One of the major challenges in precision medicine is to guide the research and development of specific therapeutic solutions through the extraction of knowledge from medical reports. These reports are often maintained as unstructured free-text and constitute a large volume of data. Knowledge and information extraction using Natural Language Processing (NLP) techniques could therefore help improve health care and medical decision-making. This stage is often preceded by a text segmentation phase to identify the part of text that are of great interest. In this article, we present our automatic segmentation system of medical reports into predefined categories and which consists of two complementary steps. First, an algotithm based on titles detection for the identification of certain sections. The second part is a supervised sentence classification task based on deep learning algorithms. This system was evaluated on 500 reports and achieved more than 96% classification accuracy.