Multivariate analysis of skin impedance data in long-term type 1 diabetic patients

The purpose of the present work is to investigate the potential of multi-frequency skin impedance as a versatile method for early detection of diabetes-related changes in the skin. The method implies application of a small sinusoidal voltage at various frequencies through the outermost layer of the skin. Skin impedance measurement is a quick and easy alternative to the expensive and slow study of skin biopsies by microscopy. However, the amount of data is overwhelming and the measurement noise is rather high and therefore multivariate data analysis is used. Different models of data transformation and data reduction for extraction of the parts of the data that correlate with patient status and other diagnostic parameters are discussed. Multivariate methods make it very easy to give a visual interpretation of the results. The results demonstrate how a regression model between skin impedance and other diagnostic data for diabetic patients and control groups can be developed into a novel diagnostic tool for the early discovery of possible complications for diabetic patients.