Hybrid architectures of CNN-ViT and SE-CNN with grad-CAM for monkeypox disease detection
Abstract
In the aftermath of the coronavirus disease 2019 (COVID-19) pandemic, the monkeypox virus (MPV) emerged as a significant global health concern, leading the World Health Organization (WHO) to declare it a public health emergency. Accurate detection of monkeypox (Mpox) is challenging due to its similarity to other pox viruses such as measles and chickenpox. Conventional diagnostic methods often struggle to distinguish between these diseases, necessitating the development of more advanced diagnostic tools. This paper presents a novel approach to detect Mpox using two efficient deep learning (DL) models: the first one is a hybrid architecture of a Convolutional Neural Network—Vision Transformer (CNN–ViT), and the second model is a Squeeze-and-Excitation Convolutional Neural Network (SE-CNN). These models are employed for the classification of medical images into two categories: Monkeypox and Others. The proposed models are applied to a dataset of 1140 images, and utilize fine-tuning for the models’ hyperparameters. Data augmentation techniques are implemented for training images to maximize the models’ performance. Additionally, a gradient-weighted class activation mapping (Grad-CAM) visualization technique is applied to explain the decision of the models and aid doctors in clinical decision-making. Experimentally, the accuracy, error matrix, area under the receiver operating characteristic (ROC) curve, Recall, F1-score, and Precision are used to measure the performance of the proposed models. The CNN–ViT and SE-CNN achieve a testing accuracy of 97.80% and 96.05%, respectively, while the micro-average area under the ROC curve is 97% for CNN–ViT and 99% for SE-CNN. The weighted F1-score, Precision, and Recall for CNN–ViT are all 97.26%, 97.33%, and 97.27%, respectively, while for SE-CNN are 96.04%, 96.13%, and 96.05%, respectively. The proposed models provide a rapid and accurate method to differentiate Mpox from other cases, offering a valuable tool for healthcare professionals in the screening and diagnosis of Mpox.
Related
- Investigating the deterministic and stochastic dynamics of Monkeypox: a hybrid numerical approach using spectral method and feedforward neural networks
- Subject-independent Parkinson's Disease Detection through A Lightweight Spectrotemporal Hybrid Network.
- Interpretable Diabetic Retinopathy Classification Using Hybrid ConvNeXt–MLP-Mixer and ConvNeXt–PoolFormer Architectures
- Multi-Version Evaluation of Deep Learning Architectures for Tibial Plateau Fracture Detection and Deployment in a Web-Based Clinical Support System.
- Detection of Pathological Voice Samples Using a Text-to-Speech-Inspired Approach With Multiple Model Architectures: Method Development and Comparative Evaluation Study.
BibTeX
@article{mohsen2026hybrid,
title = {{Hybrid architectures of CNN-ViT and SE-CNN with grad-CAM for monkeypox disease detection}},
author = {Saeed Mohsen and M. Mokhtar Zayed and Ahmed F. Ibrahim and M. Abdel-Aziz},
journal = {Scientific Reports},
year = {2026},
month = oct,
doi = {10.1038/s41598-026-72050-8},
url = {https://www.nature.com/articles/s41598-026-72050-8},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Hybrid architectures of CNN-ViT and SE-CNN with grad-CAM for monkeypox disease detection
AU - Saeed Mohsen
AU - M. Mokhtar Zayed
AU - Ahmed F. Ibrahim
AU - M. Abdel-Aziz
PY - 2026
DA - 2026/10/27
JO - Scientific Reports
DO - 10.1038/s41598-026-72050-8
UR - https://www.nature.com/articles/s41598-026-72050-8
AB - In the aftermath of the coronavirus disease 2019 (COVID-19) pandemic, the monkeypox virus (MPV) emerged as a significant global health concern, leading the World Health Organization (WHO) to declare it a public health emergency. Accurate detection of monkeypox (Mpox) is challenging due to its similarity to other pox viruses such as measles and chickenpox. Conventional diagnostic methods often struggle to distinguish between these diseases, necessitating the development of more advanced diagnostic tools. This paper presents a novel approach to detect Mpox using two efficient deep learning (DL) models: the first one is a hybrid architecture of a Convolutional Neural Network—Vision Transformer (CNN–ViT), and the second model is a Squeeze-and-Excitation Convolutional Neural Network (SE-CNN). These models are employed for the classification of medical images into two categories: Monkeypox and Others. The proposed models are applied to a dataset of 1140 images, and utilize fine-tuning for the models’ hyperparameters. Data augmentation techniques are implemented for training images to maximize the models’ performance. Additionally, a gradient-weighted class activation mapping (Grad-CAM) visualization technique is applied to explain the decision of the models and aid doctors in clinical decision-making. Experimentally, the accuracy, error matrix, area under the receiver operating characteristic (ROC) curve, Recall, F1-score, and Precision are used to measure the performance of the proposed models. The CNN–ViT and SE-CNN achieve a testing accuracy of 97.80% and 96.05%, respectively, while the micro-average area under the ROC curve is 97% for CNN–ViT and 99% for SE-CNN. The weighted F1-score, Precision, and Recall for CNN–ViT are all 97.26%, 97.33%, and 97.27%, respectively, while for SE-CNN are 96.04%, 96.13%, and 96.05%, respectively. The proposed models provide a rapid and accurate method to differentiate Mpox from other cases, offering a valuable tool for healthcare professionals in the screening and diagnosis of Mpox.
SN - 2045-2322
ER -