for deployment on modest hardware and tele-dermatology workflows. Results demonstrate that combining transfer learning with a purpose-built CNN head yields substantial gains over single-backbone baselines, improving sensitivity to early, small, or low-contrast lesions while maintaining strong specificity against confounders (acne, varicella, molluscum)..
Downloads
References
[1] Sitaula, Chiranjibi, and Tej Bahadur Shahi. "Monkeypox virus detection using pre-trained deep learning-based approaches." arXiv preprint arXiv:2209.04444 (2022).
[2] Joseph Paul Cohen, Paul Morrison, and Lan Dao. Covid-19 image data collection. arXiv preprint arXiv:2003.11597, 2020.
[3] Luna-Perej´on, F. et al. (2020). Low-power embedded system for gait classification using neural networks. Journal of Low Power Electronics and Applications, 10 , 14.
[4] Ali, S. N., Ahmed, M., Paul, J., Jahan, T., Sani, S., Noor, N., Hasan, T. et al. (2022). Monkeypox skin lesion detection using deep learning models: A feasibility study. ArXiv preprint arXiv:2207.03342.
[5] Ahsan, M. M., Uddin, M. R., & Luna, S. A. (2022b). Monkeypox image data collection. arXiv preprint arXiv:2206.01774.
[6] Phi-Yen Nguyen, Whenayon Simeon Ajisegiri, Valentina Costantino, Abrar A Chughtai, and C Raina MacIntyre. Reemergence of human monkeypox and declining population immunity in the context of urbanization, nigeria, 2017–2020. Emerging Infectious Diseases, 27(4):1007, 2021.
[7] Md Manjurul Ahsan, Redwan Nazim, Zahed Siddique, and Pedro Huebner. Detection of covid-19 patients from ct scan and chest x-ray data using modified mobilenetv2 and lime. In Healthcare, volume 9, page 1099. Multidisciplinary Digital Publishing Institute, 2021.
[8] Tim Menzies, Jeremy Greenwald, and Art Frank. Data mining static code attributes to learn defect predictors. IEEE transactions on software engineering, 33(1):2–13, 2006.
[9] Linda Wang, Zhong Qiu Lin, and Alexander Wong. Covid-net: A tailored deep convolutional neural network design for detection of covid-19 cases from chest x-ray images. Scientific Reports, 10(1):1–12, 2020.
[10] Pan Pan, Yichao Li, Yongjiu Xiao, Bingchao Han, Longxiang Su, Mingliang Su, Yansheng Li, Siqi Zhang, Dapeng Jiang, Xia Chen, et al. Prognostic assessment of covid-19 in the intensive care unit by machine learning methods: model development and validation. Journal of medical Internet research, 22(11):e23128, 2020.
[11] He, K., Zhang, X., Ren, S., & Sun, J. (2016). Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 770–778).
[12] Thomas, S. M. et al. (2021). Interpretable deep learning systems for multiclass segmentation and classification of non-melanoma skin cancer. Medical Image Analysis, 68 , 101915.
[13] K. Simonyan and A. Zisserman, “Very deep convolutional networks for large-scale image recognition,” arXiv preprint arXiv:1409.1556, 2014. [9] K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in Proc. IEEE CVPR, 2016, pp. 770–778.
[14] K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in Proc. IEEE CVPR, 2016, pp. 770–778.
[15] C. Shorten and T. M. Khoshgoftaar, “A survey on image data augmentation for deep learning,” J. of Big Data, vol. 6, no. 1, pp. 1–48, 2019.
[16] K. M. Hosny, M. A. Kassem, and M. M. Foaud, “Classification of skin lesions using transfer learning and augmentation with Alex-net,” PloSone, vol. 14, no. 5, p. e0217293, 2019
Downloads
Published
How to Cite
Issue
Section
License
This work is licensed under a Creative Commons Attribution 4.0 International License.
You are free to:
- Share — copy and redistribute the material in any medium or format
- Adapt — remix, transform, and build upon the material for any purpose, even commercially.
Terms:
- Attribution — You must give appropriate credit, provide a link to the license, and indicate if changes were made. You may do so in any reasonable manner, but not in any way that suggests the licensor endorses you or your use.
- No additional restrictions — You may not apply legal terms or technological measures that legally restrict others from doing anything the license permits.
