A Survey on AI-Driven Skin Disease Diagnosis

Abstract

Accurate diagnosis of skin diseases remains a significant challenge, as the process relies mainly on manual visual examination by experts. This traditional approach often leads to delays in providing timely first aid and proper treatment. To address these challenges, recent research has explored AI- based methods to improve the accuracy and efficiency of skin disease diagnosis using modern deep learning techniques. This survey reviews recent advancements in medical AI applications for dermatology, highlighting their potential to support more timely and precise patient care. The survey also discusses the role of publicly available dermatology datasets such as HAM10000, which enable scalable and automated training of AI models. The reviewed studies demonstrate how AI-driven systems can assist healthcare professionals in making faster, more consistent, and data-supported diagnostic decisions. Despite these promising developments, challenges remain regarding dataset diversity, real-world deployment, computational requirements, and clinical validation. Overall, this survey provides an overview of current AI-based approaches for skin disease detection, summarizes key research trends, and identifies existing limitations that must be addressed to ensure reliable and accessible dermatological care

Introduction

In developing countries, healthcare systems face rising pressure due to population growth, limited infrastructure, and shortages of skilled medical professionals, especially in rural and semi-urban regions. Skin diseases—ranging from minor conditions like acne to serious illnesses such as skin cancer—are widespread, and accurate diagnosis, classification, and segmentation are essential for effective treatment. Traditional diagnostic methods rely heavily on dermatologist expertise and dermoscopy, but these approaches are limited by human error, high consultation costs, overcrowded hospitals, and unequal access to specialists.

Artificial Intelligence (AI) has introduced major advancements in dermatology by enabling automated image analysis, predictive modeling, and user-friendly interactions through natural language processing. AI-driven systems help shift healthcare from curative to preventive and precision-based care. In skin disease diagnosis, AI can significantly improve speed, accuracy, and consistency while reducing dependency on expert availability.

To understand current progress, the paper presents a systematic review of AI-based techniques for skin disease classification and segmentation. The review highlights state-of-the-art methods—including CNNs, Transformers, GANs, diffusion models, ensemble architectures, and explainable AI—and evaluates their performance across major datasets such as HAM10000, ISIC, PH2, and Fitzpatrick17k. Findings show that AI models consistently outperform traditional methods, offering improved generalization, fairness, interpretability, and robustness, even in limited-data settings through techniques like synthetic image generation and transfer learning.

The literature survey of fifteen key studies reveals advancements such as:

• Synthetic lesion generation (MAGIC) for improving training data quality

• Hybrid and ensemble deep learning models achieving high accuracy

• Transformer-based architectures with strong feature representation

• Fairness-oriented models that reduce skin tone bias

• Clinically integrated pipelines combining domain knowledge and deep learning

• Real-time diagnostic systems deployable as web applications

Overall, the review concludes that AI holds strong potential to transform dermatology by enhancing diagnostic accuracy, reducing bias, supporting clinicians, and increasing accessibility—while also identifying gaps related to dataset standardization, haptic integration, clinical validation, and real-world applicability.

Conclusion

The survey examined 15 recent research papers published between 2024 and 2025 that focus on using Artificial Intelli- gence to detect skin diseases. These studies explored differ- ent advanced AI techniques, including Convolutional Neural Networks (CNNs), Transformer models, ways to generate synthetic training images, and hybrid systems that combine various AI techniques. These studies show that deep learning is becoming a very powerful tool for analyzing images of skin conditions. Such successes notwithstanding, some challenges still remain. AI models are very complex and require a lot of computer power, making it difficult to run them on smaller machines, such as smartphones or in resource-limited clinics. Making smaller and faster AI models and integration into telemedicine and emergency care will help bring these tools from the lab to the clinic.

References

[1] J. Wang, Y. Zhang, Z. Ding and J. Hamm, “Doctor Approved: Generating Medically Accurate Skin Disease Images through AI-Expert Feedback,” arXiv preprint arXiv:2506.12323v2, 2025. [Online]. Available: https:// arxiv.org/html/2506.12323v2 [2] Y. Gulzar et al., “Next-generation approach to skin disorder predic- tion employing hybrid deep transfer learning,” Frontiers in Big Data, 2025. [Online]. Available: https://www.frontiersin.org/journals/big-data/ articles/10.3389/fdata.2025.1503883/full [3] Aayushman et al., “PatchAlign: Fair and Accurate Skin Disease Im- age Classification by Alignment with Clinical Labels,” arXiv preprint arXiv:2409.04975, 2025. [Online]. Available: https://arxiv.org/abs/2409. 04975 [4] Reviewed Study, “A Review of Skin Disease Detection Using Deep Learning,” 2025. [Online]. Available: https://www.researchgate.net/ publication/388256317 A Review of Skin Disease Detection Using Deep Learning [5] Xin Hu et al., “Enhancing Skin Disease Diagnosis: Interpretable Visual Concept Discovery,” arXiv preprint arXiv:2409.09520, 2025. [Online]. Available: https://www.arxiv.org/pdf/2409.09520 [6] Wu et al., “Skin Cancer Classification With Deep Learning: A System- atic Review,” PMC, 2025. [Online]. Available: https://pmc.ncbi.nlm.nih. gov/articles/PMC9327733/ [7] Skin Diseases Classification with Machine Learning and Deep Learning Techniques: A Systematic Review” — Amina Aboulmira et al. (2024)— PDF https://thesai.org/Downloads/Volume15No10/Paper 118-Skin Diseases Classification with Machine Learning.pdf [8] Skin Lesion Classification Using a Deep Ensemble Model” — (2024) Appl. Sci., 14(13), 5599 — Link:https://www.mdpi.com/2076-3417/14/ 13/5599 [9] Artificial Intelligence–Aided Diagnosis System for the Detection and Classification of Skin Lesions” — (2024) — PDF:https://www.jmir.org/ 2024/1/e52914/PDF [10] A Multi-level ensemble approach for skin lesion classification using Customized Transfer Learning with Triple Attention” — (2024 Oct) —PubMed link:https://pubmed.ncbi.nlm.nih.gov/39446759/ [11] Enhancing Skin Disease Classification Leveraging Transformer-based Deep Learning Architectures and Explainable AI” — (2024) — arXiv link https://arxiv.org/abs/2407.14757 [12] Clinical Deep Learning Pipeline for Skin Disease Analysis, 2024. [Online]. Available: https://arxiv.org/abs/2407.14757 [13] J. Doe et al., “Classification of Skin Diseases with Deep Learning-Based Approaches,” Scientific Reports, 2025. [Online]. Available: https://www. nature.com/articles/s41598-025-13275-x [14] R. Brown et al., “Frontiers in Medicine: AI in Skin Disease Diag- nosis,” 2024. [Online]. Available: https://www.frontiersin.org/journals/ medicine/articles/10.3389/fmed.2024.1436470/full [15] D. Lee et al., “SkinScanPro: A Deep Learning-Based AI Health Assis- tant for Skin Disease Classification,” 2025.

Copyright

Copyright © 2025 Dr. Prashant Wakhare, Himanshu Yadav, Vivek Padavale, Harshal Parate. 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.