Clinical diagnosis support system based on symptoms and remarks by neural networks

The importance of developing a support system for medical decision-making in the clinical field has been pointed out in order to improve accuracy and objectivity of the judgment of clinician, and to detect early the disease. Various attempts on clinical diagnosis support have been done in the literature, e.g. rule-based production system, analysis of ECG or the image, focusing on only specific diseases, so those can not utilized for the general internal medicine diagnosis support. In this paper, it is noticed that signs, symptoms, remarks, clinical laboratory data such as blood and biochemical check up data are essential information which show functional depression and failure of the ecosystem, and a practical software system providing the physicians with clues to clinical diagnosis. First, after analysis of the medical task and structuralization of information from the patient, the medical modelling method is provided, and it is indicated that diagnosis can be considered the identification of patient who corresponds to the controlled object. Secondly, by the concept of patient model and disease model, the system identification algorithm is proposed, and actual system constructed for the medical diagnosis is described using neural networks. The diagnaccuracy of this system using symptoms and remarks data from medical journal case reports is also described