Yukinori Suzuki

Self-organizing QRS-wave recognition in ECG using neural networks

The author has developed a self-organizing QRS-wave recognition system for electrocardiograms (ECGs) using neural networks. An ART2 (adaptive resonance theory) network was employed in this self-organizing neural-network system. The system consists of a preprocessor, an ART2, network, and a recognizer. The preprocessor detects R points in the ECG and divides the ECG into cardiac cycles. A QRS-wave is the part of the ECG that is between a Q point and an S point. The input to the ART2 network is one cardiac cycle from which the ART2 network indicates the approximate locations of both the Q and S points. The recognizer establishes search regions for the Q and S points. Then, it locates the Q and S points in each search region. The system uses this method to recognize a QRS-wave. Then, the ART2 network learns the new QRS-wave pattern from the incoming ECG. The ART2 network self-organizes in response to the input ECG. The average recognition error of the present system is less than 1 ms in the recognition of the Q and S points.

Soft Computing in Industrial Applications

The research has developed a fuzzy logic approach to handling missing data. A prototype fuzzy model was developed, using the FuzzyTech software, to assess the quality of the steel production in terms of composition, time, and temperature. As tools like FuzzyTech are not able to handle missing data, the research has introduced a fuzzy logic approach to decision making with less data. A number of workshops were carried out in the plant, and the aired expertss knowledge was the basis for the researchs development. This paper will present the state of the art research on the application of artificial intelligence and statistical techniques for handling the missing data problem

Personal computer system for ECG ST-segment recognition based on neural networks

A personal computer system for electrocardiogram (ECG) ST-segment recognition is developed based on neural networks. The system consists of a preprocessor, neural networks and a recogniser. The adaptive resonance theory (ART) is employed to implement the neural networks in the system, which self-organise in response to the input ECG. Competitive and co-operative interaction among neurons in the neural networks makes the system robust to noise. The preprocessor detects the R points and divides the ECG into cardiac cycles. Each cardiac cycle is fed into the neural networks. The neural networks then address the approximate locations of the J point and the onset of the T-wave (Ton). The recogniser determines the respective ranges in which the J and Ton points lie, based on the locations addressed. Within those ranges, the recogniser finds the exact locations of the J and Ton points either by a change in the sign of the slope of the ECG, a zero slope or a significant change in the slope. The ST-segment is thus recognised as the portion of the ECG between the J and Ton points. Finally, the appropriateness of the length of the ST-segment is evaluated by an evaluation rule. As the process goes on, the neural networks self-organise and learn the characteristics of the ECG patterns which vary with each patient. The experiment indicates that the system recognises ST-segments with an average of 95·7 per cent accuracy within a 15 ms error and with an average of 90·8 per cent accuracy within a 10 ms error, and that characteristics of the ECG patterns are stored in the long term memory of the neural networks.

Near optimum estimation of local fractal dimension for image segmentation

This paper presents an algorithm for estimating the local fractal dimension (LFD) of textured images. The algorithm is established by an experimental approach based on the blanket method. The proposed method uses the near optimum number of blankets to obtain the LFD for a small local window. The robustness of the proposed method to consistently estimate the LFD using up to a 3 × 3 local window is confirmed by experimental evaluations. The LFD maps, created from natural scenes, are utilized in an image segmentation algorithm that demonstrates the capability of rough segmentation of fine-texture regions in natural images.

SEGMENTATION OF NATURAL IMAGES USING ANISOTROPIC DIFFUSION AND LINKING OF BOUNDARY EDGES

We present a segmentation method of natural images that uses an anisotropic diffusion algorithm and a region growing algorithm. We propose a modified version of the anisotropic diffusion algorithm as a precise edge-preserving smoothing technique modified by using boundary edges. We incorporate a linking algorithm for boundary edges based on a directional potential function into the anisotropic diffusion algorithm to improve the ability of edge-preserving smoothing. As a result, unnecessary details of images are effectively smoothed before performing a region growing algorithm. Therefore, the proposed method is suitable for an accurate segmentation of natural images. Several simulated examples are presented that demonstrate the effectiveness of the proposed technique.

Vector quantization of images with variable block size

We proposed a vector quantization (VQ) with variable block size using local fractal dimensions (LFDs) of an image. A VQ with variable block size has so far been implemented using a quad tree (QT) decomposition algorithm. QT decomposition carries out image partitioning based on the homogeneity of local regions of an image. However, we think that the complexity of local regions of an image is more essential than the homogeneity, because we pay close attention to complex region than homogeneous region. Therefore, complex regions are essential for image compression. Since the complexity of regions of an image is quantified by values of LFD, we implemented variable block size using LFD values and constructed a codebook (CB) for a VQ. To confirm the performance of the proposed method, we only used a discriminant analysis and FGLA to construct a CB. Here, the FGLA is the algorithm to combine generalized Lloyd algorithm (GLA) and the fuzzy k means algorithm. Results of computational experiments showed that this method correctly encodes the regions that we pay close attention. This is a promising result for obtaining a well-perceived compressed image. Also, the performance of the proposed method is superior to that of VQ by FGLA in terms of both compression rate and decoded image quality. Furthermore, 1.0bpp and more than 30dB in PSNR by a CB with only 252 code-vectors were achieved using this method.

Noninvasive heart rate monitoring system for avian embryos based on the ballistocardiogram

A noninvasive heart rate (HR) monitoring system for avian embryos has been developed based on the ballistocardiogram (BCG). The BCG was detected using a phonograph cartridge as a record of the velocity of the minute ballistic movements of the eggshell, which are generated by recoil and impact of heart contraction and blood ejection. The autocorrelation coefficients (ACCs) of the detected signal were computed to confirm whether the detected signal contained ballistic movement. An envelope of ACC was calculated by the monitoring system to measure the intervals between peaks and/or highly correlated parts in the ACC, and then the system obtained HR by measuring these intervals. To demonstrate the valid range of the detecting method, the BCGs of chickens and Japanese quail embryos of different ages were measured. The result of the experiment shows that the BCGs of chickens and quail embryos are detected fully after about two-thirds of the incubation period has passed. The monitoring system will enable us to perform long-term HR measurement for developing avian embryos up to hatching.

Integration of local fractal dimension and boundary edge in segmenting natural images

We present a method that integrates local fractal dimension and edge information into a region growing algorithm for the segmentation of natural images. We compare two methods of estimating the local fractal dimension in the proposed segmentation algorithm. One is a blanket method and the other is a Fourier-wavelet method. We also propose a technique to store the edge information not on a pixel itself but on a boundary between pixels in the region-edge integrating algorithm in order to use the edge information more effectively and to simplify the algorithm.

Rough and accurate segmentation of natural images using fuzzy region-growing algorithm

We present a rough and an accurate segmentation of natural images using a fuzzy region-growing algorithm. First, an optimum number of the blanket for local areas is determined to estimate the optimal local fractal dimension. Then, the intensity features and the local fractal-dimension feature are integrated into the fuzzy region-growing algorithm. In the proposed method, the intensity features are used to produce an accurate segmentation, while the fractal-dimension feature is used to yield a rough segmentation in a natural image. The effectiveness of the proposed method is confirmed through computer simulations that demonstrate a rough segmentation at the fine-texture regions and an accurate segmentation at the strong-edge regions simultaneously.

Fuzzy region-growing segmentation of natural images using local fractal dimension

We present a new method that integrates intensity features and a local fractal-dimension feature into a region growing algorithm for the segmentation of natural images. A fuzzy rule is used to integrate different type of feature into a segmentation algorithm. In the proposed algorithm, intensity features are used to produce an accurate segmentation, while the fractal-dimension feature is used to yield a rough segmentation in a natural image. The effective combination of the different features provides the segmented results similar to the ones by a human visual system.