Pillai, Mayank; Han, Ruijing E.; Saknite, Inga; Tkaczyk, Eric; Bowden, Audrey K. (2026).听.听Biomedical Optics Express, 17(10), 4968鈥�4991.听
The amount of skin affected by redness (erythema) is an important measure of chronic graft-versus-host disease (cGVHD), but visual estimates by clinicians can be inconsistent and complicated by changes in skin pigmentation. We developed an automated method that identifies both redness and increased pigmentation (hyperpigmentation) using standard color photographs, without requiring clinician-labeled images or specialized imaging equipment. The method converts standard RGB images into detailed spectral information and uses it to estimate levels of hemoglobin and melanin, which indicate redness and pigmentation. Although trained only on computer-generated images, the method achieved classification accuracy above 70% when tested on both synthetic images and clinical photographs. These results suggest that standard photographs could provide a more accessible and objective way to monitor skin changes in patients with cGVHD.
Fig. 1.听An overview of our two-stage hemoglobin and melanin abundance mapping pipeline on RGB images. In this work, RGB inputs represent digital camera images or synthetic RGB images generated from hyperspectral data. Hemoglobin maps obtained from reconstructed HSIs are used to predict erythema, and similarly derived melanin maps are used to predict hyperpigmentation.