This paper introduces an expert system to realize the diagnosis of medical diseases using case-based reasoning. Inference method and the case base are constructed respectively based on the knowledge of doctor and the questionnaire of patient. After putting the information of the patient for diagnosis into a fuzzy/crisp-frame and scanning the bases, the system can give a suitable diagnosis. The case bases and the reasoning algorithm are performed by using fuzzy method. Our experimental case bases employed real knowledge of Tuberculosis and Lung Diseases(case histories)collected as an example. The system would be used for medical diagnosis, or medical education, where the target diseases can be vary from many different fields. The system was developed in Delphi 3.0, uses Paradox database engine, and runs under Windows 9x.

View full abstract

The aim of our research is to simulate a thinking of doctor experts in diagnosis of integrating western and eastern medicine using fuzzy logic and interval computation. In this paper we present a design of a such fuzzy expert system using interval computation[3]. We extend the idea of Interval Based Expert Systems[1]for the diagnosis of Integrated Western and Eastern medicine. A general structure of the system is described. We focus more on knowledge representation and inference mechanism of the system. We show the way how to combine the different inferences of western and eastern medicine in diagnosis. The implementation of the system is briefly described. Finally, some conclusion and discussion are given.

View full abstract

A prototype CAIS(cardiac arrthythmia interpretation system)based on blackboard model is presented. The issue of automatic electrocardiogram analysis is addressed firstly. The idea of bi-directional interaction is led into CAIS, which makes the system run in a kind of way rather close to human mode of thinking. The process of PVC(premature ventricular contraction)diagnosis in CAIS is described to illustrate the effectiveness of the system.

View full abstract

Many people suffer from a continuous strong pain which is caused solely by the malfunction of the pain mechanism itself. One way to ease their pain is to electrically stimulate the spinal cord. Since the equations of pain are not known, we must use heuristic methods to find the optimal pain relief stimulation. In this paper, we show how fuzzy measures and similar nonlinear models can be used in pain relief control: they can be used to determine the parameters of the model which describes the dependence of the pain relief on the applied stimulation. Thus, fuzzy measures lead to the determination, for a given pain distribution, of the optimal pain relief stimulation.

View full abstract

We have established how to regulate tissue histology by using the Fuzzy Set Chaos Theory(BMFSA Sep.Sympo., pp^<1-2> 1995). We applied this theory to metastatic tumor cells that proliferate rapidly and migrate into adjacent tissue and organs. The basic flow chart of the histological regulation was to search effective molecules in database and to plane gene expression by programming to envelop a metastatic malignant cell by commanding fibroblast to move around the metastatic cells in the tissue. The ON commands we used were 1)synthesize collagens by ascorbic acid(VC), 2)increase collagen mRNA traslation by supplying collagen and calcium enriched small dried fishes after steam heating, 3)enveloped metastatic cells by combining collagen with elastin activated by calcium, 4)proliferate fibrogenic cells by thiamin(VB_1), riboflavin(VB_2), pyridoxine(VB_6), cobalamin(VB_<12>). A cut off supply of each molecule was used for the OFF command. Commands were transmitted by blood current, effluxed from blood vessels, parts of which with increased permeability induced by bradykinin, cytokines, and chemokines around the metastatic cells. These cytokines and chemokines transmit to fibroblasts the location of the metastatic cells. The effectiveness was monitored by X-ray photography or CT-scan and was judged by patients' viability.

View full abstract

This paper summarizes our research on developing a supporting expert system for pulmonary tuberculosis and lung-disease diagnosis, which has a fuzzy reasoning engine and a fuzzy-rule knowledge-base. The system has undergone through the theoretical and practical developments, which can be found in our earlier papers, and is now completely built. It has two new features as compared to those mentioned in our previous papers: capability of diagnosing more than just one disease, and a strengthened knowledge-base containing doctors' knowledge on diagnosing lung diseases: Pulmonary Tuberculosis, Lung Abcsess, Lung Cancer, Asthma, Pneumonia, Bronchiectasis, etc.

View full abstract

Applications of sensory information processing in neural networks have made significant and varied advances with regard to senses such as vision and hearing. Although similar research has been conducted on gustatory sensation, progress in these area lags behind due to the complexity of the mechanisms involved. If the complex gustatory mechanisms could be elucidated from an engineering standpoint, however, it could help in the medical field, for example, in the diagnosis of gustatory disturbances and treatment planning, while in the food industry it could hold promise for the production and processing of better tasting, healthier foods. Our group has therefore attempted an engineering analysis of the mechanisms of gustatory sensation. The present study examines a model for the evaluation of human taste; that is, how does our sense of taste determine "good" and "bad." In our experiments we used tofu as a model, since there are relatively strong taste preferences toward this food. The sensory evaluations were enforced by professional tasters, who have a developed sense of taste. Using a neural network with these evaluation results as the master data, the possibility of evaluating human taste evaluation from sensory information was evaluated. The results indicated that, using the present method and evaluation is indeed possible[1][2][3][4][5].

View full abstract

Advances in knowledge information processing technology have brought great promise to research of pathological diseases. Using oligosaccharides as intercellular information transmitters suited to discover the causes of onset of a disease, our group has employed neural networks in developing a support system for pathological diagnosis of liver disease, and obtained good results. When considering clinical application of such systems, however, the problem of the great amount of time needed for training, due to usual and unusual data that physiological data such as that obtained through oligosaccharides includes, remains when using conventional training methods to distinguish between liver diseases. Still, there is hope for further improvements in this method. To learn this type of data efficiently, in the present study we propose selected learning and expanded selected learning, and conducted a neural network pathological discrimination of liver disease. Finally, we show that our proposed methods are effective to train usual and unusual data.

View full abstract

Electro-gastrograms, which are records of electrical activities of the stomach detected from the body surface, are approximately periodic waves. Therefore, characteristics of electro-gastrograms cannot be represented appropriately by frequency analysis. In this paper we evaluated the chaotic nature of electro-gastrograms during rest and under stress, in an attempt to quantify electro-gastrographic traces using the chaos theory. Gastric electric response of 26 healthy adults under stress of mental arithmetic for seven minutes were analyzed using the chaos theory and FFT method. Analytical results are as follows. All values of the Lyapnov exponent determined from electro-gastrograms were positive, suggesting that Electro-gastrograms have a chaotic nature. The Lyapnov exponent was significantly reduced during mental arithmetic compared with rest. However, there was no difference in the dominant frequency obtained by FFT between the period of rest and period of stress From the analytical results, it was suggested that Electro-gastrograms have characteristics of chaos. Moreover, the Lyapnov exponent, which is an index of a chaotic nature, more sensitively quantified the effects of stress than conventional frequency analysis in the analysis of electro-gastrograms.

View full abstract

The satisfiability problem(SAT)of the propositional calculus is one of the most basic and important problems in the computer science. On the other hand the SAT is a famous NP-complete problem. It requires a lot of time to solve in general as the problem size becomes large. We proposed a recurrent analog neural network called LPPH for the SAT. The basic dynamics of the LPPH have the following properties: 1)every equilibrium point of the dynamics is a solution of the analog version of the SAT, called CONSAT, and vice versa, and 2)almost all equilibrium points have some kind of asymptotical stability. Experimental results show that the LPPH can solve the SAT effectively even if it is executed by numerical simulations on conventional computers. In this paper we introduce coefficients called "coefficients of attention" to the dynamics the LPPH and show that the performance of the LPPH for solving the SAT is improved.

View full abstract

Normally EEG-signatures are analysed in the frequency domain while a common FFT- or wave-let-transformation transforms the time signals of the electromagnetic activities of the(wet)neurons into the frequency domain. In this paper, we will show that better and faster results can be obtained, if instead of these common algorithm the FD-operation is applied to the pure time-signal and a DAS-calculation for eliminating the noisy contributions is done next. Especially we will compare the signatures of some epilepsy events analysed by common signal analysis and by the class-oriented DAS- and FD-operations and discuss the results.

View full abstract

Nobody questions the power of information dissemination of Internet. Likewise, nobody probably could say precisely neither the quantity of information present on the Net today, nor of what kind it is. However, such volume of information provokes some rejection from many potential users that no longer feel able to discern or find out where is the information of their interest. The fact is that the search engines, whose features are better and better, are based on criteria classifying the information present on the Net and not on the criteria or interests of the users who, in general, have to use rather poor mechanisms. In this paper, we will present one of the new lines of Internet effective use, based on the modeling of the information in general and of the users' profiles. These models are currently under study in the European Commission ESPRIT Project n°P29158 called FLEX, "Flexible Knowledge-Based Information Access and Navigation using Multimodal Imput/Output". We will illustrate some proposals with examples in the Medicine field.

View full abstract

This paper extends the concept of functional dependency in a database in which the presence of context null values is allowed. It is shown that the set of Armstrong's inference rules forms a sound and complete axiom system for functional dependencies with context nulls as well. Some rules for the data update procedures are also introduced and examined to ensure that the database under consideration still satisfies a given set of functional dependencies.

View full abstract

Today, deficit constitution is a serious problem in hospital management. In this paper the new methodology to decide investing budget under the consideration of its risk, which enables us to earn stable profit in the hospital management. Mean-variance analysis, which maximizes the expected profit and minimizes its risk, is employed in this paper. Neural network is applied to solve the mean-variance analysis. In this paper, the mean-variance model is transformed into the energy function to solve itself by a Boltzmann machine, and the neural approach can provide us with highly approximate solution. As an explanatory example, we discuss 1)how to distribute the budget over a virtual hospital group which consists of the main hospital, several subordinate hospitals, and 2)how to decide the production of medical products in manufacturing.

View full abstract

Recently, the identification of chaotic phenomenon is required in research of dynamic system such as biological behavior in human body. Extracting the most similar chaotic pattern to the standard in the chaotic like signal, for example, detected from human body, can identify the condition of biological behavior. The processing of 2-dimensional line figure can be applied to the chaotic pattern matching, by dealing with the 2-dimensional chaotic attractor. By corresponding the x and y-coordinate of figure to the real and imaginary part of complex number, respectively, the line figure can be regarded as the complex valued function with real variable. The line figure is analyzed by the cross correlation function of the two functions which show the each line figure. In this paper, the analysis of 2-dimensional line figure and the method of applying to the analysis of 2-dimensional attractor are described. Some 2-dimensional attractors are generated by various mappings from a chaotic phenomenon. The influence of mapping is reported by calculating the similarities of 2-dimensional attractors in experiment.

View full abstract