Document Layout Analysis is a fundamental step in Handwritten Text Processing systems, form the extraction of the text lines to the type of region where it belongs. We present a system based on artificial neural networks which is able to extract not only the baselines present in the document, but geometric and logic layout analysis of the document as well. Experiments in three different datasets demonstrate the potential of the method and show competitive results with state-of-the-art methods.