JMIR Publications

ABSTRACT

Background:

Dermatology is a visually driven specialty that depends on repeated exposure to disease presentations, yet the imagery used in dermatology education underrepresents skin of color (SOC). In evaluations of major dermatology textbooks, images of dark skin (Fitzpatrick phototypes V and VI) accounted for only 4%-18% of clinical images, and medical students have been shown to be less accurate in diagnosing common dermatologic conditions in SOC. Existing web-based and SOC-focused atlases have expanded access to images but are generally organized as static, 2D collections that limit interactive learning.

Objective:

This paper describes the design and development of In Living Color: Derm Atlas, an interactive, 3D, web-based dermatology atlas intended to teach the presentation of dermatologic disease in Fitzpatrick phototypes IV-VI using curated real-patient imagery, with generative artificial intelligence (AI) used to assist platform development but not image generation.

Methods:

The platform was developed around 4 design principles: (1) use of actual patient photographs rather than AI-generated imagery; (2) a primary focus on Fitzpatrick phototypes IV-VI, with lighter skin (phototypes I-II) shown only within a side-by-side comparison feature; (3) integrated clinicopathologic correlation; and (4) an interactive 3D interface. Clinical images were sourced with permission from VisualDx and selected using predefined criteria for phototype, anatomical site, and resolution, then reviewed by a board-certified dermatologist. A 3D avatar with darker skin tone was built from a licensed base mesh and mapped with selectable anatomical hotspots. Generative AI tools (Claude, Anthropic; Gemini, Google) assisted with module coding, drafting of educational text for expert review, and interface refinement. All module content underwent iterative expert review. Psoriasis was selected as the pilot condition.

Results:

The pilot psoriasis module comprises 8 anatomical sites (scalp, elbows, thigh, lumbosacral region, hands, genitals, nails, and foot) accessed through a rotatable, zoomable 3D avatar. Each site presents 2-4 enlargeable clinical photographs with dermatologic descriptions, SOC-specific clinical notes, clinical pearls, and differential diagnoses with distinguishing features. A side-by-side comparison feature pairs presentations in Fitzpatrick IV-VI and I-II skin with text highlighting differences in morphology, pigmentation, and lesion appearance. The application runs in standard desktop and mobile web browsers without specialized hardware or institutional login. The module is in beta testing, and learner outcome data have not yet been collected.

Conclusions:

In Living Color: Derm Atlas combines authentic SOC imagery, anatomical interactivity, and planned dermatopathology integration within a single digital platform, using generative AI for development while preserving the source transparency and clinical authenticity of its images. Planned expert validation and pretest-posttest learner evaluation will determine whether this design improves visual pattern recognition and provides educational value beyond existing static resources.