Computer recognition of skin structures using discriminant and cluster analysis

Background/aims: Automated image analysis of complex tissues is usually lim ited by the difficulty of recognizing special structures by computer. The a im of this study was to test the applicability of discriminant and cluster analysis to the interpretation of skin images. Methods: Digital images from microscopic, dermatoscopic and clinical views of skin specimens were electronically dissected into elements of equal size and shape, and a set of grey level, colour and texture features was assess ed for each element. Elements were classified interactively and submitted t o discriminant analysis. Furthermore, hierarchical cluster analysis was use d to enable the system to classify the tissue elements automatically, based on the available digital information. The classification results were relo cated to the original image in order to evaluate the performance of the pro cedure. Results: The system performs well in reproducibly detecting different skin structures in digital images. Discriminant analysis of interactively classi fied elements yielded a correct reclassification in 98 to 100%of tissue ele ments. Among the cluster analysis procedures, the conservative Ward method after removal of all highly correlated features produced the best results. The method turned out to be applicable irrespective of the image source use d. Conclusions: Discriminant and cluster analysis may be helpful techniques fo r a user-independent, subjectively unbiased measurement system of skin stru ctures.