An unsupervised bayesian approach for the joint reconstruction and\n classification of cutaneous reflectance confocal microscopy images

Abstract

This paper studies a new Bayesian algorithm for the joint reconstruction and\nclassification of reflectance confocal microscopy (RCM) images, with\napplication to the identification of human skin lentigo. The proposed Bayesian\napproach takes advantage of the distribution of the multiplicative speckle\nnoise affecting the true reflectivity of these images and of appropriate priors\nfor the unknown model parameters. A Markov chain Monte Carlo (MCMC) algorithm\nis proposed to jointly estimate the model parameters and the image of true\nreflectivity while classifying images according to the distribution of their\nreflectivity. Precisely, a Metropolis-whitin-Gibbs sampler is investigated to\nsample the posterior distribution of the Bayesian model associated with RCM\nimages and to build estimators of its parameters, including labels indicating\nthe class of each RCM image. The resulting algorithm is applied to synthetic\ndata and to real images from a clinical study containing healthy and lentigo\npatients.\n