Over the past decade or so the Annual SPIE Conference on Medical Imaging has provided the leading forum for the latest advances in imaging physics as they relate to the analysis, modeling and evaluation of medical x-ray systems. During this time there has been a near- revolution in the application of modern imaging techniques, to the extent that these now provide a universal language for the description, and inter-comparison of the increasingly diverse range of technologies which have been developed for the improved efficiency of medical diagnostic imaging. It is the intent of this review to provide a survey and reminder of the course of this progress, with illustrations within the general theme of conventional screen- film systems.

The signal and noise propagation through a generalized x-ray image intensifier/TV camera chain is modeled in terms of the modulation transfer function, two dimensional noise power spectrum, and detective quantum efficiency. The model covers effects of energy dependent x- ray absorption in the cesium iodide entrance screens, K-fluorescence and K-reabsorption, electron optics, output screens, lenses and also sensing with electronic pick-up tubes (plumbicons) or CCD sensors. Several Philips x-ray image intensifiers in combination with plumbicon TV-chains and with laboratory type CCD cameras are compared with the model. Based on the model results we further present an image simulation tool. The image simulation allows a direct evaluation of the impact of the individual components of the imaging chain on image quality. It is shown that future high resolution CCD systems can exhibit a superior image quality as compared with electronic pick-up tube systems.

Conventional computed tomography (CT) images are `maps' of the x ray linear attenuation coefficient within a slice through an object. A novel approach to CT is being developed which instead produces tomographic images based on an object's low-angle (0 - 10 degree(s)) x-ray diffraction properties. The coherent-scatter cross sections of many materials vary greatly, and this coherent-scatter CT (CSCT) system gives material-specific information on this basis. The goal of this research is to produce tomographic maps of bone-mineral content (BMC), first in laboratory specimens, and potentially in patients. The concept of reconstructing tomographic images using coherently scattered x rays was first demonstrated by Harding et al. The approach described here is a modification of their method. First generation CT geometry is used in which a diffraction pattern is acquired for each pencil-beam using a CsI image intensifier coupled to a CCD. Each pattern is sectioned into concentric annular rings so that the integrated signal in each ring gives the scatter intensity at a particular scatter angle, integrated along the path through the object. An image is reconstructed for each ring, resulting in a series of tomographic images corresponding to the scatter intensity at a series of scatter angles. A test phantom was imaged (70 kVp, 50 mAs per exposure, 100 mSv average dose) to demonstrate CSCT. The phantom consists of a water-filled Lucite cylinder containing rods of polyethylene, Lucite, polycarbonate, and nylon. The resulting series of images was used to extract the angular-dependent scatter cross section for every pixel. Using pure material cross sections as basis functions, the cross section from each pixel was fitted using non-negative least squares. The results were used to create material-specific images. These results show that CSCT is feasible with this approach and that if the materials in an object have distinguishable scatter cross sections, the method has the ability to identify the materials. It may be possible to image the BMC distribution as trabecular bone mineral is replaced by lipids.

This study compares the relative response of various screen-film and computed radiography (CR) systems to diagnostic radiation exposure. An analytic model was developed to calculate the total energy deposition within the depth of screen and the readout signal generated from this energy for the x-ray detection system. The model was used to predict the relative sensitivity of several screen-film and CR systems to scattered radiation as a function of selected parameters, such as x-ray spectra, phantom thickness, phosphor composition, screen thickness, screen configuration (single front screen, single back screen, screen pair), and readout conditions. Measurements of scatter degradation factor (SDF) for different screen systems were made by using the beam stop technique with water phantoms. Calculated results were found to be consistent with experimental observations, namely, both the BaFBr screen used in a CR system and the CaWO₄ screen pair have higher scatter sensitivity than the rare earth Gd₂O₂S screen pair; the BaFBr screen in the CR front-screen configuration is less sensitive to scatter radiation than in the normal back-screen configuration; and these screens have higher scatter sensitivity as x-ray tube voltage increases.

The point-spread function (PSF) of a circularly symmetric imaging system is commonly inferred from the line-spread function (LSF), which is the image of a line source whose length must be larger than the spatial extent of the PSF. This constraint on the minimum length of the line source makes it impossible to measure the LSF of a system whose PSF is large in extent relative to the size of the system's isoplanatic patch. This impasse motivates one to consider the problem of inferring the PSF from the finite-length line spread function (FLSF), which is the image of a finite-length line source of arbitrary, but fixed, length. Formulas for calculating the PSF from the FLSF have been developed, but the numerical implementation of these formulas are either time consuming or unstable. In this presentation, we derive a formula for performing the FLSF-PSF conversion which is better suited for numerical purposes.

The objective imaging characteristics of a wide range of gandolinium oxysulfide-based, zero- crossover, screen-film combinations are presented and compared. It is shown that complex high-spatial frequency versus low-spatial frequency performance tradeoffs are found among these systems, when these systems are examined in terms of sensitometric response, modulation transfer function, noise equivalent quanta, and detective quantum efficiency.

In a recent publication we demonstrated that the increase in image noise which results from exposing a film via a phosphor screen can be attributed entirely to the increased extent of the autocorrelation interval introduced by the screen, and not to any change in the shape or scale of the probability distribution function which governs the fluctuations about the mean level. This result implies that the (0,0)-value of the autocovariance function is independent of the degree of so-called quantum mottle and since the autocovariance function, ACV(x,y), and the Wiener Spectrum, WS(u,v), are Fourier transform pairs, it follows that the integral of the Wiener Spectrum over all spatial frequencies (u,v) must share this independence. Since this result was counterintuitive to existing assumptions of the role of screen and film in defining the Wiener Spectrum (i.e., a simple additive combination of quantum mottle and film grain), we now investigate this relationship in more detail in order to provide a more complete insight. For this purpose we have simulated a set of controlled images which correspond to a wide range of screen correlation intervals, from 192 micrometers down to uncorrelated film noise. Included in this simulation we have also explored the role of the overall amplification factor, i.e., the average number of image grains associated with an x-ray quantum. The results of these simulations are presented here, and confirm our previous findings, by establishing the invariance of the scale (0,0) value of the ACV.

The identification of women at increased risk for breast cancer has important implications in both the surveillance for cancer and research into causes of the disease. The parenchymal pattern of the breast, revealed by mammography, and rated subjectively by observer, has been found to provide strong factors of risk for breast cancer. To provide a more quantitative measure of the proportion of mammographically dense tissue in the breast, we have previously described and evaluated an interactive technique in which an observer selects a threshold brightness level to separate dense from fatty tissue in the image. Measurement of mammographic density in this way provides an estimate of relative risk of 4, that is among leading indicators of the risk of developing breast cancer. To remove the variability associated with identification of thresholds by observer, we are investigating an automated threshold prediction based on independent features characterizing mammographic parenchyma. These features are based on regional measurements of image brightness variations (histogram analysis) and texture variations (fractal analysis) within digitized mammographic images. Preliminary results from an investigative model are presented.

Improvements in mammography equipment related to a decrease in pixel size of digital mammography detectors raise questions of the possible effects of these new detectors. Mathematical modeling suggested that the benefits of moving from 100 to 50 micron detectors were slight and might not justify the cost of these new units. Experiments comparing screen film mammography, a storage phosphor 100 micron digital detector, a 50 micron digital breast spot device, 100 micron film digitization and 50 micron film digitization suggests that object conspicuity should be better for digital compared to conventional systems, but that there seemed to be minimal advantage to going from 100 to 50 microns. The 50 micron pixel system appears to provide a slight advantage in object contrast and perhaps in shape definition, but did not allow smaller objects to be detected.

The conventional x-ray source for mammography, with a molybdenum (Mo) anode and Mo filter, works well for breasts of low to moderate x-ray attenuation, but is not readily adaptable to the production of higher x-ray energies that are more suitable for imaging breasts of higher attenuation. Accordingly, alternative sources with anodes of rhodium (Rh) and tungsten (W) have been developed to improve the efficiency of the examination for thick or radiographically dense breasts. We have applied previously developed multiparameter optimization techniques to imaging systems using these alternative x-ray sources. Since these sources are intended to improve mammography of high-attenuation breasts, optimizations were performed for a range of breast thicknesses. Since high attenuation is generally associated with high scatter, optimizations for each source were done with a high-ratio, air-interspace grid similar to the one developed in our previous work. Preliminary results have been obtained for optimized system configurations using a W-anode source with Mo, Rh, and aluminum (Al) filters, and for a Mo-anode source with Rh filtration. These results indicate that the alternative sources studied can significantly improve the efficiency of mammography of high-attenuation breasts.

We report on several new techniques which we are developing for scatter correction and noise suppression in single-shot dual-energy chest radiography. For scatter, we use a deconvolution technique to correct for the scatter produced within the dual-energy cassette, which can result in scatter fractions as high as 30% in the high-energy image. For patient scatter, we have developed a technique which estimates the scatter by viewing each pixel in the image as a scattering source, and summing up its scatter contribution over all pixels. For scatter suppression, we describe the methods of noise clipping and noise forcing, both which use information from the low- and high-energy images to correct pixel values in the high-energy image which are unphysical. The noise clipping can also be combined with the correlated noise reduction technique, in order to combine the strengths of the two methods.

Image blur in digital imaging systems results from both the spatial spreading of quanta representing the image in the detector system and from the integration of quanta over the finite detector element width. Linear-systems theory has often been used to describe these blurring mechanisms as a convolution, implying the existence of a corresponding modulation transfer function (MTF) in the spatial-frequency domain. This also implies that the resulting noise- power spectrum (NPS) is modified by the square of the blurring MTF. This deterministic approach correctly describes the effect of each blurring mechanism on the overall system MTF, but does not correctly describe image noise characteristics. This is because the convolution is a deterministic calculation, and neglects the statistical properties of the image quanta. Rabbani et al. developed an expression for the NPS following a stochastic spreading mechanism that correctly accounts for these statistical properties. Use of their results requires a modification in how we should interpret the convolution theorem. We suggest the use of a `stochastic' convolution operator, that uses the Rabbani equation for the NPS rather than the deterministic result. This approach unifies the description of both image blur and image noise into a single linear-systems framework. This method is then used to develop expressions for the signal, NPS, DQE, and pixel SNR for a hypothetical digital detector design that includes the effects of conversion to secondary quanta, stochastic spreading of the secondary quanta, and a finite detector-element width.

We present statistically based error maps which graphically display the spatial distribution of two computational errors ensuing from quantitative T1 measurements. These computational errors, chi-squared error and coefficient of variation, represent the measures of goodness-of-fit and of the reliability of the fitted parameters, respectively. Visualization of the spatial distribution and assessment of the magnitude of these errors for each pixel allows one to compare and optimize quantitative MR imaging techniques to estimate tissue relaxation parameters and provides clinicians a method to differentiate the unreliable regions. This error mapping technique is demonstrated for a control study of seventy-one healthy volunteers to illustrate the patterns characteristic in these examinations.

Nonlinear quantitation tasks can be viewed as parameter estimation problems, and task performance quantified by the variance of parameter estimates. At high signal-to-noise ratio (SNR), the Cramer-Rao bound (CRB), an absolute lower limit on the variance of any unbiased estimator, is a valid predictor of the variance. At low SNR, however, the CRB may not be achievable, i.e., the realizable parameter variances may exceed the CRB. The determination of this SNR threshold, below which efficient estimation is no longer possible, is of great importance for the optimization of quantitative imaging systems. One approach to this problem is calculating the Barankin bound (BB), which at low SNR predicts larger parameter variances than does the CRB. The computation of the BB, which requires selection of a set of test points in parameter space, presents numerical difficulties. Choosing the test points based on (Chi) ²-confidence regions mitigates the numerical problems and renders the BB calculation practical using very high precision calculations in a computer algebra system. Simulations of a nonlinear two-parameter quantitation task demonstrated that the BB can be used to determine the SNR threshold region were ML-estimation performance no longer achieves the CRB. The BB, however, does not converge at low SNR. Therefore, it cannot be used as an absolute standard of achievable performance, and detailed simulations are necessary to investigate optimized strategies for data acquisition and analysis at very low SNR.

Scattered radiation in radioisotope imaging is considered in an estimation of radioactivity distribution in order to improve the image quality. To utilize the scattered components in the estimation, an observation system was modeled for each energy window, and Monte Carlo technique was applied for modeling photon transport in a water-filled cylinder. The observation system was assumed to have 8 energy windows between 90 keV and 154 keV for photons emitted by ^99mTc, and in order to clearly show the contribution of scattered components, it was also assumed to collect only one view projection data. In computer simulations, generalized analytic reconstruction from discrete samples (GARDS) was applied to estimate the source distribution, and the results show that when a large number of photons are collected, scattered components could improve the image quality.

Ultrasound is an inexpensive and widely used imaging modality for the diagnosis and staging of a number of diseases; nevertheless, technical improvements are needed before its full potential is realized. We believe that 2-D viewing of the 3-D anatomy, using conventional ultrasound procedures, limits our ability to quantify, diagnose and stage a number of diseases because: conventional ultrasound images are 2-D, multiple images must be integrated in the diagnostician's mind to develop a 3-D impression of the anatomy leading to a time-consuming process with increased operator variability; the patient's anatomy or orientation sometimes restricts the image angle, resulting in the optimal image plane necessary for diagnosis being unavailable; and, it is difficult to localize the conventional 2-D image plane and reproduce it at a later time, making it suboptimal for monitoring of therapy. Our efforts have focused on overcoming these deficiencies by developing 3-D ultrasound imaging techniques that are capable of acquiring B-mode, color Doppler and power Doppler images from existing ultrasound instruments, reconstructing the information in 3-D, and then allowing interactive viewing of 3-D ultrasound images on inexpensive desktop computers.

Matrix-array ultrasound is a new medical imaging modality that steers an ultrasound beam electronically in three dimensions. It is the first imaging modality that can view the heart in 3D in real time, making possible the `volumetricardiogram,' i.e., continuous beat to beat measurement of cardiac chamber volume. To create a fully automatic real-time volumetricardiogram, we have developed the flow integration transform (FIT), which operates on 2D images produced by slicing through the 3D ultrasound data. Although lacking rotational or scale invariance, the FIT is designed to operate eventually in dedicated hardware at very high speed, permitting the application of a large battery of test shapes within the period of a single ultrasound frame (approximately 45 milliseconds). To test the FIT, we have volumetrically scanned a series of 21 fluid-filled balloons. We used the FIT to detect circular cross-sections of the balloons by applying a battery of circles over a range of radii. The detected circles were used to compute volumes, which were then compared to volumes determined independently by weight. Our results are encouraging towards further development of this completely automated method of volume determination.

The scattering of acoustic waves from three dimensional compressible scatterers is considered. Particular attention is paid to cases where the scatterers have moderate magnitude in compressibility contrast and nondimensional wavenumber. A perturbation method based on Pade approximants is developed. It is shown that the Pade approximant model allows one to represent and evaluate the characteristic features such as internal resonances and mode shapes of the scattered pressure field. These modes are a function of the compressibility contrast and the frequency of the incident pressure wave. Numerical results are in agreement with the analytical solutions for scattering from a sphere.

Ultrasound pulses utilized for medical imaging and information-gathering appear to be coherently scattered from the many inhomogeneities within a tissue. The consequent interference effects complicate the spectral (Fourier) domain properties of the received signal: this is particularly troublesome when attempting to estimate tissue attenuation from backscattered data. A novel way to describe interference effects in (ultrasound) pulse-echo data is described. Recognition of their influence is achieved via analysis of the temporal phase of the pulse-echo signal, and correction of the artefact is achieved via novel signal processing techniques which rely on adjusting the locations of dominate zeros of the analytic continuation of short corrupted data segments (as pinpointed by the recognition procedure) into the complex frequency domain. Estimation of the mean frequency of a short pulse-echo data segment by the new method gives a reduction in variance by a factor of > 4 over existing Fourier methods. The technique finds application in an imaging technique which incorporates ultrasound attenuation information in conventional B-mode imaging.

Flat-panel x-ray imaging arrays based upon thin-film electronics are increasingly under development and investigation for a variety of applications. Our research has progressed to the point where three large area designs have now been fabricated, including a new 26 X 26 cm² array. These arrays are the largest self-scanning, solid-state imaging arrays thus far reported. In all probability, they represent only the first examples of an entirely new class of real-time imaging devices whose properties offer significant advantages over current radiographic and fluoroscopic x-ray technologies. A general overview of the current state of this emerging imaging technology is presented. Our large area array designs are described and x-ray images from a high resolution array are presented. Future challenges as well as anticipated trends and developments are discussed.

We propose a novel x-ray image intensifier that incorporates the x-ray absorption, image formation and amplification stages with a simple, flat panel structure. The device, to be called the x ray light valve (XLV), is based on two key components physically coupled in a sandwich structure; a solid state electrostatic x-ray detector and an electro-optic light modulator. X-ray absorption in the photoconductive detector controls the state of the electro-optic modulator via creation of charge carriers and the spatial and temporal variations that they induce in the modulator potential. Since the x-ray detector is electrostatic and the photo charge is strongly coupled to the light modulator, high resolution imaging is possible. Moreover, the amplification achieved by light modulation can avoid a coupling secondary quantum sink. Thus the XLV image may be coupled to an optical sensor, e.g., charge coupled device (CCD) camera, to produce a quantum noise limited digital radiographic imaging system. We outline the structure and the operation of the XLV. We discuss the features of the XLV based image acquisition and compare the properties of the XLV based and a phosphor screen based imaging chains.

The operational principle of a new, patented digital radiographic system using a multi-layer structure consisting of a thin-film pixel array, selenium x-ray photoconductor, dielectric layer and top electrode is described. Under an applied electric field, a diagnostic x-ray signal is obtained by the direct conversion of x-ray energy to electron-hole pairs which are collected as electrical charges by individual storage capacitor associated with each pixel element. The electronic readout sequence is initiated immediately after the x-ray exposure, and in several seconds, the image data is available for display on a video monitor, for data storage, data transmission, and hard copy generation. Signal strength of this direct conversion method is estimated to be significantly higher than that of other indirect conversion methods where light is first generated using a scintillator or phosphor and then detected by charge-coupled devices (CCDs) or thin-film-transistor (TFT) arrays in conjunction with photodiodes. In addition, since charges generated by x-ray photons move mostly along the direction of the bias electric field, images of very high spatial resolution can be obtained. The resolution limits are principally defined by the smallest pixel that can be manufactured. Recent x-ray images obtained from experimental detector panels are presented. X-ray sensitivity, dynamic range, signal-to-noise ratio, and spatial resolution are discussed.

We are developing a large area, flat panel solid-state detector for general application to digital radiology. The proposed detector employs a continuous photoconductive layer of amorphous selenium ((alpha) -Se) to convert incident x rays to electron-hole pairs, which are then separated and drawn to the surface of the (alpha) -Se by an applied electric field. The resulting charge image is digitally read out in situ using a large area active matrix array made with cadmium selenide (CdSe) thin film transistors (TFTs). The relationship between the potential imaging properties and the design parameters of this detector concept for digital mammography were analyzed theoretically using measured characteristics of (alpha) -Se layers and CdSe active matrices.

Recent advances in a-Si:H fabrication technology have opened the way for the application of flat panel imaging arrays in a number of areas in medical imaging. Their large area (up to approximately 26 X 26 cm), thin profile (< 1 mm) and real time readout capability make them strong candidates for the replacement of more traditional x-ray imaging technologies such as film and image intensifier systems. As a first step towards a device suitable for clinical use we have created a 24.4 X 19.4 cm array with 127 micrometers pitch pixels. This device serves as a testbed for investigating the effects of design changes on array imaging performance. This paper reports on initial measurements of the spatial resolution of this device used in conjunction with an overlaying Lanex Regular screen and 90 kVp x rays. The measured pre-sampled modulation transfer function (p.s. MTF) is found to fall below the predicted value by up to approximately 8%. At least part of this reduction seems to be due to scattering of light photons between the array and the surface of the phosphor screen contacting the array.

An algorithm has been developed to correct spatial distortion in image intensifier-based projections for three-dimensional (3D) angiographic reconstruction. A piecewise spatial warping technique is used to calculate a set of correction lookup tables which store the row and column subpixel spatial shifts, based on a reference grid image. Pixel amplitudes in the corrected image are determined from bilinear interpolation of the four surrounding pixels in the observed image. The method has been tested using an x-ray imaging chain with a 30-cm image intensifier positioned at various angular orientations and x-ray source distances. Prior to distortion correction, the maximum error between observed and expected reference point locations was found to be 14 mm. After correction, the maximum and mean errors were 0.23 mm and 0.053 mm, respectively.

An image intensifier-based rotational volume tomographic angiography imaging system has been constructed. The system consists of an x-ray tube and an image intensifier that are separately mounted on a gantry. This system uses an image intensifier coupled to a TV camera as a two-dimensional detector so that a set of two-dimensional projections can be acquired for a direct three-dimensional reconstruction (3D). This system has been evaluated with two phantoms: a vascular phantom and a monkey head cadaver. One hundred eighty projections of each phantom were acquired with the system. A set of three-dimensional images were directly reconstructed from the projection data. The experimental results indicate that good imaging quality can be obtained with this system.

In order to acquire 3D data of high contrast objects such as bone, lung and vessels enhanced by contrast media for use in 3D image processing, we have developed a 3D CT-scanner using cone beam x ray. The 3D CT-scanner consists of a gantry and a patient couch. The gantry consists of an x-ray tube designed for cone beam CT and a large area two-dimensional detector mounted on a single frame and rotated around an object in 12 seconds. The large area detector consists of a fluorescent plate and a charge coupled device video camera. The size of detection area was 600 mm X 450 mm capable of covering the total chest. While an x-ray tube was rotated around an object, pulsed x ray was exposed 30 times a second and 360 projected images were collected in a 12 second scan. A 256 X 256 X 256 matrix image (1.25 mm X 1.25 mm X 1.25 mm voxel) was reconstructed by a high-speed reconstruction engine. Reconstruction time was approximately 6 minutes. Cylindrical water phantoms, anesthetized rabbits with or without contrast media, and a Japanese macaque were scanned with the 3D CT-scanner. The results seem promising because they show high spatial resolution in three directions, though there existed several point to be improved. Possible improvements are discussed.

Visual assessment of arterial lesions from angiograms is subject to considerable inter- and intra-observer variability. To overcome these limitations, we are developing a method to perform reconstruction of vascular cross-sectional images from a limited number of x-ray angiographic cone-beam projections. The projection data are simplified by identifying blood vessels in each angiogram and removing signals due to other structures. The reconstruction is performed using the method of simulated annealing. An application of this approach to projections of cerebral vessels obtained from segmented CT slices of a cadaver injected with contrast agent are shown. We have also reconstructed an excised animal heart in order to test our method under more realistic image acquisition conditions including scatter, beam hardening, and variations in background signal.

In this paper, we present and validate a model that describes directly how the resulting object contrast in CT images is affected by (1) the x-ray collimation, (2) the table speed or helical pitch, (3) the size of the object, (4) the axial distance of the object to the reconstructed slice, (5) the distance of the object to the axis of rotation, and (6) the helical reconstruction algorithm employed. This model helps evaluate the trade-off between different helical scanning strategies and facilitates the selection of the optimal table speed and the x-ray collimation. With this model, the resulting tumor contrast can be predicted for given techniques. Besides the contrast, this model also provides the slice sensitivity profile. Several clinically important conclusions derived from this model are discussed.

MOS photodetector control technology that enables non-linear charge collection at the photosite permits breakthrough performance of CCD image sensors by overcoming previous limitations in the measurement of absorption characteristics of film and other non-linear forms.

A narrowing of the line spread function (LSF) has been observed when small amounts of low ionization potential polar dopant molecules were added to gas-filled high pressure kinestatic charge detector (KCD) for x-ray digital radiography. The LSF narrowing is attributed to different coexisting physical mechanisms. In this study, the impact of long-range dipole moment forces, associated with low ionization potential polar molecules, during ion-polars collisions, is investigated. Finally, this study is implemented with experimental examples.

A novel real-time portal imaging scanning detector, based on high-pressure gaseous electronics principles and operating up to 60 atmospheres, is presented and the predicted performance of this detector is analyzed. The idea is to utilize high pressure gaseous electronics imaging detectors operating in the saturation regime, aimed at improving image performance characteristics in real time portal imaging. As a result, beam localization errors are controlled, identified and corrected accurately and the patient radiotherapy treatment becomes more effective.

The kinestatic charge detector (KCD) is a digital radiographic detector that in brief consists of a scanning drift chamber whose velocity is synchronized to the drift velocity of the ions produced therein by detected x rays, so as to cause the moving (kinetic) ions to appear at rest (static) in the patient rest frame. The spatial resolution of the KCD is limited by mobility dispersion when the detector operates with noble gases such as xenon or krypton. The magnitude and dependence on drift distance of the peak widths of ionic signal pulses produced in the KCD provide a measure of mobility dispersion. These parameters have been measured in a KCD filled with krypton gas at several high pressures and in the same gas mixed with dopants (such as amines, alkanes, ethers, etc.). Considerable improvement and probably elimination of mobility dispersion is seen. The four amines tested (ammonia, monomethyl amine, dimethylamine and trimethyamine) and dimethyl ether were successful in reducing mobility dispersion whereas none of the alkanes tested (which include methane, ethane, propane, butane and cyclopropane in order of decreasing ionization potential) were successful. A closer look revealed that even though some of the alkanes (cyclopropane and butane) had the desired ionization potential, all of them had a zero or near-zero dipole moment. This suggests that both ionization potential and dipole moment are important parameters for an effective dopant. The details of this effect are given in an accompanying paper (Giakos et al.). In addition, electron attachment and ionic recombination in a kinestatic charge detector lead to a loss in useful signal from the detector and a consequent reduction in the detective quantum efficiency (DQE) because the pulse height distribution gets broadened. Extremely small amounts (1 ppm or more) of oxygen and other electronegative impurities can cause significant electron attachment and consequent recombination loss in the detector. The addition of a polyatomic gas such as carbon dioxide (according to our findings) and methane (according to other workers, at very low pressures however) has been found to reduce electron attachment to a considerable extent via the Ramsauer Effect. The consequent reduction in the signal loss due to recombination results inconsiderable improvement in the signal obtained from the KCD. It was shown that use of 0.1 percent (1000 ppm) of carbon dioxide along with a dopant and a parent gas resulted in considerable reduction in electron attachment.

We have designed and built a system for correlated x ray CT transmission and SPECT emission imaging with an array of photon counting detectors. The scanner operates in a third generation fan beam geometry by translating a 23 element high purity germanium detector across the fan to image phantoms and small animals. The x ray CT image is used to obtain an object specific, i.e., anatomically accurate, attenuation map for the reconstruction of the SPECT image. SPECT images are reconstructed with an MLEM code and the pixel values are scaled in physical units by determining a scaling factor from a uniform water phantom with homogeneous and known attenuation. Single myocardial slices of several pigs were imaged with a ^99mTc sestamibi imaging agent which is taken up in proportion to regional myocardial blood flow. The results show that ^99mTc uptake and regional myocardial blood flow, determined in vivo from reconstructed SPECT images, correlate with the measured in vitro data. Furthermore, the correlation is markedly improved by reconstructing the images with an object specific attenuation map obtained from the coregistered x ray CT image. We were also able to restore the ^99mTc sestamibi uptake from the reconstructed images to an accuracy between 40% and 90% of the true in vitro value, depending on the selection of maximum or mean pixel values in the regions of interest.

Through use of a computer model and an aluminum low contrast phantom developed in-house, a method has been developed which is able to grade the imaging performance of fluoroscopy systems through use of a variable, K. This parameter was derived from Rose's model of image perception and is here used as a figure of merit to grade fluoroscopy systems. From Rose's model for an ideal system, a typical value of K for the perception of low contrast details should be between 3 and 7, assuming threshold vision by human observers. Thus, various fluoroscopy systems are graded with different values of K, with a lower value of K indicating better imaging performance of the system. A series of fluoroscopy systems have been graded where the best system produces a value in the low teens, while the poorest systems produce a value in the low twenties. Correlation with conventional image quality measurements is good and the method has the potential for automated assessment of image quality.

The presentation concerns evaluations that were made to demonstrate the feasibility of a solid state detector intended to produce digital images in slit scan mammography equipment. The detector designed for this modality consists of a multilinear array of x ray sensitive elements. It operates in time delay integration mode. A CsI phosphor screen is directly coupled to a TDI CCD circuit. Typical cell size is 50 micrometers X 50 micrometers . The typical number of lines in the TDI direction is 200. The image format is 4800 X 6000 for a 24 cm X 30 cm field of view. Image acquisition time is 5 seconds. The output signal is converted into a 12 bit coded digital signal. Achievable performance has been demonstrated by performing experiments on a 1024 X 1024 CCD matrix. Cell size is 19 micrometers X 19 micrometers . Cells are electronically binned to achieve a 57 micrometers X 57 micrometers pixel dimension. The matrix is operated in TDI mode along columns, over a 10 mm wide sensitive area. A CsI phosphor layer is applied on the chip surface. Screen density is 65 mg/cm². The measured conversion efficiency is 190 electrons per absorbed x-ray quantum, 20 keV in energy. The detective quantum efficiency exceeds 50%. The modulation exceeds 10% at a spatial frequency of 8.8 lp/mm (Nyquist frequency of the experimental device). Experiments were conducted to estimate the performance of a Gd202S-based detector. The conclusion is that Gd202S is prohibited in such a detector structure. The major limitation lies in the noise contribution associated with the direct interaction of x-ray quanta in the silicon itself. This contribution is negligible in the case of a CsI-based detector.

In conventional mammography, x rays transmitted through the breast are converted to light in a phosphor screen, and the light exposes a film emulsion. The information in the image is degraded in this detector due to limitations in the screen and film. Photodiode arrays can convert the x rays directly into charge and overcome these problems. A preliminary investigation of a thick crystalline silicon photodiode array as a solid state digital detector was performed. The prototype device consists of a 300 micrometers thick, 256 X 256 photodiode array of 30 X 30 micrometers ² pixels. The array was hybridized to two different readout structures for evaluation purposes, one structure being used for imaging and the other for single pixel experiments. Imaging performance, such as linearity, resolution, and noise were measured and used to predict the performance of a proposed clinical version of the prototype. Results show the detector response to be linear over the range of exposures required for mammography, the modulation transfer function (MTF) to be superior to that of screen-film detectors, and the noise to be dominated by x-ray quantum fluctuation. Based on results from the prototype devices, we predict that the detective quantum efficiency (DQE) of the clinical design will be significantly higher than that of a screen-film detector for all spatial frequencies of interest.

The initial performance of a digital radiographic system incorporating a large-field (2016- channel) kinestatic charge detector and data acquisition electronics is discussed. The measured modulation transfer function of the system is 20% at 4 cy/mm. The measured detective quantum efficiency is 40 - 60%. These results are comparable with or better than those of current clinical (rare-earth film-screen and storage phosphor) systems. First images from the large-field system are shown and compared with those from commercial systems. Future system improvements in process or in planning are discussed.

The clinical utility of computed radiography (CR) versus screen-film for neonatal intensive care unit (ICU) applications is investigated. The latest versions of standard ST-V and high- resolution HR-V CR imaging plates were compared via measurements of image contrast, spatial resolution and signal-to-noise. The ST-V imaging plate was found to have equivalent spatial resolution and object detectability at a lower required dose than the HR-V, and was therefore chosen as the CR plate to use in clinical trials in which a modified film cassette containing the CR imaging plate, a conventional screen and film was utilized. For 50 portable neonatal chest examinations, plain film was subjectively compared to the perfectly matched, simultaneously obtained CR hardcopy and softcopy images. Grading of overall image quality was on a scale of one (poor) to five (excellent). Readers rated the visualization of various structures in the chest (i.e., lung parenchyma, pulmonary vasculature, tubes/lines) as well as the visualization of pathologic findings. Preliminary results indicate that the image quality of both CR soft and hardcopy are comparable to plain film and that CR may be a suitable alternative to screen-film imaging for portable neonatal chest x rays.

The computed radiography (CR) system consists of three processes: reading, image processing, and display. Image noise from the reading process consists of quantum noise and fixed noise. Quantum noise is dependent on exposure but fixed noise is independent of it. Quantum noise can be divided into light photon noise and x-ray photon noise. The former is inversely proportional to the light detection efficiency and the latter is independent of it. We separated the noise components of the Fuji computed radiography (FCR) 7000 system. At a spatial frequency of 1 cycle/mm, the ratio of x ray photon noise to light photon noise to fixed noise was about 8:1:1 at 1 mR(2.58 X 10^-7 C/kg). In the new CR system (FCR9000), we decreased x-ray photon noise and fixed noise. As a result, the FCR9000 system yields DQE of approximately 1.4 times higher at a spatial frequency of 0.5 cycle/mm and approximately 1.2 times higher at 1.0 cycles/mm at 1 mR (2.58 X 10^-7 C/kg).

We are using a beam port at the National Synchrotron Light Source Facility at Brookhaven National Laboratory as a source of monoenergetic photons. The photon source is radiation from a bending magnet on the x-ray storage ring and provides a usable x-ray spectrum from 5 keV to over 50 keV. A tunable crystal monochromotor is used for energy selection. The beam is 79 mm wide and 0.5 mm high. We imaged the ACR mammography phantom and a contrast-detail phantom using a phosphor plate as the imaging detector. Phantom images were obtained at 16, 18, 20, and 22 keV. Phantom thickness varied from 15 mm to 82 mm. These images were compared to images obtained with a conventional dedicated mammography unit. Subjective preliminary results show that image contrast of the monoenergetic images is similar to those obtained from the conventional x-ray source with somewhat sharper and cleaner images from the monoenergetic source. Quantitative analysis shows that the monoenergetic images have improved contrast compared to the polyenergetic derived images. Entrance skin dose measurements show a factor of 5 to 10 times less radiation for the monoenergetic images with equivalent or better contrast. Although there remain a number of technical problems to be addressed and much more work to be done, we are encouraged to further explore the use of monoenergetic imaging.

A new 2D imaging system structure is being studied in CEA-LETI for medical and industrial applications, beginning with dental applications. It consists of a bulk CdTe:Cl detection medium connected to a 2D electronic read-out circuit using the indium bumps techniques developed for infrared imaging. The feasibility of such a structure was tested first with 64 X 64 pixels, 100 micrometers pitch. The 900 micrometers thick CdTe sample suits well for x rays up to 100 keV. High absorption efficiency and high spatial resolution can be reached together by using this new x-ray detector. Due to the fact that the electric field channelizes the created charges, unlike structures using scintillators, this new structure requires no compromise in defining the thickness of the x-ray detector medium. Characterization was performed with 70 kV x rays from a standard dental x ray source. The performances (linearity, signal-to-noise ratio, spatial resolution) and the images obtained with the first prototypes confirm the advantages of such detectors. Based on these results, a 20 X 30 mm² imager for dental applications is now being developed by SOFRADIR and CEA- LETI.

A liquid nitrogen cooled CCD TV camera from Astromed, Ltd., has been used for quantitative x-ray medical imaging. The CCD is coupled to a Gd₂O₃(Eu) transparent ceramic scintillator on loan from the Ceramics Division of the General Electric Research Laboratories with an 80 mm f 1.3 oscilloscope camera lens optimized for 2:1 demagnification. High-resolution single-energy x-ray images have been acquired of lead bar patterns, human heel bones, and human teeth. Dual-energy bone-mineral densitometry images have been acquired of the foot and the femur of a rat.

An x-ray image detection system consisting of a fluorescent screen optically coupled to a CCD camera can produce a large uniform image. In this system, it is difficult to produce a noiseless image because of loss in the optical system. In order to improve the efficiency of light collection, we have developed a new screen with an interference filter. The angular distribution of the light from the new screen was improved, and the amount of light collected by the CCD was increased. The spatial resolution of the new screen was a little lower than that of the conventional screen. In this paper, we report a new application of an interference filter to a fluorescent screen, and report the performance.

This paper describes the performance characteristics of two high resolution charged-coupled device (CCD) film scanners for radiological imaging. The two models of recently developed CCD film scanners made by DBA Systems have been available on the market for ultra high resolution film digitization. One model of the scanner digitizes the film at 21 micrometers and the other one at 42 micrometers . Both systems can be interfaced to a PC. Line-pair, star-pattern and single edge on films were used to test the spatial resolution in the directions perpendicular and parallel to the CCD scan line. Step wedges generated on films through a mammographic system and print transparencies were employed to test the gray value versus the optical density response and variations on a `uniform area.' Geometric distortion of the digitized images was determined to be negligible at less than 1%. This gray value versus optical density response was linearly plotted from optical density (OD) 0 to 2.8. Depending upon optical density regions, gray value fluctuations varied. Both ultra high resolution CCD scanners showed reasonable performance. However, some digital noises were shown in the high OD range.

We have developed a high definition real-time DR system capable of 2000 line scan adopting a high resolution TV camera using a 2' Saticon imaging tube and a 16' high definition x-ray image intensifier. As a result, we obtained the dynamic range of a DR image, namely latitude of x rays, of 79 that is comparable to that of the film-screen system, and a limiting spatial resolution of 2.5 lp/mm for a maximum field-of-view of 16'.

It has been stated that digital mammography will reduce the exposure required for mammography. This poster explores the effects of decreased exposure on the information present in digital mammography. In general, the digital system performed better than screen film mammography with lower exposures. With the usual exposures used for screen film mammography, performance was equal. With high exposures sufficient to result in a dark film (OD 1.5), the digital system performed better than screen film with very small test objects. Proposals have been made to decrease the tube loading required for slot scanning devices by increasing KVP. This would result in their being less object contrast due to the decreases in the absorption coefficient of calcium compared to water at higher KVP. This poster looks at the potential for correcting the loss in object contrast that would result from the use of high contrast look up tables. It was found that in the tested system, one could restore the information in one of the two test objects used (but not the other), but that the image processing methods used would result in an image that radiologists would probably find inadequate for interpretation.

Image quality in digital fluoroscopic and angiographic systems has been limited by camera resolution and image processing capabilities especially for large fields of view. An intense research and development effort has led to an increase in the resolution performance of a digital image system designed for fluoroscopy and angiography applications. The result of this effort has been recently realized by the introduction of a 2000 line digital fluoroscopic/angiographic x-ray system capable of processing 2048² digital images at 7.5 frames per second. This paper presents the latest theoretical information and recently obtained clinical results demonstrating the improvement in resolution performance compared with the traditional 1000 line system. Several recent technological advances in the performance of image intensifiers, camera tubes, display monitors, and the semiconductor industry have enabled cost-effective solutions to earlier obstacles to 2000 line imaging. The various components of the 2000 line digital fluoroscopic imaging system affecting resolution are presented individually with emphasis on the total impact on system performance. These key components include large field of view image intensifiers, camera tubes, optics, image displays, hardcopy devices and image processing hardware. The processing hardware referenced in this paper is described in detail in a paper titled '2048 Line by 2048 Pixel High- Speed Image Processor for Digital Fluoroscopy' which is presented in the Image Display section of this conference. The MTF characteristics of these various components are illustrated and used to demonstrate their impact on the clinical images displayed on the monitor or hard copied on a laser printer. The results presented in this paper demonstrate the improvement in resolution performance with increased visibility of image details of clinical benefit that can be realized from a 2000 line digital fluoroscopic imaging system when compared with traditional 1000 line systems.

Compact sized computed radiography (CR) systems based on photostimulable phosphor technology are now being more widely used in radiology departments. Measurement of imaging performance at time of installation is essential to ensure that the CR system is operating within the manufacturer's specifications and producing a clinically acceptable image quality. At any given radiation dose, CR imaging performance primarily relates to image contrast, spatial resolution, and noise. Tests for these key aspects of CR performance are described and typical results, as obtained with a commercial system, are presented.

The relationships between the fundamental parameters of imaging statistics -- the probability density function, noise power spectrum, and autocorrelation function -- are analyzed for systems containing correlated noise sources. A calculation of the power spectrum for coherent events is rigorously derived and used to study the dependence of imaging statistics on correlation mechanisms, size of physical basis states, and characteristics of measuring devices. Examples of simple idealized systems illustrate fundamental phenomena and behaviors. Models of screen-film and storage phosphor systems are utilized to interpret reported experimental data.

A 1k X 1k frame transfer progressive scan imager was developed, based on a 1k X 1k interlaced frame transfer imager, the FT12. The sensors are designed for 2/3' optical format. They have a high resolution (1024 X 1024 pixels), overexposure handling by means of vertical anti-blooming, and square pixels of 7.5 micrometers by 7.5 micrometers . The frame rate is 30 frames per second. The single output register can run up to 40 MHz. The register can be clocked bi-directionally, which makes it possible to produce a mirrored image. The paper describes the sensors, simulations, and measurements.

While specialized phantoms for quality assurance have been provided with CT scanners since these devices were first marketed to radiology departments, there has been little in the way of integrated software and procedures to use these phantoms on an on-going basis. Typically, they are used initially when the scanner is installed, and then used only very intermittently thereafter, usually by the vendors' service personnel. Although calibration scans are performed routinely, these typically only establish the baseline for the accuracy and uniformity of CT numbers, and do not actually measure the resolution which the images are capable of achieving. Over the last four years, a software package to automatically analyze images from CT scanners has been developed, and this was adapted to use with MRI scanners in 1993. An additional software package has been developed to handle the results of the individual quality assurance scans in a data base, and allow for easy analysis and graphing of the results.

A multi-center study has previously evaluated the use of 2-dimensional transvaginal ultrasound (TVS) to measure the thickness of the endometrium as a risk indicator for endometrial abnormality in women with postmenopausal bleeding. In this paper we present methods using 3-dimensional TVS in order to improve the measurement, shape analysis and visualization of the endometrium. Active contour techniques are applied to identify the endometrium in a 3D dataset. The shape of the endometrium is then visualized and utilized to do quantitative measurements of the thickness. The voxels inside the endometrium are volume rendered in order to emphasize inhomogeneities. Since these inhomogeneities can exist both on the outside and the inside of the endometrium, the rendering algorithm has a controllable opacity function. A 3-dimensional distance transform is performed on the data volume measuring the shortest distance to the detected endometrium border for each voxel. This distance is used as a basis for opacity computations which allows the user to emphasize different regions of the endometrium. In particular, the opacity function can be computed such that regions that violate the risk indicator for the endometrium thickness are highlighted.

A high-resolution, portable, digital x-ray imaging device which replaces current film based systems has been developed. The system is intended to be used in field hospitals where on-line verification is required during treatment. Image acquisition is performed by a 3 X 4 matrix of charge-coupled-device (CCD) imaging sensors which view the output of a standard x-ray scintillation screen via an off-the-shelf optical system. The use of multiple, moderately priced CCD units results in a high resolution system with a low cost of production relative to other digital imaging systems of comparable resolution. The fields of view of each CCD are purposefully overlapped so as to facilitate image reconstruction. The acquisition of each radiographic image formed on a scintillation screen results in the production of twelve sub- images. A software algorithm is employed to detect the regions of overlap and create a single, continuous digital radiograph from the raw CCD data. Software methods are utilized to correct for barrel distortion affects that are caused by the use of low cost lens components.

Computed radiography (CR) based on storage phosphor (SP) detector is a new and complex technology for obtaining medical digital imaging. The CR system was developed such that its image quality in diagnosis (wide latitude, certainty of visual diagnosis), speed (image sensitivity), image processing, and the overall imaging performance exceeds or at least equals those of the conventional screen-film (SF) radiography systems which are considered to be `gold standard.' The unique capability of the CR system places a new responsibility on the medical physicist, radiologist, and radiologic technologist to ensure that the digital images give the same information at least as film, and improve rather than degrade the image. For that reason the quality of the image must be ensured and each component of the system must function properly. The clinical knowledge about quality control (QC) and the standard procedures for CR devices have not yet been established and are still under development. This paper presents the acceptance/QC of the CR image reader and laser imager. The procedure includes the acceptance of the equipment for its image quality, image sensitivity, and overall imaging performance. The procedures are currently and routinely being utilized in our institution.

Projection radiography, based on the storage phosphor (SP) imaging medium is a promising but challenging technology. SP plates consist of phosphor particles that are embedded in a polymer binder like film and then coated onto a flexible backing. Image quality is dependent not only on the performance of the plate, but also on the handling of the plate-cassette combination by the user and the reading system itself. The imaging plates included in this study are the 14' X 17' and 10' X 12' standard plates (ST-V) used for general radiography and the high resolution (HR-V) plates used in mammography and for extremity exams. The cassettes or image plate holders correspond to the plate sizes with the 14' X 17' being lead backed. Recorded downtime of the reader due to plate-cassette operation failures, diagnostic image quality affected by plate artifacts, and plate and cassette replacement rates are for a period of twelve months. Records are maintained through a detailed maintenance log, the image reader's internal record for plate use, and the system's internal error log for physical malfunctions. By keeping a detailed log on maintenance and performance of specific system components, communication to the vendor has been timely and effective in solving equipment failures. Plate-cassette maintenance must be an integral part of overall quality control (QC) that includes technologist training, physical plant (clinical environment) changes, and routine evaluation of the operation and performance of CR system components. Quality control combined with equipment improvements have minimized the downtime of the system, image artifacts and replacement rates of plates and cassettes.

Parameters are needed to assess quality assurance in a radiology department where computed radiography (CR) is the principal means of image acquisition. Laser-printed computed radiographs were collected for all patients examined over a period of several days. A sample of 1200 was sorted by subject anatomy and the associated exam information was entered into an EXCEL spreadsheet. Sensitivity (S) numbers were sorted into histogram and analyzed using standard descriptive statistics. Each film was over-read by a board-certified radiologist to assess whether the image was diagnostic and to determine if there were pathologic findings. A significant proportion of images were acquired using inappropriate menu codes. The histogram of S numbers for a given menu code describes a log normal distribution. The S number depends on the technologist's ability to control the technique. A significant proportion of the images were deemed non diagnostic, and many correlated to excessive S numbers. Some were a result of mispositioning. The S number is a valid retrospective measure of radiographic quality assurance. Departments using CR should strive for control on menu codes selected and S numbers produced. Such data should be available from PACS databases.

Experimental studies of the broadening of the full width at half maximum (FWHM) of the arrival time spectrum (ATS) in the research KCD system as a function of drift distance and x- ray tube mA settings were carried out. Results demonstrate that the broadening of the FWHM is mainly due to two factors, the space charge spreading of the ionic plane and the presence of more than one charge carrier with different mobilities in the ionic plane. Diffusional broadening is estimated to be almost negligible.

A kinestatic charge detector (KCD) with segmented signal-collection fingers was constructed to evaluate the dual-energy x-ray imaging performance of a KCD. The front segments of the KCD signal-collectors produce a digital low-energy image and the back segments produce a digital high-energy image. A gap between the front and back signal-collectors is used as a filter to increase the separation between the mean energies absorbed in the front and back segments. Preliminary measurements have been performed on the dual-energy KCD to determine its dual-energy imaging characteristics. The KCD output signal has been measured as a function of depth in the chamber. The ion drift velocity, modulation transfer function (MTF), detective quantum efficiency (DQE) and Wiener spectrum have been determined for both the front (low-energy) and back (high-energy) signal detection regions of the KCD.

The exact weighting function in 3D image reconstruction from 2D projections with cone beam geometry is obtained as the volume of intersection of a pyramidal ray with a cubic voxel. This intersection yields a convex polyhedron whose faces are formed by either the side of the pyramid or the voxel face. For each face of a voxel, we maintain a vertex link map. When one of the four pyramidal ray planes clips the voxel, we obtain a new face and a set of new vertices, while updating existing faces and their vertex link maps. Progressively clipping the voxel by the necessary ray planes yields the intersection polyhedron, whose faces and vertices are provided by the face list and its associated vertex link maps. To generate the weight, the volume of the polyhedron is calculated by dividing the polyhedron into tetrahedrons, whose volumes are summed. The exact calculated weights were used to reconstruct 3D vascular images from simulated data using a ROI (region of interest) limited ART (algebraic reconstruction technique). Comparing the results to those obtained from length approximation indicates that more accurate reconstruction could be achieved from the weights calculated with the new method.