Abstract

A training problem exists in the field of psychiatry since the traditional methods of varying effectiveness in diagnosing and selecting treatment options are passed onto new medical residents, learning from established professionals. The lack of consensus within the psychiatric community towards diagnosis and treatment selection probably impacts the quality of patient care. Difficulties in the diagnosis processes may lead to inaccurate diagnoses and delays in administering timely and effective treatment alternatives. To address this educational need, it is proposed that the development of a multi-layered neural network application with back-propagation facilities that would be taught to categorise the various forms of schizophrenia and the various treatment options should be created as a tool to assist medical residents in the diagnosis and treatment selection for schizophrenia. An overview of neural networks, psychiatric disorders, the steps in the diagnostic process and current assessment tools for schizophrenia is provided. The feasibility of developing a neural network as a diagnostic and treatment selection tool for schizophrenia is assessed, taking into account current neural network's limitations. It is concluded that the lack of consensus amongst psychiatric professionals towards current diagnosis and treatment selection process will result in data inadequate to train such a neural network, thereby resulting in potentially false results. The long-term methods to overcome these limitations and some alternatives towards providing a short-term computer-based educational solution are presented.