Abstracts of Medical Imaging Technology of 2001.

January(Vol.19, No.1)

[-> return]

title	author	page
Reconstruction Algorithm for Multi Slice CT	Yun SHEN	3-10
Key words : Multi Slice CT, Reconstruction, Interpolation, Cardiac CT, Helical Pitch
Abstract : Multi-slice CT scanning is one of the most important medical equipment. Multi-slice CT scanning offers many new helical applications. The multi detector row and reconstruction algorithm are the most important hardware and software techniques for Multi-slice CT, respectively. The purpose of this paper is to display several helical reconstruction algorithms and cardiac reconstruction algorithms for multi slice CT. This paper helps you to understand the concepts about multi helical pitch, optimal z-sampling, optimal time-sampling, linear interpolation, longitudinal filtering algorithm, cardiac segmentation algorithm, and cardiac multi sector algorithm.

[-> return]

title	author	page
The Engineering Aspects and the Future of Multi-Slice CT Scanner	Toshihiro RIFU	11-16
Key words : Computed tomography, CT, Multi-Slice, Multi detector row, Helical Scan
Abstract : The biggest differences between multislice CT systems (MSCT) and single-slice CT systems are the detector system and the reconstruction algorithm. Four manufacturers has introduced MSCT with the detector of 4 rows, commonly. It is very interesting to see the difference of the detector configulation, which is unique each other. The reconstruction for MSCT has been develop to solve the multi-slice helical artifact. And also it has contributed to improve the image noise a lot. The isotropic imaging is the feature that MSCT made it realistic to use. The image noise improvement by the MSCT reconstruction has helped to use the isotropic condition by the thin slice scanning. MSCT will move to have the detector system with more number of rows. It's going to become very important to have to maintain the basic image quality for the thin slice image, to have the real time reconstruction technology to keep the reproducibility of the contrast enhancement, and to improve the tempral resolution to apply to the cardiac imaging.

[-> return]

title	author	page
Basics and Application of Multi-Detector Row CT	Katsutoshi MURATA	17-20
Key words : Computed Tomography, CT, Multi-Detector Row CT, MDCT, Cardiac CT
Abstract : Multi-Detector Row CT (MDCT) provided by Siemens is based on new hardware. Moreover, operator can set more flexible parameters because new concept for MDCT was introduced, which did not exist on conventional helical CT. Siemens also developed new application for heart disease. This new technique provides high temporal resolution, 125msec/image.

[-> return]

title	author	page
Clinical Impact of Multislice CT --Significance of Isotropic Volumetric Data and Real-time Reconstruction--	Kazuhiro KATADA	21-27
Key words : Multislice CT, Volume visualization, Isotropic voxel, CT fluoroscopy
Abstract : Most significant advantage of Multi-detector Row CT is the wide scanning range provided by the employment of the multislice detector. This primary characteristic of MDCT brings other two major advantages of MDCT-shorter scanning speed and finer z-axis resolution. Most significant advancement which is realized by MDCT is the ability to acquire "true" isotropic voxel by means of 0.5-mm collimation. This helpes to eliminate partial volume effect and realized the identical spatial resolution in axial, coronal and sagittal plane. Further enhancement of the applications of real-time CT was obtained with the use of multislice detector and 0.5-second rotation. Three plane simultaneous CT fluoroscopy enables us to eliminate couch-top sliding for the search of needle tip.

[-> return]

title	author	page
Clinical Applications of Multislice CT	Go KAWAKAMI	28-32
Key words : Multislice CT, CT angiography, Dynamic CT, Isotropic data
Abstract : Multislice CT (MSCT) is the new modality of faster gantry rotation speed and multiple detector row system, which enables to obtain higher resolution images with faster scan time and wider scan range. The near isotropic data enable to create high resolution MPR and MIP with an easy to use post-processing workstation. But on the other hand, the routine work of radiologists and radiographers increased because of the post-processing work of huge amount of row data.

[-> return]

title	author	page
Classification of Sarcomas Based on Textural Features	Toshiyuki TANAKA, Yoko MURASE, Tomoo JOKE	33-41
Key words : Textural feature, Sarcoma, Classification system, Pattern matching
Abstract : There are many types of sarcomas. It is therefore very difficult for physicians to classify sarcomas precisely, except for physicians who specialize in sarcomas. The precision of classification depends on the knowledge of the particular in physician. A system for classifying sarcomas is therefore necessary. This paper presents a classification system for sarcomas based on textural features. We assume that sarcomas have a single texture and that the same scale is used when acquiring images of sarcomas. Eighty sarcomas were selected under these assumptions. First, we constructed a database of templates of sarcomas based on textural features. The intensity histogram, Fourier power spectrum, run-length matrix, fractal dimension, and co-occurrence matrix were used as textural features. Second, the sarcomas were classified by template matching, based on the degree of similarity between each sample and the templates in the database. The success rate in this study reached 92%. The results of this study demonstrate the usefulness of the proposed system.

[-> return]

title	author	page
Segment Classification Method for X-Ray CT Images of Liver Cancer Based on Application of Inductive Learning to Segment Features	Masaki ISHIGURO, Ichiro MURASE, Noriyuki MORIYAMA	43-49
Key words : Pattern Recognition, Feature Extraction, Image Segmentation
Abstract : We propose a classification method for tumor segments in CT images of the liver based on decision rules derived by inductive learning techniques. Features of each segment or a group of segments are defined based on the morphological features that are typically used to discriminate cancers. The rules for discriminating between malignant and benign tumors are generated by applying decision tree analysis, and segments are scored based on these rules. Discrimination analysis is then applied to the scores and original feature values to determine the weight of each value for identifying cancer segments. Experimental results of ROC analysis have shown that classification methods incorporating rules derived by inductive learning outperform those without learned rules in terms of generalizability.

March(Vol.19, No.2)

[-> return]

title	author	page
Activities for Generalizing DICOM Standards in Japan	Shun KABATA	75-80
Key words : MIPS, DICOM, JMCP
Abstract : Nowadays we could say that without DICOM Standards it is not possible to communicate for transmitting and receiving medical image data and their related information among PACS. The author was appointed as the Chairman of the DICOM Committee in JIRA from July 1997 to February 2000 and made efforts to generalizing the DICOM standards in Japan. Through that time period, he could make to change the Japanese National Standards from MIPS Standards into DICOM Standards, to register the translated DICOM Standards to the server at the Cancer Institute and at the Keio University for publication, and to demonstrate DICOM networks at the JMCP Exhibitions for constructing the consensus among hospital staff towards DICOM systems.

[-> return]

title	author	page
DICOM and Relation to Other Standards	Hidenori SHINODA	81-86
Key words : DICOM, IHE, EMR, Image integration, HL7
Abstract : DICOM is now recognized as a de facto standard worldwide. DICOM Standards Committee is now shifting their target to standardize messaging and communications between imaging systems and other information systems like RIS and/or HIS. Recently, the committee agreed with HL7, Inc. to develop an image integration model that will be needed to establish electronic health record or EHR, which will have a link to imaging exam results. In the age of information society, IT is expected to able to improve quality of healthcare. In order to make this possible, various standards should be utilized in harmony. Here, the relationship of DICOM with other standards will be described after IHE that is a model for the future healthcare system is introduced.

[-> return]

title	author	page
Shortcomings and Expected Problems of DICOM	Michio KIMURA	87-92
Key words : DICOM, Hospital information system, PACS
Abstract : Shortcomings and expected problems of DICOM are illustrated. Some troubles in implementation are avoidable by precise conformance statements. Image standard other than DICOM, e.g. JPEG, may be used in hospital wide service. Harmonization with DICOM protocol should be established. Also, medical information standard other than DICOM, e.g. HL7, should be used in the same time in the same hospital system. Preciseness of order entry in Japan is much higher than that in the States. Therefore, we should contribute in making detailed codes which are used in DICOM messages. There codes include examination modalities, examination region, etc. Also highly required in Japan is patient exposure record. When using DICOM supplement 17 for billing and record keeping, some elements should be "mandated in Japan", even on "optional" elements in original DICOM. DICOM already has a way to accomodate multiple byte character sets. ISO 2022 escape technique is used for this.

[-> return]

title	author	page
DICOM standard and Electronic Storage, Display and Print Function	Yutaka ANDO	93-99
Key words : Electronic storage of medical images, Grayscale display function, DICOM standard
Abstract : The Ministry of Health and Welfare (MHW) showed the criteria concerning the electronic storage of the medical record. The medical information system development center opened the common standard for medical image electronic storage to the public. In the electronic storage the images that we use clinically should be stored. The image format of the stored images is not restricted to the DICOM format whenever the image is displayed correctly. The irreversible image compression had a problem that the image that we use clinically may be deferent from the image that we read from the storage. If the electronic medical information becomes popular, it is necessary for us to make a diagnosis with a CRT without a film. Recently the liquid crystal display (LCD) device is improved for the image display, the LCD can display the same ability as a CRT. The part 14 Grayscale Standard Display Function in DICOM standard was standardized. In the case of electronic storage of medical images, it is very important to display images in the suitable reproduction. The criteria of the MHW do not contain the reproduction of display in correct grayscale.

[-> return]

title	author	page
DICOM Structured Reporting	Kiyonari INAMURA	101-107
Key words : DICOM, Structured Reporting, Radiological reports, Image diagnosis, Radiology information system
Abstract : This paper interpret, explains and describes how to make use of DICOM Structured Reporting SOP Classes. First, history and concept of SR, professionals who can use SR, simple example of SR, its features, method of referring DICOM images and associated other languages are described. Then, method of defining content of SR, concepts of "containing" and "templating" are explained. Lastly relevant standards such as HL7, CDA, XML, and HTML are raised, because SR must be cooperated with such languages and standards for the purpose of applying Structured Reporting. This paper also involves what is objective of DICOM SR, what should be contained in it, how diagnostic reports are orgnized in it, and who should employ it. Structure Reporting will be very effective when any of diagnostic reporting has any of following factors: hierachical relationship within reports, logical statements, lists of names of diseases or anatomical description to be coded by standard nomencleature such as SNOMED, the use of relationship between content in reports, and the presence of embedded reference to DICOM image and waveforms.

[-> return]

title	author	page
The Present Situation of Security in DICOM -- Include activity in Japan --	Satoru SHIMANISHI	108-113
Key words : DICOM, Security, Online Electric Storage, ISCL, TLS, Electric Signature, Encode technique, Media Encode, Value Encode
Abstract : The spread of a network progressed, and the opportunity that did Internet in home increased. A network occupies an important position in each industry. Medical care is not an exception, too. However, data through a network are disturbed, and it is fact that the stolen damage increases.

[-> return]

title	author	page
Reduction of Cardiac Motion Artifacts in Lung HRCT Images Acquired by Half-second CT -- Evaluation of the High-speed Half Scan Technique --	Shuji YAMAMOTO, Takeshi JOHKOH, Toshihiro RIFU, Takashi MATSUMOTO, Shohzoh NAKANISHI, Yoshifumi NARUMI, Hiroaki NAITO, Hironobu NAKAMURA	115-123
Key words : Half scan, Cardiac motion artifact, HRCT
Abstract : Image quality assessment and studies to evaluate the reduction of cardiac motion artifacts were conducted for lung high-resolution CT(HRCT) images acquired using the half scan technique, in which images are reconstructed from data acquired during only a part of the scanning period in half-second (0.5 s/rotation) scanning. Images were acquired using a combination of the helical scan and half scan techniques, and the physical characteristics of these images were assessed based on analysis of the slice thickness, noise, and density histogram. The reduction of artifacts generated by the motion of the margins of the heart was also evaluated by motion vector analysis. In these investigations, it was found that the X-ray quantum noise in half scan was higher than in 360¡ëaxial scan but showed a high degree of correlation(correlation coefficient 0.9943) with images obtained using the 180-degree linear interpolation technique. In addition, image distortion and cardiac motion artifacts, which cannot be reduced by the conventional half scan technique, were markedly reduced in HRCT images.

May(Vol.19, No.3)

[-> return]

title	author	page
Three-dimensional image processing algorithms -- General Survey --	Jun-ichiro TORIWAKI	135-141
Key words : Three dimensional image, Topology preserving transformation, Distance transformation, Visualization, Algorithm
Abstract : In this article, the author presents a survey of algorithms for three dimensional image processing. Extension of digital images to three dimensional ones and four dimensional moving pictures, characteristic features of 3D image processing and general trends in the study of 3D image processing algorithms are introduced first. Then extendibility of 2D algorithms to 3D images is explained for each categories of functions such as filtering and topology preserving transformation. A little detailed descriptions of recent studies are provided concerning density value manipulation, topology preserving transformation, distance transformation and related algorithms, image processing expert system and visualization method. Important literatures are listed in the end of the text.

[-> return]

title	author	page
Feature Extraction Algorithms Based on 3-D Curvatures	Yoshiki KAWATA Noboru NIKI	142-153
Key words : 3-D curvature, feature extraction, classification, pulmonary nodule
Abstract : Three-dimensional (3-D) curvatures such as Gaussian, mean and principal curvatures possess intrinsic surface properties and provide important features to analyze 3-D object surfaces. This paper presents 3-D curvature based feature extraction algorithms. These algorithms are applied to the internal structure analysis of pulmonary nodules and the¡¡volume information are shown to provide vital cues on benign and malignant classification.

[-> return]

title	author	page
Evaluation Method of Concentration Degree and Convergence Index Filter	Yukiyasu YOSHINAGA Hidefumi KOBATAKE	154-160
Key words : Vector convergence, Concentration index, Rounded convex region, curvilinear structure, Edge enhancement, Gradient vector orientation
Abstract : The evaluation of convergence degree of lines and/or vectors is very important for the diagnosis of medical images. The existence of radiating structure of blood vessels and/or spicules around a tumor is a reliable signal of malignancy. This article presents evaluation methods of convergence degree which is defined as the concentration degree of line segments and is known as a parameter effective in identifying malignant tumors. The vector convergence index filter has unique characteristics that it can enhance the target regardless of the contrast to its background. In this article, characteristics of two dimensional convergence index filters are introduced and, by showing the similarity between the 2D and 3D images, the possibility of the extension of the convergence index filters to 3D images is shown.

[-> return]

title	author	page
Algorithms for Extracting Line Figures from Three Dimensional Gray Images	Yasushi HIRANO Jun-ichiro TORIWAKI	161-167
Key words : Line figure, Center line extraction, Ridge tracking, Thinning, Three-dimensional image
Abstract : In this paper, we introduce the methods to extract line figures from three-dimensional (3D) gray images. It is very useful to use the characteristic line figures of objects, when we analyze the shape of objects or register the objects in 3D gray images. The volume rendered (VolR) images or the MIP (Maximum Intensity Projection) images are used in order for the human observer to understand the shape of objects or the structure of objects. In the simpler way, there are many cases that it is enough to understand the shape of objects using line figures. The algorithms to extract line figures can be divided roughly into three categories, the tracking algorithm, the thinning algorithm and the algorithm to use an intersection of two surfaces. The tracking algorithm and the thinning algorithm are often used to get line figures from figures like thick lines or cylinders. The algorithms introduced in the paper for extraction of center lines of vessels or bronchus belong to either of them. The algorithm to calculate an intersection of two surfaces is introduced as the method to extract characteristic lines of skulls in this paper.

[-> return]

title	author	page
Morphology for Three Dimensional Image Processing	Hidefumi KOBATAKE	168-173
Key words : Morphology, Dilation, Erosion, Opening, Skeleton, 3D image processing
Abstract : Morphology is one of the important and basic tools for image processing. Its mathematical operation is in general defined in multi-dimensional space and therefore it can be applied to three-dimensional CT and MR images. In this article, morphological methods for segmentation, detection of lung cancers, analysis of pulmonary vascular tree structure, and shape-based interpolation are introduced. Computational load is another important factor in applying mathematical morphology. Fast algorithms are also introduced.

[-> return]

title	author	page
New Image Reconstruction Algorithm for Increasing Spatial Resolution in the Trans-axial Plane with Helical CT: Principles and Physical Characteristics	Katsuyuki TAGUCHI, Yukihiro OGAWA, Soji MIYASHITA	175-186
Key words : Computed tomography, Helical scan, Spatial resolution
Abstract : Two helical reconstruction algorithms (helical-QQs) have been developed to achieve higher spatial resolution in the trans-axial plane. The proposed algorithms consist of the following two parts: 1) channel doubling, which halves the sampling pitch in the trans-axial plane, and 2) longitudinal (helical) interpolation. Both physical evaluations (spatial resolution and image noise) and preliminary clinical studies involving the temporal bone were performed to evaluate their performance compared with current algorithms (360LI and 180LI). The proposed methods markedly improved the spatial resolution with helical scanning, one (360QQ) permitting allows us to achieve 13.7-14.2 lp/cm in the trans-axial plane. Further clinical evaluations are expected in temporal bone, head CTA, and lung imaging.

[-> return]

title	author	page
Coronary Artery Reporting System with Arterial Figure Generation and Stenosis Detection for Computer-Assisted Diagnosis	Yoshio YANAGIHARA, Hiromitsu HAMA, Tetsuo SUGAHARA	187-195
Key words : Vessel figure, Arterial stenosis detection, Vessel extraction, Computer-assisted diagnosis, Medical image processing
Abstract : The measurement of vessel diameters by a computer system is important to avoid errors in the visual interpretation of coronary arterial stenosis. Such errors include inter-observer variability, intra-observer variability and excessive interpretation. In this paper, we describe an automatic stenosis analysis system which automatically generates an arterial figure showing the clinical names of the coronary arterial branches and indicates the location of on the figure if stenosis is present. The operator enters the position connecting two main vessels and the exposure direction into the system. Six cineangiograms were used in this study. The results of the study showed that 4 of the 5 stenotic segments were detected and 2 other positions were overestimated. The average extraction rate for the main arteries (LAD and CX) was about 95% and the error detection rate for branches was about 14%.

[-> return]

title	author	page
Approach Based on Eigenvalues of Hessian Matrix of Gaussian-Convolved Volume Density to Discriminate Pulmonary Nodules from Vessels within Data from Multi-Detector Row CT Examination of the Thorax	Tomoya KURAMITSU, Kenji SHIMIZU, Yuji SHIBATZ, Minoru TANAKA,¡¡ Yoshifumi YASUHARA, Junpei IKEZOE, Yoshinobu SATO,¡¡Shinichi TAMURA, Hiroshi YANAGIHARA	196-207
Key words : Three-dimensional Gauss convolution, Three-dimensional derivative feature, Hessian matrix, Pulmonary nodule, Multi-detector row CT
Abstract : With the advent of multi-detector row CT, a vast amount of isotropic volumetric data can be acquired in the clinical setting without the need to interpolate the axial slices along the z-axis, which promotes a truly 3-D approach in computerized schemes to automatically detect pulmonary nodules within the image data from CT examination of the thorax. In computerized schemes for the detection of nodules, discrimination of actual pulmonary nodules from normal structures such as vessels is essential to reduce the number of nodule candidates. In this technical report, we show that the product of eigenvalues of the Hessian matrix of the volume density function can discriminate the differences in three-dimensional local density structures of nodule and vessel models mathematically formulated as three-dimensional objects, both by theoretical analysis and numerical calculations with synthesized data. This three-dimensional derivative feature, combined with a three-dimensional isotropic Gauss convolution operation for blurring, has been applied to clinical data acquired with multi-detector row CT and has demonstrated its potential to discriminate actual pulmonary nodules from vessels in selected cases.

September(Vol.19, No.5)

[-> return]

title	author	page
Depiction of a Cerebral Arterial Network from 3D MRA Image Data Using A Blood Vessel Model Based on an Interpolation Curve	Masayuki HIKICHI, Masahiko HATANAKA	345-354
Key words : MR angiography, Blood vessel model, B-spline curve, Bezier curve
Abstract : Magnetic resonance angiography (MRA) is one method for obtaining 3D images of enhanced blood vessels. Recent developments in MRA have made it possible to obtain images with good contrast between blood vessels and soft tissues, and there has recently been great interest in developing methods for navigational diagnosis along vascular networks.In this report, we describe a "blood vessel model" based on an interpolation curve for depicting a cerebral arterial network, which is necessary for tracking blood vessels, and present an algorithm for determining the control points of the interpolation curve. Using the interpolation curve, we attempted to depict a cerebral arterial network fitted to arteries extracted from a 3D MRA data set. We then evaluated the differences between the obtained network and the center lines of the extracted arteries.

[-> return]

title	author	page
Use of Dual Fresnel Transform Pairs to Improve Signal-to-Noise Ratio in Magnetic Resonance Imaging	Satoshi ITO, Yoshifumi YAMADA	355-369
Key words : SNR, MR image, Fresnel transform, Wiener filter
Abstract : A new method for improving the SNR in MR images using dual Fresnel transform pairs is proposed. In this method, the scale of the image can be changed arbitrarily by changing the scaling parameter D as in wavelet transformation. Wiener filtering is then applied in the scaled image domain. By simulation experiments, it is shown that image SNR improvement would be almost maximum and deterioration in spatial resolution small when the scaling parameter is set to 8. The image SNR improvement is compared with those obtained by using an adaptive nonlinear filter and a Wiener filter in the wavelet domain. The results show that the proposed method has the largest SNR improvement when applied to MR image models, and the filtered image does not show the serious deterioration seen in the other two methods when the SNR is low. The proposed denoising method thus offers a practical filter which provides high SNR improvement without compromising spatial resolution significantly and which maintains the phase information required for flow measurement and angiography.

[-> return]

title	author	page
Image Reconstruction Method with Modification of the Number of Projections in a Baysian Approach with a Rescaled Block Iterative Algorithm	Kotoko MORIKAWA, Koichi OGAWA	370-376
Key words : Single photon emission CT, Iterative image reconstruction, Ordered subset, Maximum a posteriori-expectation maximization
Abstract : An ordered subset (OS) algorithm or block iterative (BI) algorithm is a useful method for accelerating image reconstruction in ML-EM or MAP-EM. Generally, the OS algorithm uses a fixed number of projections, so-called "subset levels."The subset level affects the recovered frequency component of a reconstructed image. In 1998, we proposed a new method in which the number of projections in a subset for each iteration step in an OS-Baysian reconstruction (BR) algorithm is modified to recover various frequency components in early iteration steps. In this paper, we introduce our algorithm for the rescaled BI version of the MAP-EM algorithm and report our results for the performance of the proposed method compared with conventional methods. The simulation results showed that the proposed method is superior to conventional methods.

[-> return]

title	author	page
Fourier Synthesis Method for Exact Cone-Beam Reconstruction	Hiroyuki KUDO, Noriko MIYAGI	377-388
Key words : Tomography, Image reconstruction, Cone-beam, Radon transform
Abstract : Exact cone-beam reconstruction methods for data acquisitions with non-planar vertex paths are usually based on the Tuy-Smith-Grangeat formula, which links cone-beam projections to the 3D Radon transform. The computation of the 3D Radon transform in these methods leads to drawbacks such as high computational load and significant discretization errors. We propose a new exact cone-beam reconstruction method which does not explicitly compute the 3D Radon transform. The new method is based on two familiar concepts in 3D tomography : approximate ramp reconstruction (cone-beam backprojection after ramp filtering) and Fourier synthesis using the central slice theorem. Simulation results demonstrate that the computational load of the newmethod is lower and the discretization errors are fewer than in conventional methods. In addition, as in other exact methods based on the Tuy-Smith-Grangeat formula, the new method allows flexible redundancy handling.

[-> return]

title	author	page
Shear Modulus and Viscosity Measurement Using MR Elastography	Osamu OSHIRO, Mikio SUGA, Makoto OHTA, Tetsuya MATSUDA, Sadami TSUTSUMI, Kotaro MINATO, Kunihiro CHIHARA, Takashi TAKAHASHI	389-399
Key words : Shear modulus, Viscosity, MRI, MRE, Instantaneous frequency method
Abstract : This paper describes a method of estimating the shear modulus and viscosity of an elastic material by magnetic resonance elastography (MRE) using the instantaneous frequency method. The shear modulus was calculated from the phase of the observed MRE signal and the viscosity from the amplitude. First, we performed a simulation to confirm the suitability of our proposed method. The results indicated that the method is appropriate for estimating the elasticity parameter values and is robust against noise. Then, we estimated the elasticity parameter values of a phantom and compared them with those obtained by dynamic testing. There was a good correlation between the two sets of values.

[-> return]

title	author	page
Computer-aided Diagnosis of Mitral Valve Stenosis Based on Maximum Blood Vessel Diameter Ratio in Upper and Lower Lung Fields in Chest X-ray Images	Hiromitsu AKASHI, Hiroshi MASUYAMA, Shoji KIDO, Shinichi TAMURA	400-403
Key words : Computer-aided diagnosis, Mitral valve stenosis, Blood vessel diameter, Chest X-ray
Abstract : The blood vessels in the lower lung become thicker than those in the upper lung in patients with mitral valve stenosis. On the basis of this physiological finding, we have been investigating methods for comparing the area ratio of blood vessels in the lower and upper lung fields. When the blood vessels are locally thickened and the total lengths of the blood vessels differ, conventional methods suffer from problems in evaluation and it is difficult to apply computer-assisted diagnosis. We therefore propose a new method for comparing blood vessel diameters in the upper and lower lung fields to overcome these problems. Blood vessel diameters are determined by measuring the maximum diameter of the blood vessels in an ROI using a Gauss-Hessian matrix. The blood vessels in a localized area of a 2D image can be considered to be line segments with a given width. Therefore, the value 2σ, which is the maximum value of the eigenvalue of the Gauss-Hessian matrix calculated at each point in the image using a variable scale parameter σ, can be considered to represent the maximum diameter of the blood vessels in the ROI. We can discriminate between patients with mitral valve stenosis and normal subjects by comparing the σ values obtained from the upper and lower lung fields. This method does not require the explicit extraction of blood vessels, making it possible to perform automatic diagnosis by simply setting ROIs.

November(Vol.19, No.6)

[-> return]

title	author	page
Reconstruction of Mathematical Logic for the Inverse Radon Transform Algorithms	Tamon INOUE, Takeyuki HASHIMOT0	427-235
Key words : Computerized tomography, Image reconstruction, Projections, Radon transform, Attenuated radon transform, SPECT
Abstract : The performance of analytical algorithms for CT image reconstruction from projections is evaluated in connection with the applicability in real situations. Mathematical details of those inverse problem solutions for both cases of the Radon and the attenuated Radon transforms are analyzed particularly with regard to the analytical characteristic of functions which express an object and its projections. The practical efficacy of mathematical conditions imposed for complete reconstructions is also investigated.

[-> return]

title	author	page
Image Reconstruction in Helical CT	Katsuyuki TAGUCHI	436-443
Key words : Computed tomography, Helical scan, Image reconstruction
Abstract : Image reconstruction algorithms in helical scanning with three types of scanners are described: Single-slice CT (SSCT) with single detector row, multi-slice CT (MSCT) which incorporates multiple detector rows with narrow cone-angle, and MSCT with large cone-angle. The former MSCT uses fan-beam geometries, while the latter one does cone-beam geometries. Research on image reconstruction algorithms is one of the hottest topics in medical imaging, especially for dedicated clinical applications and dynamic volumetric imaging.

[-> return]

title	author	page
Data Processing Technique in MRI: performance improvement on open MRI apparatus	Tetsuhiko TAKAHASHI, Masahiro TAKIZAWA, Kazumi KOMURA,Yumiko YATSUI	444-455
Key words : MRI, Data processing, Diffusion, Fat water separation, Temperature map, Parallel imaging
Abstract : Among current performance improvement on open MRI, one that dedicated data process algorithm contributes to realize is presented, i.e. parallel MRI, diffusion weighted image, fat water separation and temperature map. These topics also show current tendency on data processing on MRI.

[-> return]

title	author	page
Data Correction and Image Reconstruction Methods on SPECT	Nobutoku MOTOMURA	456-461
Key words : SPECT, Scatter correction, Attenuation correction, Reconstruction, Transmission CT, Iterative method, Quantitative measurement
Abstract : Data corrections in SPECT are performed for elimination of image artifacts, improvement in image contrast and quantification of data. In quantitative measurement, corrections for factors such as scatter, attenuation and collimator response are required. Currently, there are several practical methods to perform scatter correction. For attenuation correction, acquisition systems to obtain attenuation coefficient distribution and algorithms for reconstruction using the acquired data have been developed. These data correction techniques have enabled SPECT quantitative measurement to be used routinely in clinical practice. In this paper, an overview and evaluation of several scatter and attenuation correction methods are given.

[-> return]

title	author	page
Data Correction and Reconstruction Methods in PET	Keishi KITAMURA	462-467
Key words : PET, Reconstruction, Attenuation, Scatter, Random coincidence, Detector sensitivity
Abstract : In order to obtain high quality and quantity PET images, it is important to use appropriate data correction methods to reconstruct data affected by statistical noise, attenuation, scatter, random coincidence, detector sensitivity, etc. This paper describes the principle and practice of data correction and reconstruction methods based on a PET system model.

[-> return]

title	author	page
Detection of Lung Nodules on Digital Chest Radiographs	Jun WEI, Yoshihiro HAGIHARA, Hidefumi KOBATAKE	468-476
Key words : Digital chest radiography, CAD, Adaptive ring filter, ROC analysis, Lung cancer
Abstract : Lung cancer is one of the most serious cancers. Computer-aided diagnosis (CAD) systems have proved to be effective in assisting radiologists in X-ray image diagnosis. This paper describes a new system for detecting lung nodules on digital chest X-ray images. The proposed system consists of four processing steps. These steps are the identification of candidate lung nodules with an adaptive ring filter, the extraction of their regions, the extraction of feature parameters, and the discrimination of normal and abnormal regions using a maximum-likelihood method. In the preliminary experiments, 56 features were evaluated by ROC analysis and eight of them were selected. Furthermore, the experiments showed that the output image of the adaptive ring filter and the differential image of the original image contain information that is effective in differentiating between normal and abnormal regions. The performance of the proposed system was evaluated using 247 chest radiographs. The experimental results showed that the area under the ROC curve in the proposed system was 0.82, and the system sensitivity and the average number of false positives per image were 80% and 5.6, respectively. These experimental results have demonstrated the effectiveness of the proposed system.

[-> return]

title	author	page
Fast Volume Rendering by Software : Usefulness in a Virtual Endoscopy System	Masahiro YOSHIOKA,Kensaku MORI, Yasuhito SUENAGA, Jun-ichiro TORIWAKI	477-486
Key words : Fast volume rendering, Interactive visualization, Multimedia instruction set, Virtual endoscopy system
Abstract : In the medical field, volume rendering has recently been widely employed as a method for visualizing 3-D gray-scale images such as X-ray CT and MR images. It is expected that fast volume rendering methods will be developed in order to realize interactive manipulation environments for volumetric images. Our implementation of the fast volume rendering method by software employs optimization of program codes using SIMD instruction sets and also parallel processing using low-cost, general-purpose PC clusters. This parallel processing method balances the load on the PCs in order to equalize the processing time in each processor. We have implemented this method in a virtual endoscopy system and have confirmed that the method can speed up rendering processing.

[-> return]

title	author	page
Implementation and Evaluation of an All-Digital Secure Health Examination System	Masato MORI, Mayumi KITANO, Masato HORlNO, Shinichiro NlSHIDA, Toshio TAMARI, Minoru HOSOBA, Kotaro MINATO	487-494
Key words : PACS, Secure communication, Health examination, ISCL (Integrated Secure Communication Layer)
Abstract : In this paper, we assess the impact of the introduction of a PACS at a health examination center. The all-digital system was implemented based on the guidelines of MEDIS-DC concerning the electronic storage of clinical records. In this practical-scale system with secure communication, we measured the working hours of medical doctors and technicians in order to evaluate the effectiveness of the system and found a 7% reduction in mobile DR and a 24% reduction in in-house DR in comparison with a conventional film-based system.