Hopfield neural network with prespecified time convergence for the segmentation of brain MR images

We present contributions to improve a previously published approach for the segmentation of magnetic resonance images of the human brain, based on an unsupervised Hopfield neural network. We formulate the segmentation problem as the minimization of an energy function constructed with two terms, the cost-term as a sum of squared errors, and the second term temporary noise added to the cost-term as an excitation to the network to escape certain local minima with the result of being closer to the global minimum. Also, to ensure the convergence of the network and its utilisation in the clinic with useful results, the minimization is achieved with a step function which permits the network to reach stability corresponding to a local minimum close to the global minimum in a prespecified period of time. We present segmentation results of our approach for data of patient diagnosed with a metastatic tumor in the brain, and we compare them to those obtained from, previous work using Hopfield neural networks, the Boltzmann machine and the conventional ISODATA clustering technique