Artificial intelligence–enabled precision medicine for inflammatory skin diseases

Comprehensive Summary

In this article, Tang et al. provide a review of current generative AI and machine learning models being applied in dermatology, particularly regarding their abilities in diagnosis, phenotyping, treatment personalization, drug discovery, and clinical care delivery. Additionally, the disease-specific applications of AI and ML were analyzed.The authors conducted a literature review through keyword searches to target works which involved both AI/ML models and inflammatory skin diseases. All works selected were evaluated for quality and relevance prior to integration into the review. Tang et al. found that the majority of AI applications involved image-based classifications supplemented by clinician evaluation, using convolutional neural networks and transformer-based models. Most recent developments involve multimodal data to integrate patient background, genomics, and other information into diagnosis to mimic a clinician approach to diagnosis. For less studied diseases, such as vitiligo, dermatomyositis, and cutaneous lupus, applications of ML and AI models remain limited as the reference material itself cannot fully supplement image data.

Outcomes and Implications

The authors emphasize that, while AI and ML models serve as reliable clinical aids increasing efficiency of diagnosis, further developments are required such as reducing dataset bias and managing interpretability concerns. Further development of multimodal data integration will assist in progressing diagnostic and clinical care development. This research highlights concerns of AI/ML model development, which are vital to push the development of such models. Integration of such models, while beneficial to efficiency and accessibility of dermatological care, poses risks should it be applied without clinician supervision. Tang et al. recommend further development prior to clinical integration.