Neural Networks and Application on Psychiatry Dataset for Classification Diagnosis

Abstract

Objective: Diagnosis of diseases is among the major problems in medicine. Recently, artificial neural networks are widely used in medical diagnostic problems. In this study, disease dataset was modeled by using various neural networks. Among these networks, the five different learning algorithms of multi-layer perception (MLP), which gives the best results, were tested and the real performances of the algorithms were shown. Material and Methods: In this study, by the aid of artificial neural networks methods, demographic characteristic of 196 patients who presented to the Gaziosmanpasa Hospital from January 2008 to April 2008 with various complients we re recorded and they were given the Back depression scale and beck anxiety scale to make a diagnostic classification and to assess how compatible this classification was with the initial diagnosis, which was based on DSM-IV-TR diagnosis criterion. Results: Of the participants, 51 (26%) were males and 145 (74%) were females; 130 (63.3%) were married; 79 (40.3%) were primary school and 67 (34.2%) were high school graduates; and 97 (49.5%) were housewives. Ninety-three participants (47.4%) depression and 52 (26.5%) had anxiety disorder. Conclusion: Beck depression scale should be used whenever there is a suspician of depression in the initial assessment.