Studies On Medical Knowledge Assembly Techniques And An ANN-GA-based Clinic Decision Support System

A medical expert system (MES) attempts to simulate the thoughts and the processes of medical experts when making medical diagnose based on the principles and methodologies developed in the fields of artificial intelligence and expert system, and it can help doctors solve complicate medical problems as an important assistant tool of diagnosis, therapy and prognosis. Medical expert system is a closed combination of knowledge-based engineering and medicine. It is also a crossing of medicine and computer technique, network technique, communications technique and database technique et al, and it has become a significant component of modern medical science.In the development of the knowledge-based systems (KBS), very often knowledge acquisition (KA) becomes a leading problem to efficiently build and expand knowledge bases with high accuracy. Conventionally, knowledge acquisition is accomplished through close co-operations between medical experts and knowledge engineers, which gives a heavy labor burden for both of them. In addition, the speed of conventional knowledge acquisition is quite low because significant amount of time and effort are required to create and maintain these knowledge bases, which blocks the research of medical expert system. With the development of computer technique and network, some medical knowledge acquisition tools are exploited to solve the choke point. However, the knowledge acquired by these tools is hard to apply widely and easily due to their complicate inference and representation. Inspired by the effectiveness of industry assembly techniques, a conceptive of medical knowledge assembly techniques is proposed, which means the process of constructing a large-scale medical knowledge base should like a "product line", this line extends from medical knowledge acquisition and representation (construction of knowledge base) to knowledge complementarity (maintain of knowledge base), then to application of knowledge (medical expert system), finally to feedback of knowledge (renewal of knowledge base). Therefore, the author proposes a network-based and reliability-based knowledge representation system to acquire medical knowledge. The author demonstrates that a medical knowledge acquisition/management system can be built upon a three-tier distributed client/server architecture. The knowledge in the system is stored/managed in three knowledge bases, and the maturity of the medical know-how controls the knowledge flow through these knowledge bases. In addition, to facilitate the knowledge representation and application in these knowledge bases as wellas information retrieval, an 8-digit coding scheme with a weight value system is proposed. Current results have showed that the method is a viable solution to construct, modify, and expand a large-scale medical knowledge base through the network.With the development of artificial intelligence, machine-learning-based clinic decision support systems (CDSS) develop quickly. Many methods, such as rule synthesis, decision tree, case-based learning and Bayes network and so on, have been applied in clinic decision support systems. In this paper, the author presents a diagnostic decision support system based on the multilayer perceptron (MLP) neural network architecture for diagnosing five common heart diseases (coronary heart disease, rheumatic valvular heart disease, hypertension, chronic cor pulmonale and congenital heart disease). In particular, 40 input variables (sysmptoms and manifestations) related to the diagnosis of the heart diseases of interest are categorized into four groups and then encoded using the proposed coding schemes. The system is trained using a back propagation (BP) algorithm augmented with the momentum term, the adaptive learning rate, the forgetting mechanics, and an optimized algorithm based on conjugate gradients method step by step. A total of 352 medical records of patients suffering from various heart diseases have been collected to test the system. Three assessment methods, cross validation, holdout and bootstrappi...