Skin Disease Detection and Classification

Abstract

Skin diseases affect millions of people worldwide, ranging from common conditions like acne and eczema to severe cases such as melanoma and carcinoma. Early and accurate diagnosis is crucial for effective treatment and improved patient outcomes. However, traditional diagnostic methods rely heavily on clinical expertise, which can be subjective and time-consuming. This research focuses on developing an automated skin disease detection and classification system using image processing and machine learning techniques. The proposed approach utilizes Convolutional Neural Networks (CNNs) to analyze dermatological images and classify different skin conditions, including melanoma, psoriasis, eczema, and fungal infections. A large dataset of labeled skin images is used to train the deep learning model, enabling it to recognize subtle patterns and variations in skin texture, color, and morphology. Techniques such as data augmentation, transfer learning, and feature extraction enhance the model’s accuracy and generalization ability. Experimental results demonstrate that the proposed system achieves high classification accuracy compared to traditional machine learning models. The automation of skin disease detection reduces human error, minimizes diagnosis time, and enables remote diagnosis through telemedicine applications. The integration of artificial intelligence in dermatology holds great potential in aiding dermatologists, especially in regions with limited access to specialized healthcare services.

Introduction

1. Overview

Skin diseases, ranging from minor rashes to deadly melanoma, affect people globally. Traditional diagnostic methods, though effective, are time-consuming, expensive, and reliant on dermatological expertise. This has led to the adoption of AI-based systems, especially Convolutional Neural Networks (CNNs), for automated skin disease detection.

2. Role of CNNs

CNNs are highly effective in medical image analysis due to their ability to learn and extract patterns from raw image data. They can identify key visual cues such as color, shape, and texture—critical for distinguishing skin conditions. Public datasets like HAM10000 and ISIC have accelerated progress in CNN training and benchmarking.

3. Challenges

Key challenges include:

• Class imbalance (rare diseases are underrepresented).

• Model interpretability, which affects trust in medical settings.

• Dataset diversity, essential for generalization.

Solutions involve rebalancing techniques, explainable AI (XAI), and diverse training data.

4. Literature Review

Multiple architectures and methods have been applied:

• CNNs for melanoma detection showed high accuracy but had false positives.

• ResNet and VGG16 improved multi-class classification but required large datasets.

• U-Net focused on lesion segmentation.

• MobileNet and YOLO were adapted for lightweight and real-time mobile detection.

• Hybrid models (CNN + SVM, ensemble CNNs) combined strengths of multiple techniques.

5. Proposed System

The newly proposed system leverages YOLOv8 for real-time, multi-modal skin disease detection via:

• Static image analysis

• Video input

• Live camera feeds

Key Features:

• Trained on 10,000+ labeled images.

• Achieves 94.76% accuracy, outperforming traditional CNNs.

• Developed using Flask for a web-based interface.

• Lightweight and suitable for telemedicine and mobile health apps.

6. Architecture & Implementation

• Uses CNNs trained on HAM10000 with data preprocessing (resizing, normalization).

• Architecture includes multiple Conv2D layers, ReLU activations, dropout, and a softmax classifier.

• Deployed using Flask for user-friendly interaction and real-time inference.

7. Results & Evaluation

• YOLOv8 outperforms CNN-based models with:

  • 94.76% accuracy

  • High sensitivity and specificity

  • Stable loss and accuracy curves

• Outperforms previous models in speed, efficiency, and scalability.

Conclusion

The growing prevalence of skin diseases and the critical need for early diagnosis have underscored the importance of automated detection and classification systems. Accurate identification is essential for timely medical intervention, reducing the risk of disease progression and improving patient outcomes. Many dermatological conditions exhibit similar visual characteristics, making precise classification a challenging task. The application of deep learning in skin disease analysis enables rapid and reliable assessments, enhancing the efficiency of dermatological diagnostics. This study presents an advanced classification model designed to detect and differentiate multiple skin disease categories using state-of-the-art feature extraction techniques. A comprehensive dataset was compiled for both training and evaluation. The model’s performance was benchmarked against established deep learning architectures, including VGG-16 (which achieved 89.75%), ResNet-50 (which recorded 93.70%), and ResNet-101 (which yielded 83.33%). The optimized framework outperformed existing models, attaining a classification accuracy of 98.40%. Further analysis focused on refining predictions using region-based segmentation, improving contextual accuracy in medical image classification. The proposed system demonstrated a robust ability to classify various dermatological conditions with high accuracy, sensitivity, and specificity. This framework provides a valuable tool for assisting dermatologists in diagnosing skin diseases and supporting clinical decision-making. The study’s findings hold significant potential for widespread applications in telemedicine, AI-assisted diagnostics, and mobile health platforms. Future research will explore the integration of real-time diagnostic capabilities, advanced explainable AI models, and expanded datasets to enhance classification performance. Additionally, further studies could investigate the recognition of rare and visually ambiguous skin conditions, thereby improving the model’s generalizability and clinical reliability.

References

[1] Han, S. S., Park, G. H., Lim, W., Kim, M. S., Na, J. I., Park, I., & Chang, S. E. (2021). Deep neural networks for automated classification of skin diseases based on dermatological images. Journal of Dermatological Science, 102(2), 110-116. [2] Tschandl, P., Rinner, C., Apalla, Z., Argenziano, G., Codella, N., Halpern, A., ... & Kittler, H. (2020). Human–computer collaboration for skin cancer recognition. Nature Medicine, 26(8), 1229-1234. [3] Esteva, A., Kuprel, B., Novoa, R. A., Ko, J., Swetter, S. M., Blau, H. M., & Thrun, S. (2017). Dermatologist-level classification of skin cancer with deep neural networks. Nature, 542(7639), 115-118. [4] Brinker, T. J., Hekler, A., Enk, A. H., Klode, J., Hauschild, A., Berking, C., ... & Haferkamp, S. (2019). A convolutional neural network trained with dermoscopic images performs similarly to 77 dermatologists in skin cancer diagnosis. European Journal of Cancer, 111, 148-154. [5] Xie, F., Fan, H., Li, Y., Jiang, Z., Meng, R., & Bovik, A. C. (2020). Melanoma classification on dermoscopy images using a neural network ensemble model. IEEE Transactions on Medical Imaging, 39(8), 2506-2517. [6] Nida, N., Javed, A., Anwar, S. M., Khan, M. A., & Irtaza, A. (2020). Melanoma lesion detection and segmentation using deep region-based convolutional neural network and fuzzy C-means clustering. International Journal of Imaging Systems and Technology, 30(3), 621-635. [7] Argenziano, G., Soyer, H. P., De Giorgi, V., Piccolo, D., Carli, P., Delfino, M., & Ferrari, A. (2022). Artificial intelligence-based skin disease recognition: From clinical practice to future applications. JAMA Dermatology, 158(5), 525-531. [8] Li, Y., Shen, L., Zhu, H., & Zhang, C. (2021). Skin disease classification with a lightweight convolutional neural network using dermoscopic images. Biomedical Signal Processing and Control, 68, 102654. [9] Qian, W., Li, H., Gao, L., & Zhang, X. (2021). A hybrid deep learning framework for skin lesion segmentation and classification. Pattern Recognition Letters, 144, 1-8. [10] Mahbod, A., Schaefer, G., Ellinger, I., Ecker, R., & Pitiot, A. (2020). Transfer learning using a multi-scale and multi-network approach for skin lesion classification. Computer Methods and Programs in Biomedicine, 193, 105475. [11] Das, K., Conjeti, S., Roy, D., Chatterjee, S., & Sheet, D. (2020). Skin disease classification using deep learning networks with minimal data preprocessing. Neural Computing and Applications, 32(16), 12049-12061. [12] Pham, T. C., Luong, C. M., Visani, M., & Hoang, V. D. (2021). Deep CNN and data augmentation for skin lesion classification. Computers in Biology and Medicine, 134, 104457. [13] Attia, M., Hossny, M., Nahavandi, S., & Yazdabadi, A. (2020). Skin melanoma segmentation using an optimized U-Net deep learning approach. IEEE Journal of Biomedical and Health Informatics, 24(4), 1062-1069. [14] Tan, C., Sun, F., Kong, T., Zhang, W., Yang, C., & Liu, C. (2020). Skin cancer classification using attention-based convolutional neural networks. IEEE Access, 8, 112679-112691. [15] Pacheco, A. G., & Krohling, R. A. (2021). The impact of pre-processing techniques on deep learning models for skin cancer diagnosis. Expert Systems with Applications, 165, 113898. [16] Jinnai, S., Yamazaki, N., Hirano, Y., Sugawara, Y., Nakagawa, H., & Matsushita, S. (2020). The utility of deep learning in skin disease classification: A systematic review. Journal of Dermatological Science, 100(3), 191-198. [17] Liu, X., Faes, L., Kale, A. U., Wagner, S. K., Fu, D. J., Bruynseels, A., ... & Keane, P. A. (2019). A comparison of deep learning performance against healthcare professionals in detecting diseases from medical imaging: A systematic review and meta-analysis. The Lancet Digital Health, 1(6), e271-e297. [18] Mendonça, T., Ferreira, P. M., Marques, J. S., Marcal, A. R., & Rozeira, J. (2020). PH2—a dermoscopic image database for research and benchmarking in melanoma detection. Computerized Medical Imaging and Graphics, 45, 57-66. [19] Codella, N. C. F., Gutman, D., Celebi, M. E., Helba, B., Marchetti, M. A., Dusza, S. W., ... & Halpern, A. (2018). Skin lesion analysis toward melanoma detection: A challenge at the 2017 International Symposium on Biomedical Imaging. IEEE Journal of Biomedical and Health Informatics, 23(2), 501-512. [20] Simonyan, K., & Zisserman, A. (2015). Very deep convolutional networks for large-scale image recognition. International Conference on Learning Representations (ICLR).

Copyright

Copyright © 2025 Nagarjuna G R, Aishwarya G, Akshatha H M, Sreejith H S, Srujana R S. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.