羽石 秀昭

Acquisition and analysis of three-dimensional movement of knee joint is desired in orthopedic surgery. We have developed two methods to obtain dynamic volume images of knee joint. One is a 2D/3D registration method combining a bi-plane dynamic X-ray fluoroscopy and a static three-dimensional CT, the other is a method using so-called 4D-CT that uses a cone-beam and a wide 2D detector. In this paper, we present two analyses of knee joint movement obtained by these methods: (1) transition of the nearest points between femur and tibia (2) principal component analysis (PCA) of six parameters representing the three dimensional movement of knee. As a preprocessing for the analysis, at first the femur and tibia regions are extracted from volume data at each time frame and then the registration of the tibia between different frames by an affine transformation consisting of rotation and translation are performed. The same transformation is applied femur as well. Using those image data, the movement of femur relative to tibia can be analyzed. Six movement parameters of femur consisting of three translation parameters and three rotation parameters are obtained from those images. In the analysis (1), axis of each bone is first found and then the flexion angle of the knee joint is calculated. For each flexion angle, the minimum distance between femur and tibia and the location giving the minimum distance are found in both lateral condyle and medial condyle. As a result, it was observed that the movement of lateral condyle is larger than medial condyle. In the analysis (2), it was found that the movement of the knee can be represented by the first three principal components with precision of 99.58% and those three components seem to strongly relate to three major movements of femur in the knee bend known in orthopedic surgery.

An image editing system that deal with spectral reflectance instead of RGB values in retouching or changing the color of an object is proposed, for the purpose to achieve high accuracy and enhanced functionality in color reproduction.

実世界にある素材を正しい色，質感でCG再現するためのBTF/BRDF取得方法，レンダリング方法，表示方法を提案する．例えばある生地を用いて作成した衣装の質感，色を実際に試作することなくCGとして正確に再現しようとするものである．多バンド且つマルチアングルで素材の反射特性を計測することにより，所定の照明下での色を正確に再現するとともに，観測角度による色の見えの違いを正確に再現することができるようになる．（担当：実験装置試作支援、研究連携支援）<br />

実世界にある素材を正しい色，質感でCG再現するためのBTF/BRDF取得方法，レンダリング方法，表示方法を確立すべく開発した、多バンド，マルチアングルの画像撮影装置およびその基本原理を説明。（担当：装置試作支援、研究連携支援）<br />

We have developed a method that combines a small semiconductor gamma camera with an optical camera to synthesize the two kinds of images and help surgeons to easily identify sentinel lymph nodes on the gastric surface in the cancer resection surgery. The proposed method includes some key techniques such as distortion correction of the optical camera image, distance estimation between the camera head and the object surface using a laser and the optical camera, and perspective transformation of the gamma camera image to fuse with the optical camera image. The method and preliminary experimental results with a prototype setup are presented.

We are developing a high-performance brain PET scanner, jPET-D4. This scanner is designed to achieve not only high spatial resolution but also high sensitivity using four-layered depth-of-interaction (DOI) detectors. The scanner has five detector rings with the ring diameter of 390 rum and each detector ring consists of 24 DOI detectors. In order to identify the crystal of interaction and its energy from PMT output, we have previously proposed the DOI detector calibration method. This method creates position and energy Look-Up-Tables (LUTs) from the uniform irradiation. The position LUT converts location of 2D position histograms to crystal addresses. The energy LUT corrects light output of each crystal. Five detector rings have been installed and calibrated. The average energy resolution for 120 DOI detectors is optimized to 17 % 1.4 %. The scatter fraction for this system is 38 % with energy window of 400-600 keV. The sensitivity for point source is 112 kcps/MBq with a 250 keV LLD.

The jPET-D4 is a high-performance brain PET scanner which achieves not only high spatial resolution but also high scanner sensitivity by discriminating 4-layer depth-of-interaction (DOI) information. The scanner is designed to have 5 rings of 24 detector blocks each, and the detector block consists of 1,024 GSO crystals of 2.9 mm x 2.9 mm x 7.5 mm, which are arranged in 4 layers of 16 x 16 arrays. In this paper, we investigated the imaging performance of the jPET-D4 prototype where one of the 5 block-detector rings was assembled. In order to reduce computational cost while retaining the advantage of DOI information in iterative image reconstruction, we have proposed the DOI compression (DOIC) method which reduces data dimensions with suppressing resolution loss. We have also proposed an approximated system model for the jPET-D4 which enables fast system matrix calculation while preserving image quality. The experimental results show that almost uniform spatial resolution of 2-3 mm is obtained over the field-of-view by using the 4-layer DOI information. A Hoffman phantom image clearly shows the excellent imaging performance of the jPET-D4.

A respiratory-gated SPELT allows image reconstruction at various phase of the respiration and therefore the blur in an image can be avoided. However, it requires long time to accumulate sufficient number of counts at each phase. If the acquisition time is not long enough, the detected count becomes inadequately small and hence the reconstructed image becomes noisy. We have been proposed a method for correcting the motion between different phase images obtained with the respiratory-gated SPELT. By summing motion-corrected phase images, a less noisy and less blurred SPELT image can be obtained. In this paper, the other advantage of motion correction information produced in the middle of motion correction and summation process is described in addition to the review of the proposed method. We found that there are clear differences in respiratory motion between normal lung and lung with disorder, especially, emphysema. Detail analysis of the motion of the lung may serve as an aid for diagnosis. In this paper, the clinical usefulness is discussed through investigation including quantitative evaluation of the respiratory motion, comparison with ventilation planar images and physician's observation on them.

We present a proposal to achieve accurate color reproduction by fitting the CIE-XYZ tristimulus values of a given color stimulus, and using the 16-band multispectral images, as well as the CIE-XYZ tristimulus values of a known color test chart. We propose a simple linear combination of the multispectral images as interpolation function, which is equivalent to fitting the data to a straight line in a 16-dimentional space. By using this interpolation function, we minimize the merit function of the mean error between the measured and estimated CIE-XYZ tristimulus values of the color test chart. We show, by making a visual comparison between the results achieved using this proposal and the results achieved using a spectral reflectance estimation technique, that the proposed interpolation function when using 16-channel multispectral images produces high quality results in terms of color reproduction fidelity.

The development of Natural Vision (NV) system as an advanced color imaging system was discussed. It was found that NV can reproduce images with exactly the same color and also simulate objects under different illumination. NV is also found to be an advanced color imaging system consisting of image capture, processing, transmission, display, recognition and deals with primary colors depending upon application requirements. It was found that the three primary colors of a conventional display form a color triangle and colors within the triangle could be reproduced accurately by using NV technology.

The Next Generation PET scanner named jPET-D4 has been developed at the National Institute of Radiological Sciences (NIRS) in Japan. The jPET-D4 has 4 layers DOI detectors [1] and its data size becomes huge to handling. Then at first stage, we plan to apply the reconstruction techniques after the histogramming of list-mode data because it is easy to implement and effective to decrease computational cost. The conventional histogramming technique is inapplicable to 4 layers DOI detectors, so we proposed the new histogramming method which is suitable for the multi-layer DOI detector. In this method, DOI-PET list-mode data is transformed into sinogram with an accurate system model of the crystal block. In this paper, we implemented this method with a full three dimensional system model. And we also applied the method to experimental data with prototype system for the jPET-D4 scanner. Experimental results show that the proposed method keeping the advantage of DOI information, and the image quality of the jPET-D4 is superior to a conventional NonDOI PET scanner.

Radiation therapy for lung cancer using heavy ion particles is going on at the National Institute of Radiological Sciences of Japan. In such a field, image registration between different modalities or different images is very effective for overall and more precious diagnosis and treatment and its realization is strongly desired. In the radiation therapy, three kinds of medical images, namely, CT, SPECT and dose map are used. There are two kinds of CT image, one is CT for treatment planning and the other is CT for diagnosis. For image registration, we use CT images to relate all images. At first, two kinds of CT images are registered by a non-linear transformation method newly developed. Next, CT image for diagnosis and SPECT images are registered by a linear transformation or a rigid body transformation. The dose map has already a good agreement with the CT for treatment planning. Then, the dose map is nonlinearly transformed using the transformation parameters obtained through the CT-CT matching and a SPECT image is linearly transformed using parameters obtained through the CT-SPECT matching. Finally those images are superposed on the CT for diagnosis and evaluation of lung function is performed. The framework of image registration is presented with detail description on non-linear transformation of CT images and the application to the evaluation of lung function is shown.

A project for developing a next generation PET scanner named 'jPET-D4' is going on at the National Institute of Radiological Sciences of Japan. In this project, a new four-layer depth of interaction (DOI) detector is proposed to achieve high resolution and high sensitivity. The DOI detector consists of Gd2SiO5: Ce (GSO) crystal array with proper optical reflectors and position sensitive photomultiplier tubes (PSPMT). Previously we have developed a Monte Carlo simulation for the DOI detector of 2x2x4 crystal elements. In this study, in order to simulate the latest version of DOI detector with more crystal elements, we developed a much faster, simplified simulator. The content of simplified simulator and some simulation results for multiple scattering in detector are presented.

The Next Generation PET scanner named jPET-D4 is being carried out at the National Institute of Radiological Sciences (NIRS) in Japan. At this stage, we plan to apply the reconstruction techniques after the histogramming of list-mode data because it is easy to implement and has low computational cost. The jPET-D4 has 4 layers DOI detectors and it can obtain true list-mode data. However, the conventional histogramming technique is inapplicable to 4 layers DOI detectors, so in this paper we propose a new histogramming method which is suitable for the multi-layer DOI detector. In this method, DOI-PET list-mode data is transformed into sinogram with an accurate system model of the crystal block. We also applied the method to simulated data for the jPET-D4 scanner. Numerical simulation results show that the proposed method reduces sampling error effectively while keeping the advantage of DOI information, and the image quality of jPET-D4 is superior to a conventional NonDOI PET scanner.

A high-performance brain PET scanner, jPET-D4, which provides depth-of-interaction (DOI) information, is under development. The scanner is designed to achieve not only high spatial resolution but also high scanner sensitivity with the DOI information obtained from multi-layered thin crystals. The scanner has 5 rings of 24 detector blocks each, and the detector block consists of 1,024 GSO crystals of 2.9mm x 2.9mm x 7.5mm, which are arranged in 4 layers of 16 x 16 arrays. At this stage, a pair of detector blocks and a coincidence circuit has been assembled into an experimental prototype gantry. In this paper, as a preliminary experiment, we have investigated the performance of the jPET-D4's inherent spatial resolution and noise characteristics using the prototype system. Spatial resolution and background noise were measured by using a line source and a uniform cylinder phantom respectively. To avoid systematic error and reduce computational cost in image reconstruction, the DOI compression (DOIC) method followed by the ML-EM with accurate system modeling, that we have previously proposed, was applied. Experimental results showed that the jPET-D4 achieves less than 3mm spatial resolution over the field-of-view. Trade-off results between spatial resolution and background noise also promise an excellent performance of the scanner.

A detector proposed by Murayama et al. for detection of depth-of-interaction (DOI) in PET consists of three-dimensionally arranged crystal elements with proper optical reflectors and is coupled to an array of photomultiplier tubes. This detector has a great advantage in easiness and cost in fabrication. We implemented a simulator of this detector that allows us to find appropriate values of parameters such as optical properties of crystal or detector unit geometry before making detectors. The simulator is based on the Monte Carlo method that traces the migration of optical photons generated by interaction of a gamma ray with crystal. First, the simulator performance was validated by comparing with the experimental data obtained with some prototype detectors. Then, on some parameters including refractive index of inter-crystal material, reflectance of optical reflector and detector geometry, appropriate values were investigated for accurate discrimination of crystal element of interaction.

The Next Generation PET scanner named JPET-D4 is being carried out at the National Institute of Radiological Sciences (NIRS) in Japan. At this stage, we plan to apply the reconstruction techniques after the histogramming of list-mode data because it is easy to implement and has low computational cost The JPET-D4 has 4 layers DOI detectors and it can obtain true list-mode data However, the conventional histogramming technique is inapplicable to 4 layers DOI detectors, so in this paper we propose a new histogramming method which is suitable for the multi-layer DOI detector. In this method, DOI-PET list-mode data is transformed into sinogram with an accurate system model of the crystal block. We also applied the method to simulated data for the JPET-D4 scanner. Numerical simulation results show that the proposed method reduces sampling error effectively while keeping the advantage of DOI information, and the image quality of JPET-D4 is superior to a conventional NonDOI PET scanner.

Radiation therapy for lung cancer using heavy ion particles is going on at the National Institute of Radiological Sciences of Japan. In such a field, image registration between different modalities or different images is very effective for overall and more precious diagnosis and treatment and its realization is strongly desired. In the radiation therapy, three kinds of medical images, namely, CT, SPECT and dose map are used. There are two kinds of CT image, one is CT for treatment planning and the other is CT for diagnosis. For image registration, we use CT images to relate all images. At first, two kinds of CT images are registered by a non-linear transformation method newly developed. Next, CT image for diagnosis and SPECT images are registered by a linear transformation or a rigid body transformation. The dose map has already a good agreement with the CT for treatment planning. Then, the dose map is nonlinearly transformed using the transformation parameters obtained through the CT-CT matching and a SPECT image is linearly transformed using parameters obtained through the CT-SPECT matching. Finally those images are superposed on the CT for diagnosis and evaluation of lung function is performed. The framework of image registration is presented with detail description on non-linear transformation of CT images and the application to the evaluation of lung function is shown.

Natural Vision Image Data File Format has been developed in order to realize spectral-based color reproduction. This paper introduces the profile format of Natural Vision Image Data File Format. It is based on ICC Profile Format. New Profiles, Tags and Tag Types, therefore, have been developed to cover the lacks of ICC Profile Format for spectral-based color reproduction. The design concepts of Natural Vision System are introduced with respect to the requirements. Input Profiles and Display Profiles are expanded in order to provide spectral characteristics to color reproduction system. ColorSpace Conversion Profiles is expanded in order to provide rendering illuminant and CMFs to Color Reproduction Intent Management System. Relationships between the expanded Profiles and image processing are also descried.

現状のフラットパネルディスプレイ(FPD)は,自然界の全ての反射色を表示するために十分に広い色再現範囲を有していない.本報告では、ディスプレイの色再現範囲を拡大するために,従来の3原色表示に代わり,4原色表示のFPDを提案する.4原色表示のFPDを実現するために,4原色のスクロールLEDバックライトと2フレームのフィールドシーケンシャル(FS)駆動の2つの技術を開発した.その結果,対角15インチのXGA IPS TFT-LCDにおいて,色再現範囲122%を実現した.

"Natural Vision" is currently being studied for the purpose to realize natural color reproduction and highly realistic image display in visual communication systems. By introducing spectrum-based color reproduction scheme, it becomes possible to capture, transmit, store, and display high-fidelity color information, which has been difficult in trichromatic color imaging systems. This article introduces the overview of natural vision, as well as the technologies of spectrum-based color reproduction system, multispectral image capture, and multiprimary display.

Backward model for multi-primary display, which gives digital count in order to present a given color stimulus value XYZ on a screen, is proposed in this paper. Performance of the proposed method is evaluated with 6-primary display. Error in predicting XYZ of colors projected on a screen was 0.8 on average and 2.9 at maximum.

TAO* has developed six-primary HDTV display system. Two DLP™ projectors that have different three primaries superimpose two images on a screen. These six primaries almost covers (Pointer + SOCS) gamut, which is a typical example of natural reflective objects around us. Volume of six-primary display gamut is larger than sRGB display gamut at 1.6 times.

At the IEEE MIC 1999, we presented a Monte Carlo simulator made for designing a depth encoding multicrystal detector and reported some simulation results. While the basic performance of the simulator was validated from the comparison between the simulator outputs and experimental results using a prototype detector, many factors to be implemented were still remaining for simulating real phenomena inside the detector. This time, we improved the simulator by including Compton scattering, more accurate migration of optical photons at every interface of materials, and so on. Using this simulator, discrimination characteristics were investigated with respect to thickness of scintillator block, refractive index of inter-stage material, reflectance of optical reflector and surface roughness of scintillator. From these investigations, useful information for detector design was obtained.

Accurately estimating the spectral reflectance of art paintings from low-dimensional multichannel images requires that both image-acquisition hardware with appropriate spectral characteristics and appropriate estimation software be applied to the captured multichannel image. In this study, a system that incorporates both factors is designed and developed on the basis of the minimum-mean-square error criterion. The accuracy of spectral estimation by use of this system is evaluated, and the system's high performance is demonstrated. (C) 2000 Optical Society of America OCIS codes: 100.2000, 100.3010, 120.2440, 220.4830, 330.1690.

The vector error diffusion method was applied to spectral color reproduction of hardcopy. Reflection spectra of the object and primary colors used in the printer were estimated by low dimensional linear model. The accuracy of spectral color reproduction was evaluated and analyzed by computer simulation. It was shown that the spectral color reproduction can be improved by the proposed method.

Maximum likelihood criterion accounting for statistical behavior of photon is very useful for image reconstruction from projection at low count rate. EM (expectation maximization) algorithm is a powerful tool which provides such a maximum likelihood estimator successfully. In this article, the fundamental algorithms for emission and transmission image reconstruction originally proposed by Lange and Carson are reviewed. Here the likelihood incorporating the Poisson nature of photon counting is formulated and the procedure of the EM algorithm for maximizing the likelihood is shown.

A detector proposed by Murayama et al. for detection of depth-of-interaction (DOI) in PET consists of three-dimensionally arranged crystal blocks with proper optical reflectors and is coupled to an array of photomultiplier tubes. This detector has a great advantage in easiness and cost in fabrication. We built a simulator of such a detector which allows us to find appropriate values of parameters such as optical properties of crystal or detector unit geometry before making detectors actually. The simulator is based on the Monte Carlo method which traces the migration of optical photons generated by interaction of a gamma ray with crystal. First, the simulator performance was validated by comparing with the experimental data obtained with some prototype detectors. Then, on some parameters including refractive index of crystal, detector geometry, and light yield per interaction, appropriate values were investigated for accurate block discrimination from PMT output signals.

It is well known that an image has a particular gazing area. From the experimental results of our previous papers, we found that total image quality is highly influenced by quality of the gazing area. In this paper, we evaluated total image quality of degraded images by a physical quality criterion (PQC) calculated from the gazing area. The experimental result showed that the calculated PQC is not dependent on the scene. On the other hand, PQC calculated from the whole area of the image is dependent on the scene content. In this paper, the effectiveness of the evaluation of image quality under the consideration of gazing area is presented and discussed.

We have developed the multiband color imaging systems in the frame work of the IPA project as New Color Management System Based on Human Perception. This system consisted of multiband CCD camera, computer, CRT and projection type monitor. Five band images were taken by CCD camera with 2048x2048 pixels. Reflectance spectra of the object were estimated by the principal component analysis and Wiener estimation method. Spectral transmittance of separation filter was optimized by simulated annealing method based on Wiener estimation. Color adaptation model was introduced for accurate color reproduction of paintings based on human perception. In this paper, estimation of spectral reflectance of paintings, developed multiband camera, various software for image composition and control of multiband camera are introduced and demonstrated.

We are developing a spectral characteristics database for evaluating color reproduction in image input devices. The database is designed so that spectral characteristics are systematically divided by category, according to the purpose to which they are to be applied and every category has a sufficient variety of samples. The categories are as follows: photographic materials, graphic color printing, computer color printers output, paints, flowers, leaves, human skin colors and historical Krinov data. The total number of colors is 49, 672. The database is being proposed to be published as Japanese / International standard technical reports for use in establishing a new color reproduction evaluation method.

The design of spectral characteristic of the light source in an optical endoscope is discussed. First, two criteria are introduced for the evaluation of the Light source and the evaluation result for two CIE standard illuminants and 15 light sources commercially available is presented. Next, an optimization method using POCS (projection onto convex set) is proposed for the design of the light source. It is shown that the optimum right source achieves much better score for the introduced criteria than the 17 light sources. Simulation of image reproduction of mucous membrane under an arbitrary light source is also described.