Technique Standards for Skin Lesion Imaging

Abstract

IMPORTANCE Variability in the metrics for image acquisition at the total body, regional, close-up, and dermoscopic levels impacts the quality and generalizability of skin images. Consensus guidelines are indicated to achieve universal imaging standards in dermatology. OBJECTIVE To achieve consensus among members of the International Skin Imaging Collaboration (ISIC) on standards for image acquisition metrics using a hybrid Delphi method. EVIDENCE REVIEW Delphi study with 5 rounds of ratings and revisions until relative consensus was achieved. The initial set of statements was developed by a core group (CG) on the basis of a literature review and clinical experience followed by 2 rounds of rating and revisions. The consensus process was validated by an extended group (EG) of ISIC members through 2 rounds of scoring and revisions. In all rounds, respondents rated the draft recommendations on a 1 (strongly agree) to 5 (strongly disagree) scale, explained ratings of less than 5, and optionally provided comments. At any stage, a recommendation was retained if both mean and median rating was 4 or higher. RESULTS The initial set of 45 items (round 1) was expanded by the CG to 56 variants in round 2, subsequently reduced to 42 items scored by the EG in round 3, yielding an EG set of 33 recommendations (rounds 4 and 5): general recommendation (1 guideline), lighting (5), background color (3), field of view (3), image orientation (8), focus/depth of field (3), resolution (4), scale (3), color calibration (2), and image storage (1). CONCLUSIONS AND RELEVANCE This iterative process of ratings and comments yielded a strong consensus on standards for skin imaging in dermatology practice. Adoption of these methods for image standardization is likely to improve clinical practice, information exchange, electronic health record documentation, harmonization of clinical studies and database development, and clinical decision support. Feasibility and validity testing under real-world clinical conditions is indicated.