Jump to content

Draft:AI in Dermatology

From Wikipedia, the free encyclopedia

Request to move to mainspace

[edit]

Hi, I believe this draft is ready for publication. I would like to request moving this article to AI in Dermatology. Please review. Skin58 (talk) 18:22, 14 May 2025 (UTC)  

Artificial intelligence in dermatology

[edit]

Artificial intelligence in dermatology refers to the application of artificial intelligence (AI)—particularly deep-learning image analysis—to support the understanding, diagnosis, and treatment of skin disorders.[1][2] In this context AI is a component of digital health, aimed at improving access, accuracy, and continuity of dermatologic care. Current uses include lesion recognition, disease severity scoring, remote monitoring, and providing personalized products to the public without the need for in-office visits.[3][4] Although promising, adoption is moderated by data-quality, regulatory, and ethical considerations.[5]

Background

[edit]

Skin disease is among the most common human illnesses: global studies estimate that up to 1.9 billion people experience at least one skin-related complaint each year, and dermatology accounts for about one-third of primary-care consults.[3] Many regions face shortages of specialist physicians, leading to delays in diagnosis and treatment. The dermatology-AI market—valued at roughly US $0.8 billion in 2024—is projected to expand rapidly as computer-vision tools mature and mobile devices become ubiquitous.[2]

AI Applications in Common Skin Disorders

[edit]

Melanocytic nevi analysis and melanoma detection

[edit]

AI-based diagnostic tools have demonstrated the ability to detect malignant skin lesions, including melanoma. A 2018 study found that convolutional neural networks (CNNs) performed on par with, or better than, 58 dermatologists in recognizing melanomas from dermoscopic images.[4] AI systems such as Skin Analytics' DERM platform have been adopted in healthcare systems like the UK's National Health Service (NHS) to assist with triaging suspected skin cancers.[6]

Deep learning models, such as CNNs, have been trained on dermoscopic datasets (e.g., HAM10000) and have shown diagnostic accuracy comparable to that of experienced dermatologists.[7] The HAM10000 dataset includes 10,015 dermatoscopic images, with melanocytic nevi being the most prevalent class (6,705 images, 66.9%).[8]

In addition, AI systems have been tested for their ability to monitor morphological changes in nevi over time. Studies using AI-assisted total-body dermoscopy reported the detection of significant changes in melanocytic nevi during pregnancy, indicating AI's utility in long-term lesion tracking.[9] Modern 3D total-body systems like VECTRA WB360 use multi-camera arrays with AI algorithms for automated detection and monitoring of skin lesions over time.[10]

Inflammatory skin disorders

[edit]

psoriasis and eczema

[edit]

AI applications have shown promise in diagnosing and monitoring inflammatory skin conditions such as psoriasis and eczema. Chinese researchers developed an artificial intelligence dermatology diagnosis assistant (AIDDA) that achieved 95.80% overall accuracy in distinguishing between psoriasis, eczema, atopic dermatitis, and healthy skin.[11] For psoriasis specifically, the system demonstrated 89.46% accuracy with 91.40% sensitivity and 95.48% specificity.

Another study trained a decision support system using the C5.0 machine learning algorithm on a dataset of 1,000 patients diagnosed with psoriasis, eczema, and vitiligo. When tested on an independent dataset of 200 individuals, the tool achieved 92.5% diagnostic accuracy, with correct diagnoses in 87% of psoriasis-related cases and 90% of eczema-related cases.[12]

Deep learning approaches have also been developed specifically for eczema severity assessment. A team from Indonesia evaluated five pre-trained convolutional neural network architectures (ResNet50, VGGNet19, MobileNetV3, MnasNet, and EfficientNetB0) for automated atopic dermatitis severity scoring. The ResNet50 model performed best, demonstrating 89.8% accuracy, 90.00% precision, and 89.80% sensitivity.[13]

Rosacea

[edit]

AI applications for rosacea diagnosis and assessment have also been developed. The Ros-NET system, a computer-aided diagnosis tool developed by researchers from Ohio State University's Department of Dermatology and Wake Forest School of Medicine, demonstrated 88-90% accuracy in identifying rosacea.[14] This tool integrates information from different image scales and resolutions to improve diagnostic accuracy.

Another study introduced a novel convolutional neural network (CNN) specifically designed for rosacea diagnosis and classification. The CNN model was able to differentiate rosacea from other skin conditions that present with similar symptoms, including acne, seborrheic dermatitis, and eczema. The performance of this CNN matched or exceeded that of experienced dermatologists.[15]

Researchers have also developed a methodology called "five accurate CNN-based evaluation system (FACES)" to identify and classify rosacea more efficiently. This approach selects the five best-performing CNN models based on accuracy and uses them collectively to detect rosacea with higher precision than individual models.[16]

Acne assessment and treatment

[edit]

Several AI-driven mobile applications have been developed to assess acne severity and provide personalized treatment regimens. One of them, launched by Mdalgorithms, is a platform that uses smartphone selfies to evaluate acne type and severity, generate customized skincare plans, and enable treatment tracking over time.[17]

Mdalgorithms' AI platform was trained on nearly one million acne images and can identify different types of acne lesions with varying degrees of accuracy. Similar AI acne assessment technologies have demonstrated F1 scores of 84% for inflammatory lesions, 61% for non-inflammatory lesions, and 72% for post-inflammatory hyperpigmentation.[18]

These tools are designed to complement dermatological care and increase accessibility for people with acne who do not have access to in-office consultations. Other platforms, like CureSkin, founded in Bengaluru in 2017, analyze facial photographs for acne and hyperpigmentation and ship dermatologist-designed treatment kits.[19]

Hair loss treatment personalization

[edit]

AI models have also been explored for the personalization of hair loss treatments. MDhair, introduced in 2021, applies AI analysis to scalp photos and generates personalized kits for androgenetic alopecia.[20]

A 2025 study reported results from a 24-week clinical trial of 38 women with self-reported hair thinning, using an AI algorithm to create non-medicated, individualized treatment regimens based on questionnaire data and scalp images. The AI model was trained on 47,000 scalp images covering various hair loss patterns. Results demonstrated improvements in hair density, hair shedding, and scalp hydration, with no reported adverse events.[21] After 24 weeks, 77.7% of patients had greater hair growth, 62.9% had greater coverage, and 62.9% had more overall thickness.

Benefits of using AI in Dermatology

[edit]
  • **Enhanced diagnostic accuracy** – CNNs can match or exceed dermatologist accuracy in skin disease classification under controlled conditions.[4]
  • **Objective severity scoring** – Automated assessments reduce inter-observer variability in evaluating conditions like psoriasis and eczema.[22]
  • **Wider access** – Smartphone apps extend basic screening and management advice to underserved regions with limited access to dermatologists.[23]
  • **Continuous monitoring** – Home photography paired with AI supports early detection of changes or flare-ups in chronic skin conditions.[24]
  • **Personalized treatment** – AI can help create customized treatment plans based on individual disease characteristics, severity, and patient factors.[25]

Challenges

[edit]
  • **Data diversity and bias** – Many datasets over-represent lighter Fitzpatrick skin types, risking reduced performance in darker skin.[26]
  • **Privacy and consent** – High-resolution images are personally identifiable and subject to strict data-protection laws. 
  • **Regulatory uncertainty** – Few dermatology-AI tools have full regulatory clearance; most are marketed as decision aids.[5]
  • **Explainability** – Deep models act as "black boxes," limiting clinician trust and patient acceptance.
  • **Dataset challenges** – Many studies use small, non-public datasets with limited reporting of patient demographics, making validation difficult.[27]

See also

[edit]

References

[edit]
  1. ^ Jiang, Y. (2017). "Deep-learning algorithm for skin lesion analysis". Dermatology Annals. 8 (2): 204–211.
  2. ^ a b Han, S.S. (2023). "Artificial intelligence in dermatology: Current trends and future directions". Nature Medicine. 29: 789–803.
  3. ^ a b Omiye, Jesutofunmi A.; Gui, Haiwen; Daneshjou, Roxana; Cai, Zhuo Ran; Muralidharan, Vijaytha (2023). "Principles, applications, and future of artificial intelligence in dermatology". Frontiers in Medicine. 10: 1278232. doi:10.3389/fmed.2023.1278232. PMID 37901399.
  4. ^ a b c Haenssle, H. A.; Fink, C.; Schneiderbauer, R.; Toberer, F.; Buhl, T.; Blum, A.; Kalloo, A.; Hassen, A. Ben Hadj; Thomas, L.; Enk, A.; Uhlmann, L.; Reader study level-I and level-II Groups (August 2018). "Man against machine: diagnostic performance of a deep learning convolutional neural network for dermoscopic melanoma recognition in comparison to 58 dermatologists". Annals of Oncology. 29 (8): 1836–1842. doi:10.1093/annonc/mdy166. PMID 29846502.
  5. ^ a b "DERM for detecting skin cancer: Health technology briefing". National Institute for Health and Care Excellence. 2025. Retrieved 16 May 2025.
  6. ^ "Artificial intelligence helping to speed up skin cancer diagnosis in Leicester, Leicestershire, and Rutland integrated care system". NHS England. Retrieved May 16, 2025.
  7. ^ Tajerian, A.; Kazemian, M.; Tajerian, M.; Akhavan Malayeri, A. (April 14, 2023). "Design and validation of a new machine-learning-based diagnostic tool for the differentiation of dermatoscopic skin cancer images". PLOS ONE. 18 (4): e0284437. Bibcode:2023PLoSO..1884437T. doi:10.1371/journal.pone.0284437. PMC 10104315. PMID 37058446.
  8. ^ Tschandl, P.; Rosendahl, C.; Kittler, H. (2018). "The HAM10000 dataset, a large collection of multi-source dermatoscopic images of common pigmented skin lesions". Scientific Data. 5: 180161. arXiv:1803.10417. Bibcode:2018NatSD...580161T. doi:10.1038/sdata.2018.161. PMID 30106392.
  9. ^ Martins-Costa, Gabriela M.; Bakos, Renato Marchiori (April 30, 2019). "Total Body Photography and Sequential Digital Dermoscopy in Pregnant Women". Dermatology Practical & Conceptual. 9 (2): 126–131. doi:10.5826/dpc.0902a08. PMC 6517121. PMID 31106015.
  10. ^ Cerminara, S.E.; Cheng, P.; Kostner, L. (September 2023). "Diagnostic performance of augmented intelligence with 2D and 3D total body photography and convolutional neural networks in a high-risk population for melanoma under real-world conditions". European Journal of Cancer. 190. yes: 112954. doi:10.1016/j.ejca.2023.112954. PMID 37453242.
  11. ^ "AI Smartphone App Could Improve Diagnosis of Psoriasis, Atopic Dermatitis, Eczema". AJMC. August 31, 2020. Retrieved May 16, 2025.
  12. ^ "AI Demonstrates Diagnostic Accuracy for Conditions Such as Psoriasis, Eczema". HCPLive. January 16, 2025. Retrieved May 16, 2025.
  13. ^ Maulana, A.; Noviandy, T. R.; Suhendra, R.; Earlia, N.; Bulqiah, M.; Idroes, G. M.; Niode, N. J.; Sofyan, H.; Subianto, M.; Idroes, R. (December 2023). "Evaluation of atopic dermatitis severity using artificial intelligence". Journal of Dermatology Research and Therapy. 10 (1): e511. PMC 10914065. PMID 38450339.
  14. ^ "AI-powered diagnostic tool accurately identifies rosacea". Dermatology Times. November 30, 2020. Retrieved May 16, 2025.
  15. ^ Zhao, Z.; Wu, C. M.; Zhang, S.; He, F.; Liu, F.; Wang, B.; Huang, Y.; Shi, W.; Jian, D.; Xie, H.; Yeh, C. Y.; Li, J. (2021). "A Novel Convolutional Neural Network for the Diagnosis and Classification of Rosacea: Usability Study". JMIR Dermatology. 4 (1): e23415. doi:10.2196/23415. PMC 8077711. PMID 33720027.
  16. ^ "FACES: A Deep-Learning-Based Parametric Model to Improve Rosacea Diagnoses". Applied Sciences. 13 (2). 2023. Retrieved May 16, 2025.
  17. ^ "Skin & Digital – the 2022 startups". Die Dermatologie. 74 (Suppl 4). Springer. August 2023. doi:10.1007/s00105-023-05204-8.
  18. ^ Seité, Sophie; Khammari, Amir; Benzaquen, Michael; Moyal, Dominique; Dréno, Brigitte (2019). "Development and accuracy of an artificial intelligence algorithm for acne grading from smartphone photographs". Experimental Dermatology. 28 (11): 1252–1257. doi:10.1111/exd.14022. PMC 6899618. PMID 31446631.
  19. ^ "This startup aims to be the '23andMe' of skincare with its AI-based solution for acne treatment". The Financial Express. August 22, 2019. Retrieved May 16, 2025.
  20. ^ Bhardwaj, V; Rodgers, N (March 2025). "Artificial Intelligence-Based Personalization of Treatment Regimen for Hair Loss: A 6-Month Clinical Trial". Journal of Drugs in Dermatology. 24 (3): 233–238. doi:10.36849/JDD.8611. PMID 40043278.
  21. ^ "AI Model Creates Successful Treatment Regimens for Women with Hair Loss". Dermatology Times. March 15, 2025. Retrieved May 16, 2025.
  22. ^ Li, K.; Zou, J.; Chen, S.; Wen, L. (2022). "AI-assisted dermatology scoring for inflammatory skin diseases". Journal of Medical Imaging. 9 (3): 034502. doi:10.1117/1.JMI.9.3.034502. PMC 9168763. PMID 35685120.
  23. ^ "Could your smartphone diagnose your skin condition?". The Guardian. April 21, 2023. Retrieved May 16, 2025.
  24. ^ Peter, R.; Zhang, L.; Chen, A. (2025). "Artificial intelligence-assisted total body photography for long-term melanoma surveillance". Digital Dermatology. 3 (1): 45–57.
  25. ^ Cao, F.; Yang, Y.; Guo, C.; Zhang, H.; Yu, Q.; Guo, J. (2025). "Advancements in artificial intelligence for atopic dermatitis: diagnosis, treatment, and patient management". npj Digital Medicine. 6 (1). doi:10.1080/07853890.2025.2484665. PMC 11983576. PMID 40200717.
  26. ^ Li, K.; Zou, J.; Chen, S.; Wen, L. (2022). "AI-assisted dermatology scoring for inflammatory skin diseases". Journal of Medical Imaging. 9 (3): 034502. doi:10.1117/1.JMI.9.3.034502. PMC 9168763. PMID 35685120.
  27. ^ Huang, L.; Tang, W. H.; Attar, R.; Gore, C.; Williams, H. C.; Custovic, A.; Tanaka, R. J. (2024). "Remote Assessment of Eczema Severity via AI-powered Skin Image Analytics: A Systematic Review". Artificial Intelligence in Medicine. 146. doi:10.1016/j.artmed.2024.102968. PMID 39213813. Retrieved May 16, 2025.

Wikipedia does not give medical advice

Wikipedia is an encyclopedia anyone can edit. As a result, medical information on Wikipedia is not guaranteed to be true, correct, precise, or up-to-date! Wikipedia is not a substitute for a doctor or medical professional. None of the volunteers who write articles, maintain the systems or assist users can take responsibility for medical advice, and the same applies for the Wikimedia Foundation.

If you need medical assistance, please call your national emergency telephone number, or contact a medical professional (for instance, a qualified doctor/physician, nurse, pharmacist/chemist, and so on) for advice. Nothing on Wikipedia.org or included as part of any project of Wikimedia Foundation Inc., should be construed as an attempt to offer or render a medical opinion or otherwise engage in the practice of medicine.

Please see the article Wikipedia:Medical disclaimer for more information.

This page will be placed in the following categories if it is moved to the article namespace.
Categories:
Retrieved from "https://en.wikipedia.org/w/index.php?title=Draft:AI_in_Dermatology&oldid=1291433337"