Neuron Networks in Recognition the Brain Lesions.

In this paper, we principally present the recognition stage of computer vision system for helping the doctors in tracking the brain lesions. We briefly present the algorithms used for image filtering, lesion detection, compute the morphological characteristics of lesions. The proposed algorithms are based on various theories such as: morphological geodesy and genetic algorithms. We present also in details the recognition stage. For that, we propose to use the neuron networks. They are specified by their capacity of training and generalization. We choose the multi-layer neuron network with retropropagation training since it allows a supervised training for exploiting the information about lesion. Our algorithms are tested on several brain images, we give one example result