The project is part of a global interdisciplinary context concerning brain molecular imaging. The objective is to develop statistical methodologies for the classification of textures using a Bayesian approach which combines random field modeling with Deep Convolutional Neural Networks.