In clinical diagnosis, ultrasound is superior to X-ray in respects of safety and ability to image characteristic of soft tissues. A pulse-echo ultrasonic imaging is now widely applied to the clinical use. Unfortunately, however, the images are not quantitative and their resolution is not sufficient since the imaging uses reflected ultrasound signals on the assumption that the ultrasonic beam is very sharp and travels in straight lines. The assumption is not true in all biological tissues. Images of ultrasonic CT by using technique of X-ray CT also have artifacts because of the same reasons.
Diffraction tomography, which is CT considering wave diffraction effect, is expected to reconstruct quantitatively a higher resolution image. In clinical use, however, the range of insonifying angle is often limited. It is therefore an important problem how accurate image under this ill-posed condition can be reconstructed. In regard to this problem, we have proposed a new restoration method by using maximum entropy method and iterative revisions. Simulation results have shown that it is possible to restore the image under this ill-posed condition.
In this paper, we present the results of experimental investigation on this method. The results coincide with the same tendency as simulation results.

抄録全体を表示

Extracting the contour of an object in image is one of the main tasks of image processing. Snakes, active deformable contours, are well known as optimizing approach. Practically, the objective function is the potential defined over image plane. The local minima of the potential are interpreted as “attractor” of the snake. Though this interpretation seems quite natural, actual potential function has complicated shape especially near the boundary of an object, because of shadow, blurring as well as noise. So snake is often trapped in a fake equilibrium, and cannot detect edges or curves correctly. We apply stochastic relaxation method to this problem. Our algorithm guarantees that the snake can achieve true equilibrium within selected relaxation range. The efficiency of our algorithm is demonstrated via some examples.

抄録全体を表示

A biosensor is needed to be light, strong, durable, as well as to satisfy the safety conditions against virulence, stimulation, carcinogenicity and allergic reactions. Carbon is one of the most suitable elements which has the highest blood compatibly, and have been used as an intrvascular implant material. Because of these high degree of adaptabilities, carbon is expected to apply as an in-vivo bio-sensor. In this paper, we developed a new bipolar biosensor using advanced PAN-based carbon fibers (CF). The CF electrodes have been estimated to have enough sensitivity and stability as a nerve signal detector presumably because of the high degree of affinity of carbon fiber to an organism. The CF biosensors were set on the sympathetic nerves of small test animals, and the electric signals of sympathetic nerves activity for the hepatic, the cardiac and renal were detected. The fast Fourier transform (FFT) was carried out on the signals, and the FFT power spectrums were lined up along the time axis in order to transfer to color images of the signals. It has been suggested that the images indicate the clear information from noisy original signal and evaluated to be useful, especially to analyze the influence of bleeding. It is shown that the sympathetic nerve activity of the each viscera is analyzed in detail by CF sensor combined with this image processing system.

抄録全体を表示

Measurement of P₃₀₀ which is a component of evoked potential (EVP) is important for clinical diagnosis. P₃₀₀ component is usually estimated by summing up a large number of responses for reducing the background wave such as alpha activities since its amplitude is smaller than background wave. The present method therefore unable us to analyze the time-varing state of response in detail.
In this work, we propose a new method of estimating P₃₀₀ component from a single trial response using the trend model. Assuming that each trial response consists of EVP, background wave and noise component, we model the EVP as the trend of time series which varies stocastically, background wave as AR process and noise as white. Considering the properties of three components Kalman-filter is constituted in order to estimate three components from a single trial response. The performance of the method is evaluated using model waves and mesured signals of event related potentials. Results showed that the method have the ability to estimate P₃₀₀ component precisely even in the case of low S/N.

抄録全体を表示

In recent years, numbers of traffic accidents are caused by a drowsiness during vehicle driving. For preventing the traffic accidents, there are increasing demands to detect a decline in the driver's arousal level before drowsing. The purpose of this research is to develop a system for detecting the decline in the arousal level without equipping any sensor on the driver. According to various reports, as an arousal level reduces, a blinking duration becomes longer. This paper describes the techniques for measuring the blinking durations using the driver's facial images taken by TV camera. We present a method for taking the facial images which is less affected by the changes in the ambient lights. In this method, we take the images using the electric shutter function of a TV camera and the pulse emission of an infrared light source at the same time. We also present a method for extracting the blinks which is less influenced by the differences of the shape around individual eye.

抄録全体を表示

We have suggested that the systolic coronary circulation is dominated by optimal control principle. The coronary circulation was expressed by electrical circuit model which was comprised of extra myocardial, epicardial and end cardial compartments each of which have arterial and venous sides. The optimality was expressed by minimizing the pressures, flow rates at each compartment and aortic pressure, intra myocardial pressures. The optimal solution could reproduce, 1) arterial reversal flow, 2) 2 phases forward flow in venous system, 3) augmentations in reversal flows in arterial system by increased resistances and decreased compliance of vessels and 4) physiological values of aortic pressure and intra myocardial pressures.

抄録全体を表示

In the conventional ultrasonic pulse Doppler method, target velocity is obtained from the Doppler-shift at the center frequency of the transmitted ultrasonic pulse wave. In this method, since Doppler signals are sampled by the pulse repetition frequency, there is a limit of the measurable velocity due to the sampling theory. This limit gives a serious problem to measure fast velocity components in the practical applications.
In this paper, we propose a new method for determining the Doppler-shift not only at the center frequency but also in the frequency band around it, so that velocity components which are faster than the velocity limit in the conventional method, can be successfully measured. The principle of the new method is confirmed by the theoretical consideration and the experiment in a water tank. The velocity limit of the new method is also investigated theoretically.

抄録全体を表示

This paper presents a fuzzy diagnosis system for fatty liver using ultrasonic imaging. Firstly this system extracts the unit regions of both liver and kidney from original ultrasonic image, and then calculates the possibility of fatty liver by using “Liver-Kidney Contrast”. The system utilizes fuzzy inference in two parts. In the former part, the system utilizes the information of two organs' location after classifying each unit region into two categories, i.e. kidney and liver, with 5 image features, which are fractal dimension, two local fractal dimensions, average and contrast. In the latter part, the system calculates “Liver-Kidney Contrast” using regions of liver and kidney that extracted by the method in the former part. Then the possibility of fatty liver is calculated. 16 cases were applied to evaluate the performance of our system. 10 fatty liver cases and 6 normal liver cases had been diagnosed by a doctor for the reference. Our system diagnosed 7 fatty liver cases and all 6 normal liver cases had been diagnosed by a doctor for the reference. Our system diagnosed 7 fatty liver cases and all 6 normal liver cases correctly. The accuracy is 81.3%. Therefore, our system is considered to be effective for supporting ultrasonic imaging diagnosis in case of screening fatty liver.

抄録全体を表示

Measurements and analyses of chlorophyll fluorescence transients (induction) curves in vivo have been a powerful tool in monitoring nondestructively the photosynthetic apparatus injury caused by environmental stresses. Recently the intracellular potential and the 685-nm chlorophyll a fluorescence of intact leaves have been found to change quickly in response to high concentration CO₂gas exposure: the quick response properties may be used for CO₂ gas sensing. In this work, we demonstrate the quick response measurements in vivo of intact leaf to chemical stresses on the basis of Ar laser-induced chlorophyll fluorescence induction, where a fluorescence intensity emitted at 685nm is measured continuously with time. The chlorophyll a fluorescence induction curve and the surface potential of intact leaves of scheffiera octophylla have been found to change quickly in response to 100% CO₂ gas and acetic acid vapor exposure. The difference in chlorlophyll fluorescence induction response between chmical stresses is discussed in detail.

抄録全体を表示

An image obtained by ultrasonic medical equipment is poor in quality because of speckle noise, that is caused by the quality of ultrasonic beam and so on. Thus, it is very difficult to detect internal organs or the diseased tissues from a medical ultrasonic image by the processing, which is used only gray-scale of the image. To analyze the ultrasonic image, it is necessary to use not only gray-scale but also appropriate statistical character. In this paper, we suggest a new method to extract regions of internal organs from an ultrasonic image by the discrimination function. The discrimination function is based on gray-scale and statistical characters of the image. This function is determined by using parameters of the multi-dimensional autoregressive model.

抄録全体を表示

Information of sleep stage is one of the most important clue for diagnosis of mental condition and disease in psychiatry. So far we have proposed a method of detection of characteristic waves in sleep EEG and diagnosing the sleep stages of the segmented short-terms by neural networks analysis. And we showed that it was able to diagnose the sleep stages to some extent by recognizing the time-varying spectral patterns of characteristic waves. There remains, however, a problem that results in stage diagnosis often become unstable, since the contextual relation between the present and adjacent segments is not considered. In this work, a method of diagnosing the sleep stage more accurately is proposed and its performance is evaluated. In the method, the additional neural networks processing is combined with the previous system for recognizing the context of stage sequences. As a result, it is proved that detection ratio is improved to a considerable extent by utilizing the contextual information on stages and the proper duration exists for obtaining high performance.

抄録全体を表示

The quantitative evaluation of hepatic fibrosis on histological sections is of great interest for the staging and follow-up of chronic liver disease. In this study, an imaging system based on personal computer is used to measure the area of fibrosis, and the results for 22 biopsy specimens are well correlated with the clinicopathological findings. For the morphometric analysis of fibrotic area, we employed the fractal dimension estimated from the differential box counting method. The results of analysis for 31 biopsy specimens, i.e. 160 fibrotic areas, show that the fractal dimension method is useful for the staging of fibrotic changes in the histological findings for liver biopsy specimens, though continuous work is need to improve the staging accuracy.

抄録全体を表示

Metallic thin films are widely used in microelectronics devices such as inter-connections of electric circuit and electrodes of the devices. As a novel process of spatial selective deposition of the metal thin film, self-aligned deposition of copper thin films with combination of krypton fluoride excimer laser doping and succeeding electroplating has been demonstrated. The N-plus layer of N type gallium arsenide is formed with the incorporation of high concentration silicon atoms using the reduction-projection laser doping system. In the electroplating process, high-contrast selective depositions of copper thin films have been obtained. The contrast ratio of selective deposition is estimated to exceed 10⁷, calculated as the ratio of the current densities on the doped and undoped semiconductor surfaces. The contact between the deposited copper thin film and gallium arsenide has revealed ohmic characteristics. The specific contact resistance is estimated to be 2.32×10^-5ohm-cm², which is about 1/30 as small as that of the conventional alloyed ohmic contacts.

抄録全体を表示

Medical treatments for hypertension have made to avoid myocardial infarction and cerebral hemorrhage. Over-dosage is dangerous. It causes the reduction of blood pressure during sleep. It is important to presume the variation of blood pressure after drug administration from the fluctuation of blood pressure before drug administration to prevent a secondary reaction and to improve the quality of life. In this paper, a 24-hour mean blood pressure in 12 essential hypertensive subjects was measured. Power spectrum and phase spectrum in mean blood pressure time series are obtained by a FFT. Power spectrum and phase spectrum were discussed before and after drug administration. According to the relationships of power spectrum and phase spectrum before and after drug administration, mean blood pressure time series after drug administration would be presumed.

抄録全体を表示

In this paper, an application of genetic algorithm to a feedforward controller of speed servo and positioning systems is considered. The system consists of a conventional feedback controller and a feedforward controller with a genetic algorithm. Parameters of the feedforward controller are regarded as a chromosome of gene. The genetic algorithm is used to tune the parameters of the feedforward controller, the objective of which is to make the feedforward controller a dynamic inverse of the plant. The tracking error between the desired output and the actual output every sampling period is used to calculate the fitness of one gene. The range of mutation is dependent on the magnitude of the tracking error, which ensures the convergence of the parameters. The effect of the fitness function and the range of mutation on tracking performance is investigated. The experimental results demonstrate the effectiveness of the proposed genetic algorithm.

抄録全体を表示

The grades of table meat in Japan are authorized by Japan Meat Grading Association. The carcasses are inspected from two points of view such as a yield of meat and meat quality, and are graded by visual inspection of experts, so called grader. It has been difficult that we build a automatic system grading the carcasses (beef) by image processing techniques instead of grader's visual inspection, because beef grading depends on human sense of the graders.
In this paper, the authors propose new techniques such as a fuzzy shading correction, fuzzy separation of color image data, and hybrid neural network as a tool of image processing. Using the image processing with the new techniques, we construct a system which automatically grades the carcasses in accordance with the standard of Japan Meat Grading Association.

抄録全体を表示

In this paper, calculation method of higer order derivatives of Universal Learning Network (U.L.N.) which can be used as a fundamental tool to model and control large-scale complicated systems such as economic, social, and living systems as well as industrial plants is proposed. U.L.N. has an ability of learning and can optimize even the broad range systems which are difficult to be expressed as first order difference equations with one sampling time delay, First derivatives calculation method of U.L.N. has been already proposed and these have been clarified to be effective to system optimization, In this paper, extending first order theory, nth order derivatives calculation theory is presented and as an example of applying higer order derivatives to system control, a new technique is introduced to control chaotic status of U.L.N. using Hessian Matrix.

抄録全体を表示

Shaft inserting is frequent and important operation for automatic parts assemble. We have been developing a sensor that detects in real time a position of hole center and its inclination using a 2-dimensional PSD. The sensor can detect the hole position by signals for the center of blight part. When the hole is inclined, a black ellipse image is focused on the PSD, the total current of one pair of PSD electrode differs from that of another pair. Therefore the sensor can detect hole inclination by this current difference. The sensor has about 20μm and 0.3 degree resolution for the position and the inclination sensing respectively.
We developed an automatic shaft inserting system applying the position and inclination sensor to a parallel joint robot with four freedom. The output of sensor is sent to PC-98 host computer which evaluates dislocation and inclination values. These values are sent to the robot to correct the shaft position and its inclination.
Within 5 sec the robot can insert a shaft into a hole with 50μm of clearance and without chamfering with 2mm dislocation and 3 degree inclination.

抄録全体を表示

We propose an error control method to implement encoders and decoders inserting one byte parity between information symbols inside each C₁ codeword. One byte parity is designed by RS code over GF(2⁸) with minimum distance 2. The parity prevents C₁ codeword from miscorrecting for C₁ decode. Hardware to implement the parity encoder is devised to make C₁ check words by encoding information words which include parity words. Numerical results show that the probability of miscorrection is restrained enough under the detectable level, and good coincidence with simulation results.

抄録全体を表示

This paper presents an approximation method for obtaining a feasible solution of a very large-scale assignment problem with priority order. The proposed method is applicable to assignment problems whose cost coefficients cannot clearly be determined. In the method, a feasible solution is rapidly found without considering priority order, and then resources are arranged to take priority order into consideration. This causes nonzero value of the sum of infeasibilities. Hence, the following three steps are iteratively carried out to decrease the sum monotonically;
(i) a set of integer variables to be changed is chosen by means of search trees,
(ii) the trees are modified after the integer variables are changed, and
(iii) the assignment is altered so that the number of deleted arcs is as small as possible.
If the sum results in zero, resources are rearranged to satisfy higher priority order. The above procedure is repeatedly performed until the sum doesn't become zero furthermore. The method has proved to be efficient as a result of applying it to very large-scale assignment problems, especially which have exceedingly large infeasibility amounts.

抄録全体を表示

This paper proposes an information retrieval interface that provides the user not only with the requested data but also with other helpful information in linguistic forms. When there is no data that satisfies the user's request, the system shows what kinds of data are stored in the database and provides the closest one to the request as an alternative. It also shows possible compromises that the user must make if he is not satisfied with it. When too many data to examine are retrieved, it suggests a hint to reduce the retrieved data set. With these characteristics of the system, the user can easily constuct successive requests and get the best data for him. The paper starts with construction of a “macro database”, which is knowledge about distribution of the data and is expressed in fuzzy clusters with linguistic labels. Then, it discusses how to generate the helpful information described above. Finally, an application to an actual realestate database shows the effectiveness of the proposed approach.

抄録全体を表示

Response characteristics of the electronic excitable membrane model by Tsukada, et al. to perodic pulse stimuli are investigated. Use of phase plane analysis and of phase transition curve extended by Maginu is made in order to clarify the characteristics. The global bifurcation structure of the model resembles to that of squid's giant axon. It is shown that the global bifurcation structure varies with stimulus pulse width. Shorter pulse width leads to simple behavior, and longer pulse width is likely to produce various behavior.

抄録全体を表示

The earth potential difference (EPD) has been observed continuously for monitoring an earthquake at the urban area. A new measuring method of EPD is proposed to reduce the unnecessary signals induced by artificial and natural earth current sources. On this method EPD is measured through two electrodes that are buried at vertically different positions. On the earthquakes of more than 1 of intensity scale we can clearly observe EPD signals. These P-wave arraival times agree exactly with ones obtained by meteorological agency seismograph. The obtained data show that electrical field of vertical direction is higher by one order than one of horizontal direction.

抄録全体を表示