Reliability analysis of a novel measurement system for quantifying human skin color.

Abstract

Precision is crucial in determining the appropriate procedure for implementing further trials. We conducted a study to explore the reliability of a novel measuring system for human skin color. The novel skin color measuring system was used to capture the skin color of four volunteers (2 males and 2 females) from the same location on each subject by the same operator. The measurement was repeated for different poses and instrument factors (camera and shooting protocol) in the red, green, and blue (RGB) system. The average color depth in each image was calculated and converted from 0 to 255. The spread of measures and the Bland-Altman plot was displayed to determine each variance source's random error, with the interclass correlation coefficients applied to reflect the reliability. The RGB color depth in the experiment ranged from 190, 152, and 122 to 208, 170, and 142. The 95% confidential interval of the differences from the means in RGB colors for the different protocols were ±2.8, ±2.6, and ±2.1, respectively. The largest variation in the replicate trials was observed when subjects were in a supine position (standard deviation: 2). The interclass correlation coefficients were greater than 90%, suggesting that the developed system is highly precise. This study demonstrated that the developed device could stably and reliably detect human skin color across different common sources of variation, and thus could be applied clinically to explore relationships between health/disease and skin color changes.