Abstract: This article describes MIDAS, an advanced expert system that is able to suggest medical diagnosis from the radiological/clinical patient records, based on information extraction and machine learning from clinical histories of previously diagnosed patients. However, attempts to address the problem of free text processing have led to demand for software that simulates and complements what people are able to do. MIDAS was specifically designed to participate in the 2007 Medical Natural Language Processing Challenge MIDAS can be considered as one of the latest successors of MYCIN, the first expert system in history developed in the early 1970s at Stanford University, which was designed to diagnose infectious blood diseases The task of classifying physicians' diagnoses has been previously done.