Sources of unwanted variability in measurement and description of skin surface topography

Background/aims: Quantitative measurement of skin roughness has proved to be a valuable tool in the efficacy‐control of external applications, but it suffers from not yielding easily comparable results. The most important sources of inter‐observer variability are high‐pass filters used to separate roughness and waviness, and low‐pass filters which result from the finite resolution of the instrument or from the finite sampling interval of digital measurement. In the present study, the effects of high‐pass filters and sampling intervals on the roughness measured were investigated. Methods: Dynamically focusing optical profilometry was used to measure the surfaces of negative replicas of healthy human skin. High‐pass cut‐off wavelengths and sampling intervals were varied systematically. Results/conclusions: Virtually unbiased estimates for the roughness parameters K, Sk, R q , and R a can be obtained using sampling intervals of 40 or even 80 μm. Regarding these roughness parameters, it is far better to do more scans than to shorten the sampling interval. The roughness parameters R z , R p , R t , R pm , R max , P t , on the other hand are very sensitive to the influence of the sampling interval; to achieve satisfying estimates, the sampling interval should be no longer than 2 to 5 urn; as an important parameter’of the measurement, it is worthy of remark and should always be indicated. The way the mean square roughness R q depends on the cut‐off wavelength is not well described by the Sayles‐Thomas‐relation R q ∼λ c 0.5 . If the power‐spectrum | h* (v)| 2 approximates sufficiently to a power law, | h* (v)| 2 ∼v δ , a better estimate is given by R q ∼λ c γ with γ=‐(δ+1)/2. In many cases, γ=1 or R q ∼λ c will suffice.