Scholarlink

Accurate and timely diagnosis of skin diseases, including Monkeypox (Mpox), is crucial because it requires immediate attention. However, it is challenging as most skin conditions often present similar symptoms. Recently, computer aided diagnostic systems have been developed to classify Mpox and other skin conditions. This paper presents a new efficient deep learning based diagnostic system that uses pre-trained CNN models (EfficientNetB0, MobileNetV2, ResNet50) along with Convolutional Block Attention Mechanisms (CBAM) and a custom Artificial Neural Network (ANN) to detect 6 skin conditions: Monkeypox, Chickenpox, Measles, Cowpox, Hand-Foot-and-Mouth Disease (HFMD), and Healthy skin. The model is trained and tested on a balanced version of the Mpox Skin Lesion Dataset version 2.0 (MSLD v2.0), which includes 6,000 high-quality dermoscopic images. It achieves an impressive validation accuracy of 94% and AUC score of 0.97. The paper highlights the proposed approach’s superiority in identifying the subtle patterns and detecting the underlying condition. In future, deployment, explainability, and clinical integration are planned. The dataset and model architecture are made publicly available to encourage additional research on the application of AI in skin disease detection.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Skin Image-Based Monkeypox Classification: A Novel Approach Using Transfer Learning, CBAM, and ANN

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Afnan Al-Johani,
  • Mohammed Al-Sarem

摘要

Accurate and timely diagnosis of skin diseases, including Monkeypox (Mpox), is crucial because it requires immediate attention. However, it is challenging as most skin conditions often present similar symptoms. Recently, computer aided diagnostic systems have been developed to classify Mpox and other skin conditions. This paper presents a new efficient deep learning based diagnostic system that uses pre-trained CNN models (EfficientNetB0, MobileNetV2, ResNet50) along with Convolutional Block Attention Mechanisms (CBAM) and a custom Artificial Neural Network (ANN) to detect 6 skin conditions: Monkeypox, Chickenpox, Measles, Cowpox, Hand-Foot-and-Mouth Disease (HFMD), and Healthy skin. The model is trained and tested on a balanced version of the Mpox Skin Lesion Dataset version 2.0 (MSLD v2.0), which includes 6,000 high-quality dermoscopic images. It achieves an impressive validation accuracy of 94% and AUC score of 0.97. The paper highlights the proposed approach’s superiority in identifying the subtle patterns and detecting the underlying condition. In future, deployment, explainability, and clinical integration are planned. The dataset and model architecture are made publicly available to encourage additional research on the application of AI in skin disease detection.