A picture archival and communications system has been under development at the Hospital of the University of Pennsylvania for the past two years. This system features a fiber optic network, high resolution display for initial interpretation, and a lower resolution display for review and comparison. A laser optical disk will be used for mass storage. A high resolution digital camera is used to acquire images from x-ray film. This manuscript addresses an analysis of the image resolution, intensity transformations, and noise associated with the digital scanning camera and the high resolution cathode ray tube displays. A linear systems approach is utilized for this analysis.

This paper initially describes the functional specifications of a generic digital workstation for medical images. These generic specifications have formed the basis of a first generation PACS design currently in clinical evaluation. Preliminary observations of the first generation system in a clinical environment suggests an oppor-tunity to optimize both performance and cost effectiveness. Functional specifications are presented for a series of workstations specializing in four major applications, Image Acquisition, Image Review, Image Processing, and Image Archiving. The tmportance of performance and the impact of system architecture on the individual workstation specifications will be clearly noted. The relative composition of hardware and software specialization in each workstation design will also be described. The specification of a useful multimodality workstation is therefore largely determined by application and performance expectations and significantly influenced by the particular system architecture chosen.

The Radiology Department of the University of Arizona Health Sciences Center is involved in a five year project to develop a prototype for a Totally Digital Radiology Department (TDRD). One prerequisite for such a department is that the system must allow the radiologist to perform at the same level as with the existing film-based system. The amount of spatial and contrast resolution required to equal the diagnostic information in films will impact directly on the capabilities of the equipment needed to image, transmit, display, and store such images. This is a crucial question that must be addressed because of its diagnostic import and due to its engineering implications.

Electronic image display has been envisioned as a means of alleviating some of the problems that expanding radiology departments have with film management, integration of multiple imaging modalities and as a means of improving the physician's diagnostic abilities through after acquisition image contrast enhancement. In an effort to provide an electronic image display suitable for general use, several educational institutions and industry have been in the process of developing computerbased image systems called picture archival and communication systems (PACS). Functional integration of the PACS components (network, archive, database management and display) presents problems which require the development of departmental modeling, a common communication protocol, image storage devices with enormous capacities, management software and display consoles. The latter has been the object of our focus.

This paper reports on a recent effort at the Lister Hill National Center for Biomedical Communications where it is shown that a medical bibliographic database, when linked to a mini-computer's file handling software, can be used to build a demonstration database of document images. The images of the tables of contents from approximately fifty medical books were stored on magnetic disks. The medical bibliographic database, hosted on a remote mainframe computer, was linked to the minicomputer controlling the magnetic disks. The result is that users who do bibliographic searches on a particular medical subject may view, at their work station, the tables of contents and title page from selected citations. The result is also that the document image database is accessed and managed using pre-existing features of the bibliographic database and the minicomputer's operating system.

An extension of the data base language SQL is described which provides tools for handling images in an interactive environment. It is implemented in a prototype pictorial information system for radiology. The properties of the language are demonstrated with several examples. It is claimed that extending the facilities of a conventional DBMS yields an economic and powerful tool for uniform management of image and non-image data in a pictorial information system.

Six design criteria for a local area nuclear medicine PACS are discussed, with emphasis on novel but appropriate goals. A PACS which has been designed to meet these goals is described. After extensive field experience, the system has proven to meet or exceed the specified criteria.

This paper describes the application of a Realtime Digital Disk (RTDD) with processing capabilities and the increased functionality the RTDD brings to a PACS system. While the display and acquisition of images at 30 frames per second is the obvious advantage of the RTDD in PACS, the image processing and image enhancing functions make its use more effective. Realtime presentations include window and level of greater than 8 bit images, intensity transformation of 8 bit images, shape table and tiling presentation, and subsampling. Movie loop and three dimensional reconstruction sequences are viewed in realtime. Edge enhancement, smoothing, interpolated zoom, unsharp masking and other convolution filtering are near realtime presentations. Compression and decompression, histogram equalization and other transforms are performed in just a few frame times.

One of the key challenges confronting the designer of a PACS network is to reconcile the conflicting requirements for advanced functionality and cost-effectiveness of viewing consoles. This paper describes an image processing system architecture, which permits the distribution of sophisticated image processing and analysis capabilities among a large number of inexpensive consoles. The individual viewing consoles are simply the display modules of a multi-user, multi-tasking image processing system, which have been equipped with computer controllers and network interfaces. The network acts, in effect, as an extension of the image processor's data bus. The remote consoles have basic processing capability, such as gray scale manipulation and zoom, so that the majority of image review activities are performed locally and do not tie up the network. Infrequently-used, but important, capabilities such as spatial filtering are available from the parent image processor via the network when needed.

A design study of a Picture Communication Network is presented with special emphasis on those issues that differ from conventional networks. The large amount of data in an image together with the requirement for a rapid respons time make a wide-band transmission medium a nec ssity. The wide-band medium is perceived as a collection of parallel broadcast channels. Four different logical organizations are proposed for t is architecture. Design equations are developed and cost-performance curves are generated. Advantages of one organization over another under different circumstances are pointed out.

The Department of Radiology at the University of North Carolina, Chapel Hill has embarked on a computerized simulation of the Department to aid in the development of a Pictorial Archiving Communications System (PACS). The project when completed will provide us with simulation tools that support the testing of different configurations of the PACS system as well as generate performance and cost data for both PACS and non-PACS departmental operations. We have recently completed the first phase of the project which we will discuss in detail.

This paper describes the implementation of a commercial PACS system. A fairly detailed presentation of the hardware components is given, and an overview of the software is also provided. The architecture presented has divided the system functions so that user performance is very predictable. The hardware and software also allow the user to interact with the system in a very friendly manner.

This paper describes the design of a Local Area Network (LAN) diagnostic package This program will contain "confidence" type test routines, and commands designed to aid in detecting failures and improve the use of the LAN services during normal operations. The approach selected requires installing the same copy of the program in selected bus interface units(bius) and a subset in all the others bius. Those bius containing the full set of options are the controlling nodes; more than one controlling node can run concurrently its own diagnostic program. Messages generated when exercising the different resources and interfaces of the LAN will be stored and/or displayed on appropriate positions. As a minimum the LAN will have one controlling node which will be the location of the LAN's technical control. The diagnostics will provide data to aid in mana-ging the system, accepting or declining requests for connections, allocating additional ports, declaring sectors inoperable, off lining nodes, devices, functions, and reconfiguring the system as necessary. Host computers can optionally have the full blown diagnostic program, which allows them to be a controlling node, or a subset of the diagnostic program with enough capabilities to interface with its host front end biu (hfe). The diagnostics will have its own application protocol and will utilize the ser-vices of the Internet Protocol (IP) to transport messages to their destinations.

In a Picture Archiving and Communication System (PACS) the image transmission performance affects the performance of the system overall. This report places particular emphasis on evaluating the image transmission performance of the network, and on examining the parameters which are most important in the development of a PACS system. We conducted computer simulation studies to determine how far image transmission is affected by such parameters as the number of imaging diagnostic machines within the hospital, their image generation frequency, the size of the images, and the performance of the diagnostic equipment itself.

The Medinet model accounts for cost items in two major categories: materials + equipment and labor (archive and communications). By adjusting the model's parameters for local or regional variances from the Medinet standard, an imaging department can obtain a reasonably accurate assessment of costs and relative efficiency associated with their manual PACS for digital diagnostic modalities.

In a digital imaging department, images are viewed using PAC systems and CRT terminals. Although the need for hard copies is greatly reduced, it is expected that some hard copy images will continue to be required for patient charts and communication to referring physicians. This report considers the utility of non-film modalities for such applications that do not involve the primary diagnosis and compares the acceptability of three possible modalities.

Rapid changes in digital technology and its application to medical equipment demand a standard communications interface which system integrators can use as a common and stable link between equipment of different type, manufacture and vintage. A Standard Product Interface fulfilling this need is proposed for consideration for industry-wide use. It relies on a message-passing protocol with a standard message format, standard message transport functions, and a separate physical link to make it independent of a particular imaging equipment, local area network, or medical information system.

The Toshiba computed tomography scanner system TCT60A/500X is equipped with an optical disk data storage device for image data archiving. One optical disk stores 3.6 gigabytes of data, or 15000 CT images on both sides. When writing on an optical disk, one spiral of data pits is produced with a semiconductor laser by evaporating the Te-C film coated on the PMMA (poly(methyl methacrylate)) substrate. The pits are read by the same laser at a lower power along with CRC (cyclic redundancy code) error correction. A bit error rate of 1.0E-12 was attained. The IEEE488 interface bus (GPIB) is used to communicate with a host computer. The mean data transfer rate through the bus is 100 kilobytes per second.

The complexity of a Picture Archiving and Communication System (PACS) will induce the novice to guess system responses. Such guesses will be based on a conceptual model. Frustration will result if actual responses deviate from the expected ones. The paper suggests that human services offered in a radiology department can form the basis for suitable conceptual models both with respect to information and to attitude.

Ultrasonic images of the heart, i.e. echocardiographs, have been in use in diagnosis and treatment for some time, the quality of the images is not good. This paper describes a procedure for enhancing and improving the quality of echocardiographs. The procedure described has three main aspects: noise filtering or smoothing, edge extraction and contour following. A number of filtering algorithms to eliminate noise and smooth the image are presented and their advantages reviewed. Several edge operators are considered and their merits discussed, and a modification to one is presented. Contour tracing algorithms are studied and a modification proposed. The resulting procedure is illustrated in each step.

The development of computer tomography (CT) in radiography has stimulated recent interest in picture archiving and communication systems (PACS) in medicine. Electron microscopy and histopathol-ogy also use interpretation of images in diagnosis. We have implemented a prototype system for on-line digitization of tissue images as a part of PACS based on DEC PDP 8, VAX 11/730 and PDP 11/35 processors. The front end processor was directly connected to a JEOL 100B scanning transmission electron microscope (STEM). Because image digitization results in a loss of information that is present in photography, various methods of image enhancement and filtering were necessary and helpful. Data compression by discrete cosine transform (DCT) retained satisfac-tory image quality at 0.5 - 1.0 bits/pixel. It is also stressed that future PACS in medi-cine must incorporate images of diagnostic histo-pathology and electron microscopy.

Computed Tomography (CT) images reconstructed using a limited number of projections, measured over a narrow angle range, are characterized by approximately elliptical distortion along the view angles used, and poor contrast at angles not used. This systematic geometric distortion is caused by the two dimensional point spread function of the reconstruction process. In this paper, we show that such geometric distortion and other artifacts introduced in the reconstruction process can be reduced substantially by deconvolution performed via Wiener filtering using a priori knowledge derived from the given projections. The two-dimensional system transfer function used in the deconvolution is obtained from the reconstruction of a test image by a linear reconstruction algorithm (unconstrained multiplicative Algebraic Reconstuction Technique).

In the image formation methods, using nuclear magnetic resonance (NMR) techniques, static magnetic field with spatial uniformity is used. The spatial fluctuation of the static field causes distortion of the images. We propose a method for correcting the image distortion in a case that the direct Fourier transform image formation method is applied. The correction procedure is as follows. (1) The distorted image of a standard phantom, which is composed of many point objects, is reconstructed. (2) The displacements of these point objects in the image are calculated. (3) The non-uniform static magnetic field is estimated from those values. (4) The image distortion of a desired object is corrected using the estimated value. The feasiblity of the proposed method is confirmed by experimental results.

Various components of the x-ray imaging process contribute differing forms of degradation. One of the most significant sources of blur is object scatter which cannot be fully eliminated by grids and often contributes more than 60% of the final image exposure. Because the amount of scatter varies locally as a function of attenuation, the low pass filtering which this blurring creates is spatially variant, being most severe in the most attenuated areas (lowest density - most translucent parts) of the image. By taking advantage of this density dependent property we have devised a technique which restores high frequencies only in low, but not in high density areas. The method first creates a high frequency Laplacian filtered version of the original which produces reduced and enhanced high frequency amplitude in originally dark and translucent areas respectively. This high frequently image is then multiplied by a low-pass version of the original image. By then adding the product image to the original a final result is obtained in which only the translucent areas of the image receive significant high frequency restoration. The technique is capable of rapid implementation and can be additive with image processing methods aimed at other, less spatially variant sources of blur.

In order to promote the development of PACS it is necessary to develop standards for medical image exchange. This paper will present a draft for the medical standard format of a magnetic tape for digital image exchange from the Japan Medical Industries Committee. This format consists of three sections; a closed directory, data blocks and an open directory. The closed directory would include patient directories and image directories. The data blocks include a header and frame for each block. Each header comprises an image size definition record,a directory record, a link information record, a location and time information record, comment record, a modality dependent condition record and an additional information record. The image size definition record is unaltered from the form specified in the standard format of magnetic tapes for digital image exchange issued in 1976 by the Image Processing Committee of the Information Proccessing Society of Japan[1]. Medical information is added to the comment area of the header to make this format applicable for the recording of medical images of any of the current imaging modalities.

This paper discusses yet another image operation performed in real time by the LCLV. Because of the birefringent properties of the liquid crystal used in its manufacture, a light intensity distribution imposed on the input plane of the liquid crystal maps as a reflectance function proportional to wavelength on the output plane. Using this 'spectral shift' of the light reflected from the output plane an optical device can be constructed which will cause dispersion proportional to input intensity gradients, which, in turn, enables the device to produce depth and shadowing effects in real time. A brief description of the operating characteristics of the LCLV will be discussed along with a method for using the 'spectral shift' distribution produced on the output plane of the device. Algorithms for producing stereo effects are reviewed followed by a description of the configuration of a system incorporating the LCLV to produce stereo effects. Finally, possible future uses for the device are suggested and further work proposed.

THE VIDEO DIFFERENTIAL PLANIMETER (VDP)1 is a re-mote sensing instrument that can measure minute changes in the area of any object seen by an optical scanning system. The composite video waveforms obtained by scanning the object against a contrasting back-ground are amplified and shaped to yield a sequence of constant amplitude pulses whose polarity distinguishes the studied area from its background and whose varying widths reflect the dynamics of the viewed object. These pulses are passed through a relatively long time-constant capacitor-resistor circuit and are then fed into an integrator. The net integration voltage resulting from the most recent sequence of object-background time pulses is recorded and the integrator is returned to zero at the end of each video frame. If the object's area remains constant throughout the following frame, the integrator's summation will also remain constant. However, if the object's area varies, the positive and negative time pulses entering the integrator will change, and the integrator's summation will vary proportionately. The addition of a computer interface and a video recorder enhances the versatility and the resolving power of the VDP by permitting the repeated study and analysis of selected portions of the recorded data, thereby uncovering the major sources of the object's dynamics. Among the medical and biological procedures for which COMPUTER-AIDED VIDEO DIFFERENTIAL PLANIMETRY is suitable are Ophthalmoscopy, Endoscopy, Microscopy, Plethysmography, etc. A recent research study in Ophthalmoscopy2 will be cited to suggest a useful application of Video Differential Planimetry.

The constant call for shorter development cycles and improved quality, is leading many companies to adopt a vertically integrated approach to product design and development. Computer aids are playing an increasingly important part in this process, but until recently, the early stage - concept design, has been largely ignored. Computer generated "solid models" provide a useful tool at the early stages in the design process, enabling investigation and optimization of clinical performance variables.

We have developed a system which can recognize speech and generate the corresponding animation-like sign language sequence. The system is implemented in a popular personal computer. This has three video-RAM's and a voice recognition board which can recognize only registered voice of a specific speaker. Presently, fourty sign language patterns and fifty finger spellings are stored in two floppy disks. Each sign pattern is composed of one to four sub-patterns. That is, if the pattern is composed of one sub-pattern, it is displayed as a still pattern. If not, it is displayed as a motion pattern. This system will help communications between deaf-and-mute persons and healthy persons. In order to display in high speed, almost programs are written in a machine language.

This report describes a prototype dental x-ray machine designed to eliminate shortcomings of conventional dental radiographic methods by using multiple image sampling techniques, electronic x-ray detection, and computer image processing. A miniaturized intraoral x-ray tube is used having an enlarged target on which multiple focal spots can be excited sequentially under computer control. The detector is a video-based image intensifier yielding a sequence of projections which are stored digitally in real time. Suitable processing of these images permits synthesis of tomographic slices. Specific projections obtained at different examinations can be matched so that subsequent subtraction demonstrates diagnostically important changes.

Currently there are several methods for creating images that relate to either the anatomy or function of the human brain. Static pictures of anatomical structures are produced by computerized axial tomography (CAT) and the presently evolving methods of nuclear magnetic resonance (NM P). These images give excellent resolution of various degrees of structures but do not reveal any aspects of function. A variety of radioactive labeling and detection techniques are available that produce images related to brain activity.

An ophthalmic imaging system, the IS-2000, has been developed for acquisition, archival storage, and analysis of fundus imagery collected during fluorescein angiography and other ophthalmic procedures. The system consists of a conventional fundus camera equipped with a video camera; a microcomputer with color video digitizer and frame buffer; and an optical video disk for archival storage of images. In the course of development of the IS-2000 a study was performed to assess the spatial frequency content of typical angiograms and to evaluate the resolution capabilities of a variety of types of commercially available video cameras and image storage media. The frequency content of angiograms was analyzed through the application of Fourier transforms to high-resolution digitized images. Modulation transfer functions (MTF) of various types of video cameras and video disks were measured with appropriate video test images. An assessment of the suitability of currently available video equipment for collection and storage of fluorescein angiograms is provided.

Digital holograms of semitransparent objects, such as human cells, have been experimentally obtained through a microscope objective and acquired through a TV acquisition system, A/D conversion and fast storage. The problem of generating 3D images of these objects starting from their digital holograms through a computerized reconstruction algorithm is here considered.

A practical analyzing procedure has been developed to evaluate dynamic images with reference to time series data from the same object. A special recording and analyzing system has been contructed, and its validity and an analyzing procedure are demonstrated using actual dynamical cardiac images. Echocardiogram and X-ray TV images were recorded on the video tape simultaneously with physical parameters. These data were fed into the computer and analyzed in reference to each other. Actual experiments were then executed to analyze myocardial ultrasound and X-ray images quantitatively, using subtraction and densitometry techniques. In both images, abnormalities of myocardial perfusion were clearly described by the subtraction image with reference to the ECG signal and the M-mode image, in the case of the echocardiogram. The temporal change of myocardial images was also represented by subtraction, and the details were described by densitometry. These image processing techniques can be a useful tool in evaluating cardiac dynamics in daily practice.

Digital angiography is widely considered simply as a method in which images taken at different times are subtracted from each other. This paper presents some techniques which are performed in the frequency domain after the application of the Fourier Transform. Nonselective bypass angiograms and intravenous ventriculograms are taken as examples to show that simple procedures utilizing these techniques exhibit the advantages of improved signal to noise ratio in the subtraction images, reduction of motion artefacts, easy application of phase-synchronous subtraction, integration and quantitative visualization of blood propagation.

A new approach to blood vessel boundary estimation from a few whole volume projection images addressing the problems of low S/N ratio, low sampling rate and optimum data utilization, and the clinical need for three dimensional information about the vessel is presented in this paper. It is based on a recently developed estimation approach and algorithm for 3-D reconstruction from projections with incomplete and very noisy data. Using a stochastic dynamic model for the vessel, the imaging problem is reformulated as a nonlinear state estimation problem of small dimensionality, and an approximate recursive MMSE smoothing algorithm for its solution is presented. Computationally, the algorithm outperforms by orders of magnitude a previously suggested estimation method for blood vessel imaging, which was infeasible for the 3-D problem.

This preliminary report describes a segmentation algorithm which has utility in aiding the assignment of image elements to anatomical regions of interest. The algorithm does not depend on uniform image intensity throughout the region of interest and is not influenced by variance in the shape of the structures being identified.

Various methods have been proposed aimed at automatically finding the left ventricular (LV) boundary of the heart from several types of medical imaging modalities. One of the primary reasons why accurate boundaries are sought is for quantitative evaluation of the motion of the heart's walls. Few of these methods work consistently well on nuclear medicine images because of the inherently low signal-to-noise (SNR) ratio and low resolution of the data. This paper discusses a new algorithm which is a more knowledge-based approach to the LV boundary finding and wall motion determination tasks. Information obtained spatially and temporally from a gated sequence of nuclear medicine images is integrated in order to better isolate the LV boundary. Furthermore, information from several views of the heart (and possibly from other modalities) will be used to help with wall motion characterization.

Computed tomography (CT) scanners provide images of internal anatomy with unsurpassed spatial resolution. Since these images are inherently digital, computer systems can be used to simulate, plan, and guide surgical procedures to submillimeter precision. The combination of CT images, specially designed instruments, and the software to coordinate them results in improved accuracy for stereotactic surgery. This paper introduces per cutaneous cruciate ligament replacement as a new application for computer-aided ster eotaxi s. The procedure is described here with the knee firmly attached to a custom device. Twenty-five to thirty CT scans are performed to view the knee in detail, with special care taken to visualize the cruciate ligaments and their attachment to the tibia and femur. At the display console, two trajectories are chosen, using images delivered by the scanner and alternative views generated by software. These trajectories position two attachment shafts through skeletal structure in the knee to secure a replacement cruciate ligament. Interaction is illustrated that allows both the selection of the trajectories and the simulated surgery along their path. Anatomy intersected by the proposed trajectories can be reviewed in detail prior to actual surgery. Once reviewed, frame sittings are delivered by the computer system to drill the ligament attachment shafts. A replacement cruciate ligament can be passed through these two drill holes. Only an arthroscopic procedure is needed for later fixation of the ligament. In this manner, 4-6 hour open surgery of the knee is avoided and patient rehabilitation should be reduced from several months to 4-5 weeks. The frame, software, procedure, and computational aspects of the interaction are described. Test and patient results are given.

In order to describe the movement of the left ventricular wall, we have developed a method to access the movement of its centre of mass, using ECG-gated T1-201 scintigraphy of the left ventricle. To reject areas with T1-201 uptake outside the left ventricle, we had to delineate the object in scintigrams with extremely low signal-to-noise ratios, and to estimate a background count density. The DIODA-contour detection program showed even in these extreme circumstances its worth.

This paper presents a self-organized correction algorithm for the distorted measurements of the displacement pattern of the mandible recorded by the mandibular kinesiograph(MKG) based on the group method of data handling(GMDH) concept. A heuristic self-organization method aims to design the correction models characterized by the property that the corrected estimates of the mandibular measurements are the outputs of the nonlinear polynomial functions the input of which are the measurements correspondonding to the neighbouring nominal coordinates. The proposed method for building a correction algorithm was applied to the distorted trajectories of the measurements determined for the simulated mandibular incisor-point displacement performed on the stereotaxic device. The GMDH modeling yielded a mean estimation error for 45 coordinate variables of 0.27 mm. with a standard deviation of ± 0.31mm., thus suggesting a possible application of this method to the signal correction of the MKG.

Several methods for display of 3D-data are presented. Basic representations like look-through projections and surface display are combined with stereoeffects and interactively controlled motion to give a better 3D-impression. The effects of various preprocessing steps are discussed.

A generalized interactive modeling and graphical display system is being developed to treat 3-D random data. The modeling system generates an estimated surface using either a weighted average of a family of nearest neighbors or a bicubic surface patching technique. From such a model geometric properties can be calculated. Preliminary tests of the method using canine left venticles suggest that volume and surface area can be estimated within 10% of the true values. The modeled surface can be displayed as a wire-frame or shaded solid. The system can interactively display the model frame by frame with or without the original data points. A display mode is being implemented that allows displaying the model using animation techniques to simulate a beating heart.

3D83 is a software package for three-dimensional display from computed tomograms, which has been designed to put as little demand as possible on the user's time. It produces images which are similar to what would be seen if an organ was removed from the body, possibly cut open and then photographed from different directions.

This paper describes a technique to effectively visualize three-dimensional (3-D) images commonly given as sequences of two-dimensional (2-D) slice images such as microscopic serial section images and multiple CT slices. The method is based on the principle of binocular stereoscopy and the idea of semi-transparent viewing. The main problem addressed here is that of computing a stereo pair directly from a set of 2-D gray-level images. A transmitted-light model is introduced for determining each pixel value in the stereo images. Effectiveness of the method is verified in 3-D view of biomedical section images.

This paper presents a system for display-ing three-dimensional images of human brain structure by reconstructing them from CT image sequences. Functions of the system include : reslicing (pro-ducing cross-sectional images along an arbitrary plane); generation of 3D shaded surfaces of the skull, ventricle and disease lesions (hematoma and tumor etc.), windowing, and translucent display of the skull. All these functions can be combined in an arbitrary way to generate complicated 3D images. The system is expected to be useful especially for surgical planning and education in medicine.

Several straightforward techniques for displaying arbitrary solids of the sort encountered in the life sciences are presented, all variations of simple three-dimensional scatter plots. They are all targeted for a medium cost raster display (an AED-5l2 has been used here). Practically any host computer may be used to implement them. All techniques are broadly applicable and were implemented as Master Degree projects. The major hardware constraint is data transmission speed, and this is met by minimizing the amount of graphical data, ignoring enhancement of the data, and using terminal scan-conversion and aspect firmware wherever possible. Three simple rendering techniques and the use of several graphics cues are described.

In this paper, a real-time approach to the remasking of angiographic images is described and a real-time motion detection algorithm is presented. Various factors related to the algorithm's perform-ance are explored theoretically and these results are compared with experimental data.

The ability to interact with multiple images either from a temporal or spacial sequence has become a critical requirement for dislay stations in a PACS system. This paper presents the design and implementation of such a system.

At present electrocardiogram has been most developped in the area of heart diagnosis. But at the surface of body there is also the magnetic field which is induced from electromotive forces in the muscles of heart. This magnetic field has attracted doctor's considerable interest recently, and it is called magnetocardiogram(MCGD. In this paper, a method which we used for measuring the MCG data using a superconducting quantum interference device magnetometer is firstly explained. Next, in order to be able to evaluate the data more easily for doctor they are made into a movie using image processing by computer, then its contents and the result are presented.

With the growth of digital and electronic i maging techniques in medicine particularly useful are the 3-D data for diagnosis and therapy. The representation of 3-D objects adopted uses the 'octree' data structure and is derived from serial section 2-D ima ges, as in CT. The slice-images, pre-proces sed with segmentation techniques and then processed to obtain quadtrees, allow by sim ple technique to reconstruct 3-D represen tation with octrees. For the movement analx sis and generation is adopted a syntax- di rected tree-transducer.The time-varying mages are represented by a sequence of 8-trees and the matching process is performed by parser. Many rules adopted for motion a nalysis and generation are described. Key-words: quadtree, octree, serial-section motion-primitive, tree-translation, parsing.

A coronary cine-angiogram is a sequence of roentgen absorption images of the coronary arterial system, where each frame represents the projection of the arterial tree on a two-dimensional plane at a specific instant of the cardiac cycle. In single frames, the automated delineation of the projected arteries of the entire coronary tree can be achieved by means of a maximum brightness tracing procedure, followed by minimum cost contour detection. A method has been developed to guide the segmentation process in a specific frame by utilizing the a priori knowledge gathered from the previous frame. This a priori knowledge is represented in terms of a structural model of the arterial system, which is based on the tangent angle function of the various branches. As a result of this guidance user interaction may be reduced to a minimum.

This paper presents a computer classification system of rosette forming cells. In clinical immunology, it is important to distinguish lymphocytes into some subpopulations. For this purpose, rosette formation test is one of the most popular and useful methods. But recognizing rosetted and non-rosetted process depends on microscopic examination. We notice that an accumulative histgram of gray-levels presents characteristics of each rosette well. The system proposed here extracts rosettes from microscope image, constructs an accumulative histgrams of gray-levels for each rosette, and classifies rosettes into five groups using several features derived from the accumulative histgrams. Touching rosettes are also treated. The system is tested using 322 samples from 10 images. Total accuracy of classification is 73.3 %. And total correctrate of rosetted and non-rosetted dicision is 86.0 %. It is confirmed that computer classification of rosette forming cells is possible by the proposed method.

The cytologic and histologic detection and diagnosis of disease states and the monitoring of their progression and regression by human visual analysis may be improved by the design and application of an array of new instruments for and approaches to microscope imaging systems. The development of these tools and techniques allows us to visualize and analyze cells and tissues with increased accuracy, sensitivity and objectivity. The Johns Hopkins Biomedical Image Data Analysis System, BIDAS, is being developed to assist humans in their interpretation of clinical cytopathology specimens and in the evaluation and standardization of specimen quality. The BIDAS computer analyses are extremely sensitive to cell changes, they are objective, and their results can be accurately repeated. BIDAS consists of a complex of multifaceted subsystems. An overview of the four primary computer-based subsystems (1. Microscope Imaging and Data Acquisition; 2. Data Communications; 3. Data Processing; and 4. Data Analysis) is presented in this paper.

Tomosynthetic reconstructions suffer from the disadvantage that blurred images of object detail lying outside the plane of interest are superimposed over the desired image of structures in the tomosynthetic plane. It is proposed to selectively reduce these undesired superimpositions by a constrained iterative restoration method. Sufficient conditions are derived ensuring the convergence of the iterations to the exact solution in the absence of noise and constraints. Although in practice the restoration process must be left incomplete because of noise and quantization artifacts, the experimental results demonstrate that for reasons of stability these convergence conditions must be satisfied.

Two new image reconstruction algorithms, the "Metz filter method" (MF) and the "simple EM method" (SEM), have been developed for TOFPET. The results of these new methods were evaluated systematically in computer simulation studies and compared to those produced by three established algorithms: the most-likely position method (MLP), the confidence weighting method (CW), and the estimated posterior-density weighting method (EPDW). Relative computational efficiencies and image quality descriptors such as reconstructed image resolution, noise magnitude, and quantitative accuracy have been investigated. In most cases, the MF method appears to provide a superior compromise between acceptable image quality and reasonable computation efficiency.

Two major problems in the reconstruction of spatial distribution of radioisotope in single photon emission computed tomography (SPECT) are tissue attenuation and resolution degradation with distance. In order to describe the projection process in SPECT more accurately, a three-dimensional linear model which takes into consideration the medium attenuation effect and the point spread function (PSF) of the imaging system is presented. Based on this model, an iterative spatial reconstruction algorithms, i.e., the Spatially Modified Algebraic Reconstruction Technique (SMART) is proposed and tested by computer simulation and phantom study. Comparison of the result with other conventional SPECT algorithms indicates that SMART has better performance in attenuation effect compensation and discrimination of spatial features in phantom. Key Words: (1) Single photon emission computed tomography (2) Spatial iterative algorithm (3) linear model (4) evaluation of computer algorithms

A system for diagnostic evaluation of Technetium 99-m gated blood pool studies based on artificial intelligence and image understanding methods is presently under development at the University of Erlangen. A first version of the system has been completed recently. The present paper gives an overview of the system and reports first results.

Interactive or display-oriented pattern recognition algorithms can be utilized with advantage in the design of efficient computer-aided diagnostic systems. These visual methods may provide a powerful alternative to the pure numerical approach of data analysis for diagnostic and prognostic purposes. Functional as well as pictorial representation techniques are discussed in conjunction with some newly developed semi-fuzzy classification techniques. The blend between the two methodologies leads to the design of a very flexible, yet powerful diagnostic system. Results obtained when applying the proposed system on a group of patients representing several classes of liver dysfunction are also reported, to demonstrate the effectiveness of the proposed methodology.

Image processing by linear systems has some fascinating aspects due to the theorems of the Fourier transform (FT). Those, whose FT operator is a piece of glass are quite familiar to them, thinking in object space images and FT space images by reason of visual experience. But also in digital image processing the study of FT images is advantageous, e.g. when designing linear filters. The realization of an apodized Laplacian operator is shown, using an image of G-banded chromosomes as test objects. The quality of the filters is tested by the application of a contour-algorithm on-to the filtered images.

The increasing use of digital imaging techniques create a need for improved methods of digital processing, communication and archiving. However, the commercial opportunity is dependent on the resolution of a number of issues. These issues include proof that digital processes are more cost effective than present techniques, implementation of information system support in the imaging activity, implementation of industry standards, conversion of analog images to digital formats, definition of clinical needs, the implications of the purchase decision and technology requirements. In spite of these obstacles, a market is emerging, served by new and existing companies, that may become a $500 million market (U.S.) by 1990 for equipment and supplies.

Although the functions performed by the different nodes on the PACS network are many, it is possible to formulate a minimum set of service primitives such that the application software residing at the nodes can utilize those primitives to perform the functions. These primitives define the framework for the communication interface. The question of how these primitives fit into the concept of a layered network architecture is explored in this paper. The OSI model as applicable to the PACS network is described, the areas that need standardization are briefly mentioned, and the ongoing standardization efforts are addressed from the OSI perspective.

A chronic barrier to rapid progress in image processing and pattern recognition research is the lack of a universal and facile method of transferring image data between different facilities. Comparison of different approaches and algorithms on a common data base is often the only means for establishing the validity of results. Data collected under known recording conditions is mandatory for improvement of analytic methodology, yet such valuable data is costly and time consuming to obtain. Therefore, the sharing and exchange of image data may be expendient. The proliferation of different image data formats has compouned the problem of exchange. The establishement of logical formats and standards for images and image data headers is the first step towards dissolving this barrier. This paper presents initial recommendations of the IEEE Computer Society PAMI (Pattern Analysis and Machine Intelligence) and CompMed (Computational Medicine) Technical Committees' Database Subcommittees on the first of a series of digital image data standards.

Hardware and software of a computer-assisted image analyzing system used for infrared images in medical applications are discussed. The application of computer-assisted thermography (CAT) as a complementary diagnostic tool in centralized diagnostic management is proposed. The authors adopted 'Computer Assisted Thermography' to study physiological changes in the breasts related to the hormones characterizing the menstrual cycle of a woman. Based on clinical experi-ments followed by thermal image analysis, they suggest that 'differential skin temperature (DST)1 be measured to detect the fertility interval in the menstrual cycle of a woman.