Kazuhiko Hamamoto

Discrimination of Breast Tumors in Ultrasonic Images Using an Ensemble Classifier Based on the AdaBoost Algorithm With Feature Selection

This paper proposes a novel algorithm to estimate a log-compressed K distribution parameter and presents an algorithm to discriminate breast tumors in ultrasonic images. We computed a total of 208 features for discrimination, including those based on a parameter of a log-compressed K-distribution, which quantifies the homogeneity of the echo pattern in the tumor, but is influenced by compression parameters in the ultrasonic device. The proposed algorithm estimates the parameter of the log-compressed K-distribution in a manner free from this influence. To quantify irregularities in tumor shape, pattern-spectrum-based features were newly developed in this paper. The discrimination process uses an ensemble classifier trained by a multiclass AdaBoost learning algorithm (AdaBoost.M2), combined with a sequential feature-selection process. A 10-fold cross-validation test validated the performance, and the results were compared with those of a Mahalanobis distance-based classifier and a multiclass support vector machine. A total of 200 carcinomas, 50 fibroadenomas, and 50 cysts were used in the experiments. This paper demonstrates that the combination of a classifier trained by AdaBoost.M2 and features based on the estimated parameter of a log-compressed K-distribution, as well as those of the pattern spectrum, are useful for the discrimination of tumors.

Unidirectional antenna using two-probe excited circular ring above square reflector for polarization diversity with high isolation

This paper presents a circular ring antenna fed by two perpendicular probes, both of which are placed above the square re∞ector. The antenna is employed to radiate unidirectional beam for polarization diversity reception. A linear isolator is added to improve the isolation between the two probes. The antenna is proposed for the point-to-point communication of Wireless Local Area Network (WLAN) system according to the IEEE 802.11a standard in which the allocated frequency band ranges from 5.150GHz to 5.825GHz. The proposed antenna is compact and suitable for mass production. Without the dielectric material, the antenna is free of dielectric loss and capable of high power handling. The prototype antenna was fabricated and measured to verify the theoretical predictions. At the center frequency, the unidirectional pattern with the measured half-power beamwidths in two principal planes of 65 and 75 degrees is achieved. The front-to-back ratio is 31dB, and the antenna gain is 7.42dBi. The jS11j and jS21j are respectively i23:09dB and i33:99dB; the obtained bandwidth is 23.64%. Based on the aforementioned characteristics, the antenna is a potential candidate for polarization diversity of WLAN applications.

Fingerprint Image Enhancement with Second Derivative Gaussian Filter and Directional Wavelet Transform

In this paper, we propose a technique for enhancing the quality of fingerprint images. Directional wavelet transform and second derivative of a Gaussian filter are applied. The original fingerprint image is decomposed into approximation and detail sub-images. To each sub-dimension a directional filter: second derivative of Gaussian filter is applied for tuning up the image features. The enhanced image is measured for its improvement by testing the success of core point identification where Poincare technique is used. The commonly-well-known database FVC-2004 is used in this study. The obtained results offer clean visualization as well as the increase the success of true core point detection.

A cost-sensitive extension of AdaBoost with markov random field priors for automated segmentation of breast tumors in ultrasonic images.

PurposeA cost-sensitive extension of AdaBoost based on Markov random field (MRF) priors was developed to train an ensemble segmentation process which can avoid irregular shape, isolated points and holes, leading to lower error rate. The method was applied to breast tumor segmentation in ultrasonic images.MethodsA cost function was introduced into the AdaBoost algorithm that penalizes dissimilar adjacent labels in MRF regularization. The extended AdaBoost algorithm generates a series of weak segmentation processes by sequentially selecting a process whose error rate weighted by the cost is minimum. The method was tested by generation of an ensemble segmentation process for breast tumors in ultrasonic images. This was followed by a active contour to refine the extracted tumor boundary.ResultsThe segmentation performance was evaluated by tenfold cross validation test, where 300 carcinomas, 50 fibroadenomas, and 50 cysts were used. The experimental results revealed that the error rate of the proposed ensemble segmentation was two-thirds the error rate of the segmentation trained by AdaBoost without MRF. By combining the ensemble segmentation with a geodesic active contour, the average Jaccard index between the extracted tumors and the manually segmented true regions was 93.41%, significantly higher than the conventional segmentation process.ConclusionA cost-sensitive extension of AdaBoost based on MRF priors provides an efficient and accurate means for the segmentation of tumors in breast ultrasound images.

Wavelet filters comparison for ultrasonic image compression

At present, wavelet transform has become a standard tool in medical image compression. Although, the choice of wavelet filters is an essential issue that can be determined the compression image quality. The goal of this paper is to provide the comparison of applying different types of wavelet filters belonging to orthogonal and biorthogonal wavelets with different orders on the ultrasonic images. The wavelet filters used are Haar, Daubichies8, 6/10, 9/7 and 5/3 filters. The performance evaluation of the image quality is measured objectively using maximum difference (MD), normalized absolute error (NAE), and normalized mean square error (NMSE). Moreover, ultrasonographers do the subjective performance assessment. From the experimental evaluation, it can be concluded that objective assessment alone is not suitable objective scale to evaluate the quality of the compressed images. Therefore, subjective assessment is very important to take into account visual quality.

Study on medical ultrasonic echo image compression by JPEG2000 : optimization and the subjective assessment of the quality

Lossy pulse echo ultrasonic image compression by a JPEG baseline system is permitted in DICOM and many papers about it have been presented. On the other hand, application of JPEG2000 to DICOM has been investigated. JPEG2000 is a new standard of image compression whose basis is wavelet transform. Although JPEG2000 is superior to JPEG in general, the viewpoint may not be applied to medical ultrasonic echo image since features of the image are unique and different from one of natural image significantly. In this paper, it is attempted to optimize the compression condition of JPEG2000 for medical ultrasonic echo image. There are two quantization methods in JPEG2000. One is the scalar derived quantization and the other is the scalar expounded quantization. The latter can be optimized by user. In this paper, firstly, it is shown that the scalar derived quantization is not appropriate for medical ultrasonic echo image even though the PSNR is high. Secondly, the scalar expounded quantization is optimized for the image. Lastly, the images by conventional JPEG2000 and ones by optimized JPEG2000 are evaluated by ultrasonographers. The results of the subjective assessment show the images by optimized JPEG2000 preserve more important diagnostic information than conventional JPEG2000.

A robust road profile estimation method for low texture stereo images

The estimation of road profiles from low texture stereo images is a problematic task because the disparity images computed from such class of images have a large number of noisy disparities. This paper presents a new method that is based on edge maps to guide the cost aggregation process in the stereo matching problem. Using the proposed aggregation method, the disparity images are smooth at low texture regions, but the boundaries of on-road objects are still preserved. The V-disparity images computed from such reliable disparity images can clearly show the road profiles. Thereby, the road profiles can be straightforwardly extracted by the dynamic programming technique. Experiments on a long and real stereo image sequence demonstrate that the proposed method can robustly estimate the road profiles. Furthermore, on-road objects can be detected by combining v- and u-disparity images as well because their boundaries are preserved in the disparity images.

Real time eye tracking using initial centroid and gradient analysis technique

Robust and accurate algorithm in real time eye tracking system has been a fundamental and challenging problem for computer vision and biomedical engineering area. This paper proposes a new method to estimate eye-motion position and direction based on initial centroid and gradient analysis technique. The proposed method was validated by tracking eye position within high and low occlusion conditions. Experimental results show that the proposed method improves the accuracy of tracking pupil position compared to conventional window-matching method. The proposed method yields 76.07% and 87,89% accuracies while the conventional window-matching performs 44.56% and 67.89% accuracies within high and low occlusion conditions, respectively.

Ultrasonic refractive index and sound velocity tomography

Ultrasonic computed tomography (UCT) is one of the methods capable of tissue characterization. It is expected to provide not only a quantitative image but also diagnostic information. As in X-ray CT, UCT requires a projection data to reconstruct a cross-sectional image. The projection data of ultrasonic tomography is based on measurement the time delay, which is time difference between ultrasound traverse with and without object. In this paper, we investigate two different types of quantitative UCT image, refractive-index and sound-velocity image. We purpose the new method of measurement the time delay by converting the received ultrasonic pulse to frequency domain and measuring the phase shift of the center frequency of the broadband pulse. The method seems more robust to noise. Two image reconstruction techniques are used for comparison purpose including traditional filtered back-projection and simultaneous algebraic reconstruction technique (SART).

A Robust Stereo Matching Method for Low Texture Stereo Images

Computing disparity images for stereo pairs of low texture images is a challenging task because matching costs inside low texture areas of the stereo pairs are almost similar. This problem can not be solved straightforwardly by increasing the size of aggregation windows or by using global optimization methods, e.g. dynamic programming, because those approaches will smooth depth discontinued boundaries as well. Based on the assumption that disparities of pixels in homogeneous regions are similar, this paper proposes a new method that is able to robustly perform stereo matching for low texture stereo images. The proposed method utilizes the edge maps computed from the stereo pairs to guide the cost aggregation process in stereo matching. By using edge maps, the proposed method can achieve the effect of using different shapes and sizes of aggregation windows. Moreover, the computational complexity of the proposed method is independent from the window size, similar to the moving average aggregation method. Experimental results from both of an artificial and a real stereo image sequence demonstrate that the proposed method can produce a larger number of and a better accuracy of reliable disparities for low texture stereo images than the moving average method.