Objective measurement of radiation induced erythema by nonparametric hypothesis testing on indices from multivariate data

Three instrumental measurement techniques: Laser Doppler Imaging (LDI), Digital Colour Photography (DCP) and Near InfraRed (NIR) spectroscopy were tested for their potential to objectively measure radiation-based erythema in breast cancer patients. The irradiation dose intervals were 0, 8–16, 18–26, 28–34, 36–44 and 46–50 Gy. In addition, two types of skin lotion for reducing erythema were tested on the patients and these were compared to using no lotion. The measured results had very skew distributions for all three techniques making nonparametric testing necessary. The Wilcoxon Signed Rank Sum Test (WSRST) was used for this purpose. LDI was performed to produce univariate average perfusion values leading to a perfusion increment ratio. These ratios showed a good sensitivity to erythema, with a median detection limit of 18 Gy. DCP was used to extract average red–green–blue (RGB) values that were used in multivariate models. Results for a combination of principal component score values showed a marked increase in median erythema from 8 Gy on. The multivariate data from NIR spectroscopy were data-reduced to principal component scores and combinations of these were tested. The score combinations were used to show median detection limits down to 8 Gy. The difference between the lotions and using no lotion gave no significant result for the WSRST paired comparison for any used measurement technique.