Computer-supported diagnosis of melanoma in profilometry

Laser profilometry offers new possibilities to improve non-invasive tumor d iagnostics in dermatology. In this paper, a new approach to computer-suppor ted analysis and interpretation of high-resolution skin-surface profiles of melanomas and nevocellular nevi is presented. Image analysis methods are u sed to describe the profile's structures by texture parameters based on co- occurrence matrices, features extracted from the Fourier power spectrum, an d fractal features. Different feature selection strategies, including genet ic algorithms, are applied to determine the best possible subsets of featur es for the classification task. Several architectures of multilayer percept rons with error back-propagation as learning paradigm are trained for the a utomatic recognition of melanomas and nevi. Furthermore, network-pruning al gorithms are applied to optimize the network topology. In the study, the be st neural classifier showed an error rate of 4.5% and was obtained after ne twork pruning. The smallest error rate in all, of 2.3%, was achieved with n earest neighbor classification.