Related Experiment Videos
Establishing research priorities for artificial intelligence approaches in dermatology using an electronic-Delphi
Ravi Ramessur1,2, Luke Carson, Joseph Scott
1St John's Institute of Dermatology, School of Basic & Medical Biosciences, Faculty of Life Sciences & Medicine, King's College London, London, UK.
Background:
Artificial intelligence (AI) has significant potential to improve dermatological care, but most studies have concentrated on image-based skin cancer diagnostics. Broader applications remain limited, and it is uncertain whether existing research reflects clinicians' real-world priorities. Identifying these priorities is essential to ensure that AI development addresses genuine clinical needs and supports safe, equitable integration into practice.
Objectives:
To identify and prioritize clinician-derived research questions for the application of AI in dermatology through an international, consensus-based electronic-Delphi (e-Delphi) study.
Methods:
A 3-round e-Delphi exercise was conducted between May 2022 and July 2025, overseen by a 15-member steering group from the British Association of Dermatologists and the American Academy of Dermatology. Clinicians involved in dermatological care or research were invited to participate. In Round 1, participants submitted up to 5 research questions across 10 predefined domains. The steering group refined and collated these into unique or themed items for prioritization. In Rounds 2 and 3, participants rated randomized subsets of 30 questions using a 0-5 Likert scale, re-rating the same items in Round 3 after reviewing group mean scores.
Results:
In total, 101 participants contributed to Round 1, 150 to Round 2, and 78 to Round 3. Participants submitted 429 research questions, which were refined into 110 unique and themed items for prioritization. The highest-rated questions focused on skin cancer diagnostics and monitoring, particularly dermoscopic interpretation and lesion surveillance in high-risk groups. When cancer-related questions were excluded, priorities shifted towards triage, workflow efficiency, diagnostic support for primary care, communication between teams and improving diagnostic equity across skin tones. Dermatology consultants placed greater emphasis on service efficiency and medico-legal issues, while general practitioners and nurses prioritized implementation and access.
Conclusions:
This international consensus study identifies clinician-led priorities for AI research in dermatology. The findings provide a roadmap for researchers, funders and policymakers to align future AI development with clinical needs, supporting equitable, efficient and relevant innovation in dermatology.
