RGB skin erythema imaging

Background: Erythema (a blanchable redness of the skin caused by vasodilation and local increase in cutaneous oxyhemoglobin content) is a clinical indicator of various medical conditions. Visualizing erythema can be challenging, particularly in patients with high melanin content or pigmentary changes co-occurring with erythema. Objective: We aim to develop and validate image processing algorithms capable of highlighting erythema with respect to background melanin pigmentation in standard RGB (red-green-blue) images. Methods: Two computational approaches were evaluated to characterize erythema across varying skin tones. First, an RGB erythema index was developed for erythema accentuation and melanin suppression. Secondly, a pigmentation-corrected optical density (OD) model was developed to linearize RGB data, creating a melanin axis orthogonal to the erythematous signal. Additionally, a semi-automated masking tool integrating these algorithms was developed to enable ground truth annotation for machine learning applications. Results: The RGB erythema index and OD mapping demonstrated complementary performance across diverse skin tones and conditions. The RGB index performed better with specular reflection, while OD mapping required clear ROIs for effective calibration. Both methods showed context-dependent limitations, highlighting the need for an adaptive approach to assess erythema independent of melanin. Conclusions: The RGB erythema index and optical density mapping provide complementary approaches for erythema enhancement in diverse skin tones. The RGB index performs better with specular reflection, while the OD method requires clear ROIs for calibration but offers improved melanin suppression. The semi-automated masking tool integrates both methods to generate training data for future machine learning applications. This preliminary work will lay the groundwork for our development of context-aware erythema assessment in clinical dermatology.