LAUSR

Dear Editor, Digital imaging in dermatology is increasingly gaining importance in the clinical evaluation of skin diseases, and there is a need to understand the different properties of a photograph for evaluation of skin lesions consistently.1 Light can have different colors, referred to as color temperatures, depending on the light source in different lighting environments. For example, light from an incandescent light bulb appears to be more orange than sunlight. In a photograph, these differences are referred to as being “warmer” (or more orange) and “cooler” (or more blue) than natural sunlight. White balance adjustment (WBA) to a photograph balances the color temperature, so it appears as a neutral white, and counteracts the orange/yellow color of artificial light. The aim of this study was to investigate the influence of WBA on the evaluation of the severity of photographed atopic dermatitis (AD) lesions, by comparing the ratings with and without the WBA. Photographs for the present study were from a library of photographs taken with a smartphone application (Imagine— Skin Tracker, LEO Innovation Lab, Copenhagen, Denmark) from a previously conducted study.2 A total of 211 photographs of AD lesions taken with a white sticker (Ø 0.5 cm) placed near the lesion for the purpose of WBA (shown in Figure 1) were used in the present analysis. Each photograph was rated twice by two dermatologists with at least a nine weeks interval between the unadjusted photograph and the WBA photograph. AD severity assessment was based on the intensity part of the SCORing Atopic Dermatitis (iSCORAD) index3 consisting of six items: erythema, edema, excoriations, lichenification, oozing, and dryness, with each item assessed on a scale from 0 to 3. WBA was performed using GIMP 2.10 for Mac and adjusted using the builtin white balance tool in GIMP. To avoid a situation in which a photograph was calibrated based on a single pixel, the brush size of the calibration tool was selected to be at least 1/8th the size of the white sticker. For photographs with a partly shaded white sticker, the brush was sized to cover both shaded and unshaded areas, to ensure a uniform average. A BlandAltman plot was used to compare the ratings of adjusted versus unadjusted photographs. To compare the interrater reliability when evaluating WBA and unadjusted photographs, Intraclass Correlation Coefficients (ICC) with corresponding 95% confidence intervals (CI) were calculated using the icc function from the irr R package (R Development Core Team, 2019). By convention, an ICC >0.90, 0.750.90, 0.500.75, and <0.50 indicates an excellent, good, moderate, and poor agreement, respectively.4