In biomagnetic imaging, the magnetic field caused by electrical nerve impulses is measured and used to form an estimate of the location and strength of the impulse. One complicating factor in forming this estimate is the fact that the impulses induce current flow in the volume conductor surrounding the nerve.17 In this presentation we explore the properties of these volume currents. We first formulate the problem in the standard form using Ohm's law to relate the volume current to the impressed (nerve impulse) current and the conductivity distribution. We then depart from the usual derivation by making use of properties of the fourier-transformed maxwell8 and continuity equations. In fourier space, the divergence operation in a vector field becomes a simple taking of the radial component of the fourier-transformed field; the curl transforms into taking tangential components. By decomposing the current densities and using the maxwell equations, we are able to arrive at a recursive differential expression for the volume-current generated magnetic field. The driving term in the expression is the current due to the divergence of the impressed current density. We provide some examples of applying this expression to simply shaped conductors.

The feasibility of a new diagnostic imaging technique is investigated that potentially might be used for breast cancer screening with millimeter resolution, but without using ionizing radiation. It is suggested that acoustic pulses of sufficient intensity may produce small density changes within tissue which result in small but detectable changes in electrical current flowing through the tissue. The magnitude of this current fluctuation is shown to be inversely proportional to the conductivity of the tissue within the region occupied by the pulse. Measurement of the current modulation may enable small resistive inhomogeneities, such as tumors, to be detected. If the position of a pulse's wavefront can be predicted with sufficient precision at any given instant, measurement of the current modulation could be used to reconstruct the unknown electrical impedance distribution within the tissue. The rudiments of the technique are discussed and, using some simplifying assumptions, a rough estimate is made of the magnitude of the current modulation, and of the timescale necessary to obtain useful diagnostic information.

A real time volumetric ultrasound imaging system has been developed for medical diagnosis. The scanner produces images analogous to an optical camera or the human eye and supplies more information than conventional sonograms. Potential medical applications include improved anatomic visualization, tumour localization and better assessment of cardiac function. The system uses pulse-echo phased array principles to steer a two-dimensional array transducer of 289 elements in a pyramidal scan format. Parallel processing in the receive mode produces 4992 scan lines at a rate of approximately 8 frames/second. Echo data for the scanned volume is presented as projection images with depth perspective, stereoscopic pairs, multiple tomographic images, or C scans.

This paper presents a system model and its corresponding reconstruction method for synthetic aperture echo imaging using a single element ultrasonic transducer (SEUT). The system model incorporates the spherical nature of the SEUT's radiation pattern. The reconstruction method integrates the recorded signals at various coordinates of a translational or rotational SEUT via a spatial Fourier transform. It is shown that the transformed data provides samples of the spatial Fourier transform of the test object's reflectivity function. This study indicates that a synthesized array and its physical counterpart possess the same resolution despite the fact that the synthesized array's signal subspace is a subset of the much larger signal subspace for the physical array.

This paper addresses variances of estimates of power spectral densities of radio-frequency (RF) signals generated with ultrasound and magnetic resonance spectroscopy. The spectral estimation methods studied involved autoregressive (AR) and moving average (MA) models. With experimental ultrasonic data, the power spectral density estimate obtained using the MA model exhibited an appreciable reduction in vanance compared with the squared modulus of the FFT. With magnetic resonance spectroscopic data, the AR spectral estimate was comparable to, but not significantly better than, the squared modulus of the FFT.

Spatial inhomogeneities in the acoustic velocity of tissues degrade the contrast and resolution of diagnostic ultrasonic images. A system which iteratively adjusts the phases of signals transmitted and received from individual elements in a phased array scanner in order to restore image quality when imaging through inhomogeneous layers is described. The adjustments are selected to maximize a quality factor in an operator-selectable region-of-interest. The use of several quality factors in improving image quality, including the average local and the peak echo magnitude of diffuse and point-like targets and integral powers of these magnitudes were investigated. Theory and experimental results related to these investigations is presented.

We are designing an instrument which will perform correlated emission-transmission image acquisition, but which departs from previous systems by incorporating a low-power x-ray tube and generator, rather than a radionuclide source, for the transmission image. The system uses an array of high-purity germanium (HPGe) detectors and detector electronics with energy discrimination circuitry to separate x-rays (at 100 or 120 kVp) from higher energy gamma rays from the 99mTc or 123j radiopharmaceutical injected into the patient. The data acquisition electronics have time constants matching the charge collection time (50 ns) of the HPGe detectors to maximize count-rate capabilities (up to 1 million cps per detector element), while maintaining adequate energy resolution (approximately 10% FWHM). Each detector channel has two energy windows for simultaneous transmission-emission imaging or for dual-energy x-ray studies. A host computer provides system control as well as data acquisition, data correction, tomographic image reconstruction, image display, and data analysis. As a radionuclide imaging system, this instrument will function as a single-slice SPECT scanner with high-count rate capabilities and excellent energy resolution for imaging short-lived radionuclides, improved photopeak discrimination and scatter rejection, and simultaneous imaging of multiple radionuclides. The system also will generate radiographic images in either a tomographic or projection scanning mode, while dual-energy x-ray CT will provide material specific imaging. However, the novel and potentially powerful capabilities of this instrument would derive from its inherent correlation of functional information from SPECT with precise anatomic information from CT or the material-specific morphologic information from dual-energy x-ray CT. The simultaneously acquired radiographic images should relieve the deficiencies of poor statistics and limited spatial resolution commonly associated with SPECT systems. Dual-energy xray CT also can provide an energy-corrected and anatomically-correlated map of attenuation coefficients for more accurate quantitation of emission radionuclide data.

While preparing to conduct human facial surgery, it is necessary to visualise the effects of proposed surgery on the patient's appearance. This visualisation is of great benefit to both surgeon and patient, and has traditionally been achieved by the manual manipulation of photographs. Technological developments in the areas of computer-aided design and optical sensing now make it possible to construct a computer-based imaging system which can simulate the effects of facial surgery on patients. A collaborative project with the aim of constructing a prototype facial imaging system is under way between the National Engineering Laboratory and St George's Hospital. The proposed system will acquire, display and manipulate 3-dimensional facial images of patients requiring facial surgery. The feasibility of using two NEL developed optical measurement methods for 3-D facial data acquisition had been established by their successful application to the measurement of dummy heads. The two optical measurement systems, the NEL Auto-MATE moire fringe contouring system and the NEL STRIPE laser scanning triangulation system, were further developed to adapt them for use in facial imaging and additional tests carried out in which emphasis was placed on the use of live human subjects. The knowledge gained in the execution of the tests enabled the selection of the most suitable of the two methods studied for facial data acquisition. A full description of the methods and equipment used in the study will be given. Additionally, work on the effects of the quality and quantity of measurement data on the facial image will be described. Finally, the question of how best to provide display and manipulation of the facial images will be addressed.

A clinical instrument called a "Nevoscope" is used to image skin-lesions. The lesion is transilluminated by a fiber-optic annular ring light source that directs light into the skin area surrounding the lesion and thus forming a virtual source just beneath the lesion. Mirrors uniformly spaced around the lesion and tilted at various angles, provide orthographic projections of the skin lesion. Additional views are obtained by rotating the mirror assembly. These multiple views are used in a direct 3-D reconstruction of the lesion to estimate its depth of penetration. A pigment pattern analysis is performed on the direct view. This includes both color and texture segmentation. In this paper, we present preliminary results of our 3-D reconstruction and pigment pattern analyses of some lesions.

The application of confoca]. imaging systems to image the cornea and the ocular lens of the living eye is a major improvement in ophthalmic imaging. This new application of confocal imaging to the living eye has potential for both basic research and clinical diagnostics. The standard optical imaging systems have poor resolution and low contrast when applied to ocular tissue. The use of confocal systems to image the nearly transparent cornea and ocular lens has several advantages as an imaging system. The two important physical improvements are: (1) almost total rejection of out- offocal plane reflected light, and (2) increased lateral resolution. The problems of ocular imaging with standard slit lamps and specular microscopes include: low contrast images due to contributions to the image from out of focal plane reflections, and low resolution. Confocal imaging systems can image living eyes. The images show sharp, high resolution, high contrast features of submicron structures. Examples of ocular confocal images include the following: submicron optical sections through the various cells in the cornea, nerve fibers, cellular processes, and images through the nucleus of living cells comprising the cornea and the ocular lens. The ocular lens has been imaged. The lens capsule, lens epithelial cells, lens fibers and nuclei are readily observed. The use of realtime confocal imaging systems in ocular medical imaging has resulted in improved quality (resolution and contrast) of cellular images of living eyes. Further developments require the availability of long free working distance objectives with a high numerical aperture. This paper demonstrates the advantages of imaging ocular tissue with a confocal imaging system. Confocal imaging of ophthalmic structures may become a standard clinical tool with the new developments in instrumentation.

The problem of getting holograms of microscopic structures is considered, so as to apply the three-dimensional properties of holograms to biological specimens. Some optical holographic techniques are studied, pointing out problems concerning vision far from the microscope axis. A digital reconstruction method is proposed, with the aim of improving the lateral vision. Some images of optical microholograms are reported and the image formation theory is discussed in order to achieve a digital reconstruction.

There has been much evidence to suggest that diseased tissue is spectrally different from healthy tissue. Clinical endoscopy often results in fairly accurate diagnosis based on observed spectra (color), texture, and shapes of lesions. Conventional spectrometers lack the ability to coordinate adequate sampling of diseased verses healthy tissue to make a practical analytical tool. A video pixel spectrometer was developed which captures the spectra of a complete line of pixels within an image simultaneously. A color image of the suspicious tissue is continuously displayed which identifies a line through the image. This line represents the line of pixels which pass in to the spectrometer and is usually positioned to pass through the lesions. The spectrometer output spectrally disperses every pixel along the line simultaneously, resulting in the capture of the spectra of hundreds of pixels within the lesion and the surrounding tissue. The data can then be processed and correlated with shape and texture data for statistical classification of benign verses healthy tissue. This instrument was designed to couple to a variety of medical imaging instruments such as endoscopes, fundus cameras, macroscopic optics for dermatology, and microscopes.

A new type of endoscope is being developed which utilizes an optical raster scanning system for imaging through an endoscope. The optical raster scanner utilizes a high speed, multifaceted, rotating polygon mirror system for horizontal deflection, and a slower speed galvanometer driven mirror as the vertical deflection system. When used in combination, the optical raster scanner traces out a raster similar to an electron beam raster used in television systems. This flying spot of light can then be detected by various types of photosensitive detectors to generate a video image of the surface or scene being illuminated by the scanning beam. The optical raster scanner has been coupled to an endoscope. The raster is projected down the endoscope, thereby illuminating the object to be imaged at the distal end of the endoscope. Elemental photodetectors are placed at the distal or proximal end of the endoscope to detect the reflected illumination from the flying spot of light. This time sequenced signal is captured by an image processor for display and processing. This technique offers the possibility for very small diameter endoscopes since illumination channel requirements are eliminated. Using various lasers, very specific spectral selectivity can be achieved to optimum contrast of specific lesions of interest. Using several laser lines, or a white light source, with detectors of specific spectral response, multiple spectrally selected images can be acquired simultaneously. The potential for co-linear therapy delivery while imaging is also possible.

In fast MR imaging with a steady-state free precession (SSFP) pulse sequence the coherent signal that forms just before each rf pulse depends entirely on echoes from previous pulses. When moderate strength field gradient pulses are applied, this signal is very sensitive to random motions such as diffusion. A theoretical analysis of the full effects of diffusion in SSFP imaging is in good agreement with preliminary experimental results in a water phantom, but a previously used approximation does not adequately account for the qualitative or quantitative features of these effects.

An analytic solution of the eddy currents in nuclear magnetic resonance imaging is derived from Maxwell's equations and a temporal compensation technique for the recovery of distorted gradient waveform is described. The analysis is mainly focused on the frequency characteristics and the intensity variations of the eddy-current induced field depending on the system geometry. The major effects of eddy currents in NMR imaging appeared as (i) resolution degradation, (ii) misregisteration, (iii) intensity attenuation, and (iv) phase variation. Since the proposed approach is based on the analytic solution rather than experimentally determined multiexponential models as previously tried, accurate compensation can be achieved effectively. The limitation of the the temporal compensation due to the spatially-varying characteristics of eddy currents is also investigated.

The signal attenuation in nuclear magnetic resonance images where turbulent flow exists is described in terms of eddy diffusivity. An analytical expression for eddy diffusivity is given. Experimental results in phantoms are presented.

Head and body size images are discussed which were made by collecting four or eight data space lines per TR period. The images have a standard clinical number of pixels (128 x 256) and were made with no data averaging. The lines of data in the data space were collected along diagonal Cartesian echo planar trajectories. In this type of scanning, the main source of error is known to be the inhomogeneity of the main field. Phase angles caused by inhomogeneity can largely be eliminated and scanning speed increased by using the half Fourier method with full Fourier phase maps, but one phase map is required for each echo. To prevent having a speed increase only in scans taking longer than 10 seconds, some attention to GRASS and/or FLASH technology is also required. In addition, the method can be applied to directly to FLASH/GRASS as is demonstrated by a two echo GRASS scan. Fastest image time was 1.9 seconds.

We report on the behavior ofthe linear maximum a posteriori (MAP) tomographic reconstruction technique as a function of the assumed rms noise α_n in the measurements, which specifies the degree of confidence in the measurement data. The unconstrained MAP reconstructions are evaluated on the basis of the performance of two related tasks; object detection and amplitude estimation. It is found that the detectability of medium-sized discs remains constant up to relatively large α_n before slowly diminishing. However, the amplitudes of the discs estimated from the MAP reconstructions increasingly deviate from their actual values as α_n increases.

A method for comparing reconstruction algorithms is presented based on the ability to perform certain detection tasks on the resulting images. The reconstruction algorithms compared are the algebraic reconstruction technique (ART) and the maximum entropy reconstruction method (MaxEnt). Task performance is assessed through a Monte Carlo simulation of the complete imaging process, including the generation of a set of object scenes, followed by data-taking, reconstruction, and performance of the specified task by a machine observer. For these detection tasks the figure of merit used for comparison is the detectability index, d'. When each algorithm is run with approximately optimized parameters, these studies find comparable values for d'.

Unlike the attenuation densities which are estimated in conventional xray tomography the signals estimated and measured in magnetic resonance (MR) imaging are inherently timevarying. Any method for reconstructing the free induction decay (FID) signals within particular voxels of an object must take this into account. In spite of this, Fourier MR imaging methods involve decoding procedures for spatial localization of spin density and exponential decay parameters which assume that the signal amplitudes are constant in time, an assumption which is valid only under the condition that the data collection window is very small with respect to the signal decay rate. This forces conventional Fourier methods to generate images from small slices of the full FID time signal (on the order of lO msec). We have recently demonstrated in NMR spectroscopy [1] that fitting the FIB to exponentially decaying sinuosoid models with unknown amplitude,frequency and decay parameters using the method of maximumlikelihood yields far more accurate estimates of the parameters than those based on Fourier methods. This ofcourse requires collection and fitting of the entire FID time signal. Following these ideas, we now describe a new method for reconstructing spin density and T2 images from data collected in the hydrogen MR imaging mode, which models the FID from every voxel of the image as a sinusoid with unknown amplitude and decay. The paper first presents a signal model for the FID signals collected from hydrogen MR imaging based on previous work with collaborators [2], which is at the heart of our new algorithm for maximum likelihood estimation of the image parameters. Then the major focus of this paper is to describe the maximum likelihood estimation of the spin density and T2 decay images, and present its solution via an iterative expectation maximization algorithm. Finally we show the reconstruction of a simulated 2 dimensional phantom imaged using conventional phase and frequency encoding. We conclude by showing results which demonstrate the major advantage of the ML method over conventional Fourier based techniques for producing MR images.

The problem of image assessment is examined for several cases of parameter uncertainty. Several ideal and sub-ideal observers are considered and figures of merit (FOM) for describing their performance are considered. Advantages and disadvantages of these FOMs are enumerated. The spectrum of noise equivalent quanta, NEQ(f), appears to be the most useful for evaluating a broad class of practical problems since the performance of the best-linear as well as the ideal non-linear observers considered here is monotonic with its components. However, more work is required within this context to quantify the effects of the "null space," or regions in object space for which NEQ = 0. These regions derive from incomplete measurement sets and may lead to severe image degrading artifacts that are not adequately covered by any FOMs considered here.

In this paper, a continuous / discrete projection / backprojection model is presented, from which the validity of the discrete projection / reconstruction algorithm can be assessed. We mainly focus on projection sampling since angular sampling has been extensively studied previously. For this purpose a pixel intensity distribution model relating continuous and discrete original functions is proposed. Sampling of model projections is then studied, and projection filtering analyzed. Proper implementation of the discrete backprojection operator is derived, such that the resulting reconstructed function can be compared with the original one, and the overall consistency of the aproach proved. Experimental results are presented to demonstrate the validity of the theoritical approach. The consequences of properly sampling projections in practical conditions are fmally discussed.

We address the problem of applying Dual-Energy techniques to Image-Intensifierbased acquisition systems. We propose a new Dual Energy calibration and combination scheme, which takes into account the spatially non-uniform response of the Image Intensifier. During the calibration step, the coefficients of the combination are locally estimated for a set of uniformly spaced points over the image, and interpolated from these samples for the other points. In the combination step, the selective images are generated using the corresponding polynomia for each pixel. 3D selective reconstruction of bone-only structures is presented as a possible application. Experimental results obtained on anthropomorphic phantoms are discussed.

We address the problem of reconstructing a three-dimensional volume from a set of two-dimensional X-ray projections. We present a time efficient solution based on a multiscale estimation technique. Estimation is first performed at a coarse resolution. Then the resolution is increased step by step and at each step a new estimation is performed, using an initial value derived from the volume estimated at the preceding level of resolution. The method is illustrated by results obtained on geometric and anatomic phantoms.

We have constructed a table-top CT system with high spatial resolution in all three dimensions which can be used to analyze excised tissue samples in vitro. Our system uses an x-ray image intensifier, optically coupled to a time delay integration (ThI) CCD to obtain low-noise, low-scatter projection radiographs of the sample volume. Rather than collecting one projection line at a time (as in conventional CT scanners), the architecture of the TDI-CCD allows us to collect 96 image lines simultaneously. The digital radiograph is formed by scanning a slot-beam of radiation across the sample, reducing the detection of scattered radiation without excessive x-ray tube heat loading. Objects to be imaged are placed on a computer-controlled stage and projections are obtained as the sample is rotated. A water bath surrounds the sample to equalize the exposure to the image intensifier, thereby reducing the dynamic range of the input signal. CT reconstruction of this data results in a 512 volume image with 0.13 x 0.13 mm pixels in the transverse plane and a slice thickness of 0.15 mm. This table-top CTscanner has been used to investigate the properties of intact, excised human arterial samples as part of our research into the development of vascular disease.

A phantom. having similar dimensions to a heart in the human thorax, has been used to determine the sensitivity of X-ray CT scanning for the imaging of contrast agents within the myocardiuin. A theoretical determination of sensitivity has been made for elements from Z=35 to Z=92 for both monochromatic and polychromatic beams. For the monochromatic beams, calculations were made for single energy studies in which the contrast agent was added after an initial scan and for dual energy studies where differential absorption across a K or L absorption edge is used. Simulated images, in which Poisson noise was added to the projection data, have been generated for various elements in the range Z=35 to 92, with particular attention being paid to iodine, gadolithum, tungsten and gold for monochromatic and polychromatic beams. For monochromatic beams the simulated images are compared with the theoretical predictions. The phantom, imaged in a Somatom-2 scanner (with a 256x256 display) has been used to test the accuracy of the results for a polychromatic beam. Results show that the maximum sensitiyity will occur in the region of Z=64 (gadolinium).

Two years ago in these proceedings1 we reported on a new method for measuring the noise associated with the variation in light output of x-ray intensifying screens caused by absorption of x-ray quanta of equal energy, together with data for the Kodak Lanex intensifying screens. We have extended our measurements to screens in the back-screen configuration, in addition to the front-screen configuration previously reported. By back-screen configuration we mean that the x-rays are incident from the same side as that from which the emission will be measured. This has been realized by means of an integrating sphere, which allows screens to be mounted as back or front screens; or even as a pair. The light emission statistics (including the mean light output and the Swank I factor) for some Kodak Lanex intensifying screens in the front and back-screen configurations are given and compared. These data can provide a basis for understanding the depth dependent emission probability which in turn provides a useful test of theories of light propagation within the screen.

In our previous studies,"2 we had shown the utility of multivariate momentgenerating functions (MMGF) for analyzing the influence of stochastic amplifying and scattering mechanisms on the transfer of signal and noise through multistage imaging systems. Recently, we extended these studies to include cases in which the amplification or scattering parameters are themselves stochastic variables.3 In this paper we consider a special case in which amplification is followed by scattering such that the same random variable which characterizes the parameters of each amplification process also characterizes the parameters of the subsequent scattering of the amplified output events. In radiographic imaging, this can be used to describe the physics of the depth dependence of emission efficiency and light scatter in x-ray intensifying screens which was originally treated by Lubberts.4 In this work Lubberts' original results are rederived in a more general form.

We developed a new technique for the laser film digitizer using a photo diode and a special optical system with an elliptical mirror to replace a photo multiplier tube and optical fiber bundle combination. In order to maintain the highest possible signal-to-noise ratio with a photo diode, the noise problem due to the dark current had to be solved. The authors analyzed the characteristics of the dark current first, then determined the optimized laser beam power at the surface of the photo diode so that the highest signal-to-noise ratio can be kept. In order to realize this result, a new elliptical mirror system which boosted transmission efficiency from 5% to 70% was developed. In addition, the authors discuss the required features for the future laser film digitizers for radiological application.

In this paper, we discuss the requirenients, design approach, and features for a high-performance, laser-based radiographic film digitizer. The scanner is capable of digitizing a radiograph over a 3.5 diffuse density range in 15-30 seconds while not significantly reducing the noise-equivalent-quanta (NEQ) relative to the original analog screen-film image. Three resolutions are possible, corresponding to maximum film sizes of 35 x 43, 24 x 30, and 18 x 24 cm, with pixel sizes of 86, 59, and 43 microns; respectively. Preliminary results of side-by-side comparisons of scanned data reconstructed on pnnted film with the original analog film are discussed, along with future directions for research.

There is continuous effort by the medical system manufacturer and component manufacturer to improve the quality of the X-ray imaging system. As part of this effort, Philips has developed a new Plumbicon® camera tube for medical X-ray imaging applications. The new Plumbicon® camera tube type 88XQ provides improved spatial resolution, improved S/N and contrast resolution and many other functional advantages.

A scanned-slot digital radiography system is being developed for use in mammography. The system consists of a novel fiber-optic "reformatter" which couples a strip of phosphor material to a light amplifier and a specially designed CCD camera. The reformatter provides an excellent means of converting a slot-shaped image to a format more suitable for digitization. A mammogram is produced by scanning the system in steps across the breast from the chest wall to the nipple. The pixels are of dimension 56 μm by 49 μm at the detector input. Images are currently acquired using a 40 kV tungsten anode spectrum with a measured half-value layer of 0.82 mm Al. The limiting resolution [MTF(f)=O.05] in the scanning direction is 9.2 cycles/mm. The quantum interaction efficiency of the phosphor is 64%, and the low-frequency detective quantum efficiency (DQE) was measured to be 0.60 ± 0.07. The high-frequency DQE is superior to that of mammographic film-screen systems because of the ability to remove fixed pattern noise from images. Phantom images produced with the scanned-slot system, and with a state-of-the-art film screen mammography system were compared. The scanned-slot system demonstrated better contrast sensitivity using a lower mean glandular dose than the film-screen mammography system.

A fast, unique method for determining phosphor scintillation spectra by means of an electronic apparatus is described. A major advantage is that all of the controls are digital. This includes coincidence timing to determine whether an x-ray absorption event has occurred as well as control of the time window during which the number of light photons emitted in response to an x-ray absorption event is counted, and storage of the photon count itself. The apparatus and its logic are described in detail. The scintillation spectra and the resulting calculated Swank coefficient values of several phosphors obtained with this apparatus are presented as examples.

The Expectation Maximisation (EM) algorithm for computing the Maximum Likelihood image has been applied to both simulated and experimental cone-beam SPECT projection data. Reconstructions of simulated data show the superiority of the EM algorithm over a convolution and backprojection (CBP) algorithm with respect to image uniformity across the field of view and tolerance to noise. Phantom studies demonstrate the need for an accurate attenuation correction routine. A method of correcting for photon attenuation has been applied to the EM algorithm assuming the object to be a uniform attenuator. A method of extending this to allow for non uniform photon attenuation is also proposed.

A newly developed x-ray detector for use in a direct digitizer using an alkali halide photostimulable phosphor is described. For efficient protection against moisture, the design makes practical a protective layer over 1mm thick with virtually no loss of image sharpness.

We have made preliminary image quality measurements (modulation transfer function and noise power spectrum) measurements on a digital scanned projection radiography system developed in our laboratory. The DSPR system images the light output of a scanning strip of intensifying screen onto a time-delay-Wand--integrate scanning CCD imager with a high speed lens. The data show that the current design is limited in spatial resolution compared to a screen film system. In the current CCD camera electronic noise dominates both Wiener spectrum and MTF near the resolution limit. Noise power spectrum measurements indicate that the current DSPR system, with a relatively large amount of electronic noise, requires exposures an order of magnitude higher to achieve DQE comparable to screen/film systems. However, for clinical situations in which scattered radiation is a significant factor an advanced prototype DSPR with improved analog electronics is predicted to have DQE comparable to screen/film/grid systems at comparable doses. Development of a low noise DSPR system is currently underway.

We are investigating a prototype x-ray imaging system in which a scintillating fiberoptic glass plate and/or a fluorescent screen is fiberoptically coupled to a 2048 x 2048 CCD array (Tektronix). The imaging system includes a fiberoptic minifier to increase the imaging field of view to a clinically usable one. The system also allows for cooling of the CCD to reduce the effect of dark noise on image quality and the use of single-stage light amplification to act as a shutter and to provide gain control. Images are software corrected for dark current, individual pixel gain, and geometric distortion. Preliminary results indicate that high quality x-ray imaging can be obtained using this methodology. This paper describes design concepts and configuration of the system as well as characterizations of the initial x-ray images acquired with the camera.

The limited light sensitivity of the TV-pick-up devices and the desire for low dose imaging in X-ray fluoroscopy causes the use of extremly high aperture lens optics between X-ray image intensifier and TV-sensor (1). Therefore a contrast reducing optical feedback between the TV-sensor and the output screen of the X-ray image intensifier may occur, if the light of the output screen of the X-ray image intensifier is reflected from the sensor's photoconductor. The contrast relations in the case of lens coupling of X-ray IL of different output structures with some CCD-sensors and different kinds of pick-up tubes have been measured and will be discussed. One possible way to reduce contrast loss due to stray light is direct fiber optical coupling. We investigated such systems with fiber optic pick-up tubes and a CCD equipped with a fiber optic plate in front of the sensor area coupled via a fiber optical taper to an X-ray image intensifier with fiber optic output window. The drawback of every direct fiber optical coupling is, however, that the light flux to the TV-sensor cannot be controlled. Therefore an adaptation of such a system to the practical needs of varying indMdual applications cannot be performed. Some imaging properties oflens coupling, fiber optical coupling and mixed coupling (fiber optics and lens optics in one system) have been evaluated and will be compared. Some precautions against loss in contrast and requirements regarding signal to noise ratio will be outlined.

A prototype digital chest system, which uses storage phosphor technology and has the advantages over existing computed radiography systems (CR) of compactness and immediate image display, is being evaluated in our laboratory. We evaluated the imaging properties of the Konica Direct Digitizer (KDD) in order to assess its potential usefulness for general clinical use, or as a front-end for a PACS. The prototype system consists of a new stimulable phosphor (RbBr.Tl) detector read by a compact semiconductor laser scanning system, with images immediately displayed on a CRT or transferred to a host computer. The imaging characteristics of resolution and noise were evaluated, using display parameters matched to a Kodak Lanex Medium/OC system. Preliminary results using sensitive composite test objects show an increase in noise and a slight decrease in resolution as compared to conventional radiography. However, subjective comparison of a chest phantom and volunteer images indicates that these differences may not be clinically significant. Further development is needed to provide increased absorption, and thus improved image quality.

Two years ago we first reported the preliminary results from an experimental methodology that we believed would be sensitive enough to enable the measurement of very subtle changes in observer detection performance [1]. If successful, we intended to apply this methodology to the evaluation of computer enhancement techniques. The procedure had two facets: generation of digital images using a 3M chest phantom with simulated lung and heart structures, and the presentation of reference information to the observers during the psychophysical experimental session. We used our Toshiba computed radiography (TCR) 201A system to generate the digital images. The digital images were a benchmark from which we could physically measure changes due to image processing. Eventually we hoped to be able correlate physical measurements taken from the digital image to psychophysical detection measurements to get a better understanding of the effect of processing on perception. We believe that this correlatiye information will eventually help us to find or develop enhancement techniques that measurably improve detection. Once we knew that a teOhnique increased detection, then we would apply it to clinical images and evaluate the technique's increases in diagnostic accuracy.

A digital system has been developed for on-line acquisition, processing and display of portal images during radiation therapy treatment. A metal/phosphor screen combination is the primary detector, where the conversion from high-energy photons to visible light takes place. A mirror angled at 45 degrees reflects the primary image to a low-light-level camera, which is removed from the direct radiation beam. The image registered by the camera is digitized, processed and displayed on a CRT monitor. Advanced digital techniques for processing of on-line images have been developed and implemented to enhance image contrast and suppress the noise. Some elements of automated radiotherapy treatment verification have been introduced.

There is increased concern over radiation exposure to the general population from many sources. One of the most significant sources is that received by the patient during medical diagnostic procedures, and of these, the procedure with the greatest potential hazard is fluoroscopy. The legal limit for fluoroscopy in most jurisdictions is SR per minute skin exposure rate. Fluoroscopes are often operated in excess of this figure, and in the case of interventional procedures, fluorocopy times may exceed 20 minutes. With improvements in medical technology these procedures are being performed more often, and also are being carried out on younger age groups. Radiation exposure during fluoroscopy, both to patient and operator, is therefore becoming a matter of increasing concern to regulating authorities, and it is incumbent on us to develop digital technology to minimise the radiation hazard in these procedures. This paper explores the technical options available for radiation exposure reduction, including pulsed fluoroscopy, digital noise reduction, or simple reduction in exposure rate to the x-ray image intensifier. We also discuss educational aspects of fluoroscopy which radiologists should be aware of which can be more important than the technological solutions. A "work in progress" report gives a completely new approach to the implementation of a large number of possible digital algorithms, for the investigation of clinical efficacy.

A new method to measure the texture (structure) noise produced by x-ray screens has been developed. Different films are exposed with the same screen and are scanned in registration. The numerical correlation of the two sets of pixel data then contains only the texture noise component. This is a powerful technique allowing texture noise to be measured quantitatively in the presence of the quantum noise and film grain noise background. Our results support the generally accepted belief that texture noise is small compared with quantum and film grain noise. There are advantages to visualizing texture noise behavior in Cartesian space (as correlation functions) as well as in frequency space (as noise power spectra). Some examples of structure noise correlations and the corresponding spectra are presented.

We describe a method to analyze the linear imaging characteristics of rotationally invariant, radially variant tomographic imaging systems using singular value decomposition (SVD). When the projection measurements from such a system are assumed to be samples from independent and identically distributed multi-normal random variables, the best estimate of the emission intensity is given by the unweighted least squares estimator. The noise amplification of this estimator is inversely proportional to the singular values of the normal matrix used to model projection and backprojection. After choosing an acceptable noise amplification, the new method can determine the number of parameters arid hence the number of pixels that should be estimated from data acquired from an existing system with a fixed number of angles and projection bins. Conversely, for the design of a new system, the number of angles and projection bins necessary for a given number of pixels and noise amplification can be determined. In general, computing the SVD of the projection normal matrix has cubic computational complexity. However, the projection normal matrix for this class of rotationally invariant, radially variant systems has a block circulant form. A fast parallel algorithm to compute the SVD of this block circulant matrix makes the singular value analysis practical by asymptotically reducing the computation complexity of the method by a multiplicative factor equal to the number of angles squared.

A key problem in using digital subtraction radiography in dentistry is the ability to reposition the X-ray source and patient so as to reproduce an identical imaging geometry. In this paper we describe an approach to solving this problem based on real time sensing of the 3-D position and orientation of the patient's mouth. The research described here is part of a program which has a long term goal to develop an automated digital subtraction radiography system. This will allow the patient and X-ray source to be accurately repositioned without the mechanical fixtures that are presently used to preserve the imaging geometry. If we can measure the position and orientation of the mouth, then the desired position of the source can be computed as the product of the transformation matrices describing the desired imaging geometry and the position vector of the targeted tooth. Position and orientation of the mouth is measured by a real time sensing device using low-frequency magnetic field technology. We first present the problem of repositioning the patient and source and then outline our analytic solution. Then we describe an experimental setup to measure the accuracy, reproducibility and resolution of the sensor and present results of preliminary experiments.

A typical radiological TV imaging system includes an X-ray generator, an X-ray image intensifier tube which produces a visible image, an optical lens, a pick-up tube working in a TV camera, and finally an image processing device followed by an image display. The PRIMICON pick-up tubes are already well tried in various radiological systems (general fluoroscopy, digital fluoroscopy, digital subtraction angiography (DSA)) In which they provide their image performance features: high resolution, and excellent contrast and large dynamic range. Another important image feature is the time response of the pick-up tube, named lag. In general fluoroscopic mode, high lag is needed to integrate the noise caused by the low X-ray dose. On the contrary, in digital fluoroscopic mode or DSA mode, low lag is required to pick-up images with moving contents. Consequently, as standard pick-up tubes have a fixed amount of lag, two TV cameras must be used if the system has to operate in the three modes described above. Due to a special design of its photoconductive layer, the new PRIMICON tube features variable lag, the amount of lag being controlled by the combination of the signal current level and of the light bias rating. The necessary light source is located close to the optical path in front of the tube's faceplate. So this new tube (referenced Th 9957) is suited either to general fluoroscopic mode (12 % lag is got with 100 nA signal current and no light bias) or the digital fluoroscopic or DSA modes (2% lag is got with 1500 nA signal current and 100 nA light bias). As a conclusion, the new PRIMICON pick-up tube enables multirnode TV cameras to be realized, which bring substantial benefits in radiological systems - system design is simpler, - system use is easier, - system maintenance is alleviated and fewer spare components are needed.

The human chest is a demanding object for conventional radiography. There is a large difference in X-ray lransmittance between the lung and the mediastinal areas. In order to create an image which can be judged in a single view one has resorted to high kVp exposures and the use of wide latitude receptors. In a study on an out-patient population the X-ray transmittance of the chest was measured. From these data a sample response for the AMBER imaging system is derived.

Autoradiography in biochemical analysis supplies important information, but often requires long exposure times, often measured in weeks, to acquire adequate images. A commercial computed radiography system was utilized to investigate the potential application ofphotostimulable phosphor technology for direct autoradiographic imaging ofsamples labelled with beta-emitting isotopes. It was determined that the wide dynamic range of the photostimulable phosphor imaging plates could provide more information than conventional film autoradiographic techniques, however the plate reading system could not consistently utilize the full extent of the plate's acquired information.

We have developed a mammography phantom which is designed to take advantage of the recently developed Free-Response Receiver Operating Characteristic (FROC) methodology. This observer performance methodology allows for multiple abnormalities and observer responses per image and requires both detection and correct localization for a true positive event. The phantom is composed of several Lucite slabs and a series of 30 test plates to which were affixed simulated microcalcifications, fibrils and masses. The phantom also simulated a structured background. The series was radiographed using three different techniques (30 kVp with no grid, 28 kVp and 35 kVp). These three sets of 30 images were viewed by three medical physicists and one radiologist in the FROC manner. The analysis provided an index of performance for each reader and each set. As multiple abnormalities may appear on a single image, good statistics are achieved using relatively few images. Statistical comparison of the index values indicates that, with improvements in the phantom, the combination of the phantom with FROC methodology should be capable of detecting subtle changes in image quality and may be used on a regular basis as part of a detailed quality assurance program.

A prototype imaging system for digital x-ray imaging of the chest is under development in our research laboratory. The image receptor consists of a 40/30/22-cm image intensifier and an electronic camera employing a 1320 x 1035-pixel scientific grade CCD. The veiling glare-to-primary ratio and spatial resolution have been measured.

Ascatter correction technique based on artificial neural networks is presented. The technique utilizes the acquisition of a conventional digital radiographic image, coupled with the acquisition of a multiple pencil beam ("micro-aperture") digital image. Image subtraction results in a sparsely sampled estimate of the scatter component in the image. The neural network is trained to develop a causal relationship between image data on the low-pass filtered open field image and the sparsely sampled scatter image, and then the trained network is used to correct the entire image (pixel by pixel) in a manner which is operationally similar to but potentially more powerful than convolution. The technique is described and is illustrated using clinical "primary" component images combined with scatter component images that are realistically simulated using the results from previously reported Monte Carlo investigations. The results indicate that an accurate scatter correction can be realized using this technique.

We are investigating the characteristics of a prototype digital radiography imaging system in which six two-dimensional diode arrays (CCD) are directly coupled through a bonded matrix (3 x 2) of fiberoptic minifiers to either a scintillating fiberoptic glass plate or to a fluorescent screen. Images are digitally acquired at a rate of up to 30 frames/sec and software corrected for pixel gain, dark current, and geometric distortion. This paper describes the concepts and design configuration of this approach, as well as preliminary results from several phantom and animal studies. Our results indicate that high resolution (> 4 lp/rnm) and high signal-to-noise ratio images can be obtained with this method. However, the complexity associated with this concept cannot be discounted.

This paper describes a new method to determine the lower limit of patient exposure: By placing several imaging plates of a computed radiography system (CR) into the same cassette, several images of the same patient can be obtained at different exposure levels (determined by the x-ray transmission of the various imaging plates. Initial experiments indicate that exposure reduction of between 50 and 75% might be acceptable. CR provides a powerful tool to study the subject of exposure reduction.

This paper compares image quality of conventional film based portal images with the quality of portal images based on computed radiography (CR). The paper concludes that portal imaging with CR imaging plates instead of the conventional film provides images which have a significantly higher image quality. CR based portal images permit better visualization of important landmarks.

"Direct" measurement of the scatter point spread function (PSF) was obtained with a digital imaging system for the purpose of comparison of a Monte Carlo (MC) scatter distribution algorithm that takes into account polychromatic spectra and detector absorption characteristics. An image intensifier (II) TV system was used to experimentally acquire digital x-ray scatter images in a geometry designed to eliminate detection of the primary beam and to optimize the dynamic range for scattered radiation as well as correct for image variations caused by veiling glare and shading. Both computer generated and experimentally acquired scatter data were measured in concentric annular bins and individually integrated to provide a direct PSF profile. Results indicate a good match between the MC and experimental PSFs for homogeneous scatter distributions at short range airgaps and radial distances. Discrepancies at larger radial distances are likely due to the spherical II input phosphor geometry. For an ll/1'V system detector, therefore, geometric image warping correction for pincushion distortion is indicated, as well as modifications in the MC algorithm for geometry of the detector. This will help to minimize differences between simulated and measured system characteristics.

A system model for analyzing degradation in the image quality of a radiograph introduced by a film digitizer is presented. The analysis is an extension of the screen-film model of Shaw and VanMetter (SPIE 454, 128-141(1984)). By combining the screen-film characteristics for specific exam types with the properties (e.g., MTF and NPS) of a particular scanner design, the information transfer of the whole digital system can be determined. As an example, the performance of two typical film digitizers, a CCD-based scanner and a laser-based scanner, are evaluated and compared. Image quality descriptors, such as DQE and NEQ as well as equivalent bandwidth and system aperture, are used for the evaluation. By incorporating the human observer's threshold response to changes in noise levels (just noticeable differences), a criterion for negligible loss of image information can be established. This can be very useful for system optimization and determination of design tradeoffs.

We have begun experimentation with a variety of digital and optical filters used in conjunction with a digital imaging colposcope system of our design. We have successfully implemented a number of different filtering techniques and now see potential applications of multispectral imaging in colposcopic examination. We discuss some of the methodology relevant to acquisition of multispectral images. Analysis of the scope of application of this imaging to colposcopic diagnosis is an interesting problem requiring further study.

MTF computations based on a "knife edge" scan, which is a numerical representation of the so-called edge spread function (ESF), require computation of the derivative of this function, thus yielding the line spread function ([SF) whose Fourier transform (FT) is the MTF. Since direct numerical differentiation is often inaccurate, the differentiation is usually accomplished by convolution with the derivative of a filter function, typically the derivative of the sinc function, sin(nfχ)/nfχ. But, since the sinc function is itself the FT of a rectangular filter, the subsequent FT of the [SF obtained by this procedure behaves as if the LSF had been abruptly truncated (set equal to zero) at the cutoff frequency, f, so that the resulting LSF is often subject to rapid oscillations (ringing). A new filter function based on the FT of a more gradually decreasing Hann function has been derived in closed form. Convolutions with the derivative of the new function yield smooth, non-oscillating LSF's. Examples of theoretical and experimental LSF's obtained by this method are presented and the qualitative limits within which this procedure is applicable are discussed.

We have developed a table-top CT system with high temporal and spatial resolution which can be used to quantify the 3-D flow patterns and arterial wall distensibility in intact arterial samples. Our system comprises of an x-ray image intensifier optically coupled to a linear photo-diode array camera. The arterial phantom or sample is irradiated by a narrow beam and a projection image of a thin slice through the sample is obtained with a temporal resolution of 17 ms. The spatial resolution in the CT images is 2 mm. The system can be used in two modes: flow quantification and dynamic motion quantification. In flow quantification mode an image is formed of a bolus of radio-opaque dye as it flows past the plane of interest. Repeated bolus injections, performed as the arterial phantom is rotated through 180, yield a series of timeposition projection images of the flowing bolus. These projection images are reconstructed using CT techniques to produce a 3-D time-position map of the flowing bolus. In dynamic motion quantification mode, time-position maps are obtained from many orientations as fluid is pumped through an arterial sample. Pulsatile flow is provided by a computer controlled pump which generates reproducible physiological flow and pressure waveforms. The collection of each projection image is gated to the beginning of the cardiac cycle. Again, usIng CT techniques we obtain cross-sectional images of the arterial sample at 60 points of the 'cardiac' cycle.