John B. Zimmerman

Color display in ultrasonography

Color display is useful in medical ultrasonography to increase sensitivity to contrast, to improve quantitative comparability between image regions, and to allow multiparameter display. In its first section this paper expands upon these objectives, sets forth the concepts of color definition, and discusses the criteria that should be used to select a pseudocolor display scale. The second section surveys the published work applying color display in ultrasonography, and the final section presents some of our recent results on (1) producing continuous pseudocolor scales with a natural order and with the property that equal changes in driving signal produce equally perceptible changes along the pseudocolor scale; and (2) applying these scales to ultrasound images.

Adaptive Histogram Equalization For Automatic Contrast Enhancement Of Medical Images

With the large number of images that will be viewed simultaneously in a medical picture archiving and communication system (PACS) system in the diagnosis of a particular patient, image by image interactive contrast enhancement, at present by intensity windowing, becomes unacceptably time-consuming. Furthermore, windowing has disadvantages of being non-reproducible and providing adequate contrast primarily in selected image regions. The method of adaptive histogram equalization (ahe) appears to provide a solution to these problems. It is reproducible, automatic, and simultaneously provides contrast in all image regions. After summarizing the basic method, this paper will 1) describe a new contrast limited form of ahe that appears to allow its application to a wide variety of medical images, 2) present a VLSI machine design that will allow the calculation of ahe in a fraction of a second per megapixel, and 3) report the results of a study demonstrating that for chest CT images, ahe provides no measurable loss of diagnostic performance compared to the now standard windowing.

Contrast Transmission In Medical Image Display

The display of medical images involves transforming recorded intensities such at CT numbers into perceivable intensities such as combinations of color and luminance. For the viewer to extract the most information about patterns of decreasing and increasing recorded intensity, the display designer must pay attention to three issues: 1) choice of display scale, including its discretization; 2) correction for variations in contrast sensitivity across the display scale due to the observer and the display device (producing an honest display); and 3) contrast enhancement based on the information in the recorded image and its importance, determined by viewing objectives. This paper will present concepts and approaches in all three of these areas. In choosing display scales three properties are important: sensitivity, associability, and naturalness of order. The unit of just noticeable difference (jnd) will be carefully defined. An observer experiment to measure the jnd values across a display scale will be specified. The overall sensitivity provided by a scale as measured in jnds gives a measure of sensitivity called the perceived dynamic range (PDR). Methods for determining the PDR fran the aforementioned PDR values, and PDRs for various grey and pseudocolor scales will be presented. Methods of achieving sensitivity while retaining associability and naturalness of order with pseudocolor scales will be suggested. For any display device and scale it is useful to compensate for the device and observer by preceding the device with an intensity mapping (lookup table) chosen so that perceived intensity is linear with display-driving intensity. This mapping can be determined from the aforementioned jnd values. With a linearized display it is possible to standardize display devices so that the same image displayed on different devices or scales (e.g. video and hard copy) will be in sane sense perceptually equivalent. Furthermore, with a linearized display, it is possible to design contrast enhancement mappings that optimize the transmission of information from the recorded image to the display-driving signal with the assurance that this information will not then be lost by a -further nonlinear relation between display-driving and perceived intensity. It is suggested that optimal contrast enhancement mappings are adaptive to the local distribution of recorded intensities.

A psychophysical comparison of two methods for adaptive histogram equalization

Adaptive histogram equalization (AHE) is a method for adaptive contrast enhancement of digital images. It is an automatic, reproducible method for the simultaneous viewing of contrast within a digital image with a large dynamic range. Recent experiments have shown that in specific cases, there is no significant difference in the ability of AHE and linear intensity windowing to display gray-scale contrast. More recently, a variant of AHE which limits the allowed contrast enhancement of the image has been proposed. This contrast-limited adaptive histogram equalization (CLAHE) produces images in which the noise content of an image is not excessively enhanced, but in which sufficient contrast is provided for the visualization of structures within the image. Images processed with CLAHE have a more natural appearance and facilitate the comparison of different areas of an image. However, the reduced contrast enhancement of CLAHE may hinder the ability of an observer to detect the presence of some significant gray-scale contrast. In this report, a psychophysical observer experiment was performed to determine if there is a significant difference in the ability of AHE and CLAHE to depict gray-scale contrast. Observers were presented with computed tomography (CT) images of the chest processed with AHE and CLAHE. Subtle artificial lesions were introduced into some images. The observers were asked to rate their confidence regarding the presence of the lesions; this rating-scale data was analyzed using receiver operating characteristic (ROC) curve techniques. These ROC curves were compared for significant differences in the observers’ performances. In this report, no difference was found in the abilities of AHE and CLAHE to depict contrast information.

Contrast Perception With Video Displays

Distributed picture archiving and communication systems require electronic displays, today probably video displays. An obvious restriction with video displays, especially when multiple images are viewed, is on resolution both along image lines and due to the video raster. The effect of this restriction and the needs for improvement will be briefly reviewed. Perhaps a less obvious restriction of video displays is on contrast, as compared to film. Only limited grey-scale contrast is provided, and the fact that multiple images must be presented near each other on a single display means that the contents of one image will affect the perception of another. The perceptual and display mechanisms causing these effects will be described, means of specifying these contrast effects will be presented, and quantitative measures of the effectiveness of various display systems will be given. Pseudocolor scales provide a means on video displays of lessening the restriction on contrast. However, these scales bring with them problems of associability and variation in relative sensitivity across the scale. A linearization method will be presented for avoiding the second problem and thus allowing the comparison of different scales. Ppproachs for choosing sensitive pseudocolor scales without associability difficulties or contour artifacts will be presented, and specific scales superior to grey-scale will be recommended.

Concepts of the Display of Medical Images

Display is the step of the imaging chain in which the image is put into a form viewable by an observer. The objectives of display can be qualitative or quantitative, and each requires different display approaches. For each objective, producing a high-quality result requires facing issues in both the space and intensity dimensions. In the spatial dimension the major choices are of sampling and interpolation. In the intensity dimension the major choices are of sampling, display scales, assignment by the display device of display scale values to display-driving intensities, and contrast enhancement. Concepts and approaches, but not hardware, for each of these choices will be presented.

Optimization of Trade-offs in Error-free Image Transmission

The availability of ubiquitous wide-area channels of both modest cost and higher transmission rate than voice-grade lines promises to allow the expansion of electronic radiology services to a larger community. The band-widths of the new services becoming available from the Integrated Services Digital Network (ISDN) are typically limited to 128 Kb/s, almost two orders of magnitude lower than popular LANs can support. Using Discrete Cosine Transform (DCT) techniques, a compressed approximation to an image may be rapidly transmitted. However, intensity or resampling transformations of the reconstructed image may reveal otherwise invisible artifacts of the approximate encoding. A progressive transmission scheme reported in ISO Working Paper N800 offers an attractive solution to this problem by rapidly reconstructing an apparently undistorted image from the DCT coefficients and then subse-quently transmitting the error image corresponding to the difference between the original and the reconstructed images. This approach achieves an error-free transmission without sacrificing the perception of rapid image delivery. Furthermore, subsequent intensity and resampling manipulations can be carried out with confidence. DCT coefficient precision affects the amount of error information that must be transmitted and, hence the delivery speed of error-free images. This study calculates the overall information coding rate for six radiographic images as a function of DCT coefficient precision. The results demonstrate that a minimum occurs for each of the six images at an average coefficient precision of between 0.5 and 1.0 bits per pixel (b/p). Apparently undistorted versions of these six images can be transmitted with a coding rate of between 0.25 and 0.75 b/p while error-free versions can be transmitted with an overall coding rate between 4.5 and 6.5 b/p.

Algorithms For Adaptive Histogram Equalization

Adaptive histogram equalization (ahe) is a contrast enhancement method designed to be broadly applicable and having demonstrated effectiveness [Zimmerman, 1985]. However, slow speed and the overenhancement of noise it produces in relatively homogeneous regions are two problems. We summarize algorithms designed to overcome these and other concerns. These algorithms include interpolated ahe, to speed up the method on general purpose computers; a version of interpolated ahe designed to run in a few seconds on feedback processors; a version of full ahe designed to run in under one second on custom VLSI hardware; and clipped ahe, designed to overcome the problem of overenhancement of noise contrast. We conclude that clipped ahe should become a method of choice in medical imaging and probably also in other areas of digital imaging, and that clipped ahe can be made adequately fast to be routinely applied in the normal display sequence.

Evaluation Of 3D Voxel Rendering Algorithms For Real-Time Interaction On An SIMD Graphics Processor

The display of three-dimensional medical data is becoming more common, but current display hardware and image rendering algorithms do not generally allow real-time interaction with the image by the user. Real-time interactions, such as image rotation, utilize the motion processing capabilities of the human visual system, allowing a better understanding of the structures being imaged. Recent advances in general purpose graphics display equipment could make real-time interaction feasible in a clinical setting. We have evaluated the capabilities of one type of advanced display architecture, the PIXAR1 Image Computer, for real-time interaction while displaying three-dimensional medical data as two-dimensional projections. It was discovered during this investigation that the most suitable algorithms for implementation were based on the rendering of voxel rather than surface data. Two voxel-based techniques, back-to-front and front-to-back rendering produced acceptable, but not real-time, performance. The quality of the images produced was not high, but allowed the determination of an image orientation which could then be used by a later, high-quality rendering technique. Two conclusions were reached: first, the current performance of display hardware may allow acceptable interactive performance and produce high-quality images if a scheme of adaptive refinement is used wherein succesively higher quality images are generated for the user. Second, the correct algorithm to use for fast rendering of volume data is highly dependent upon the architecture of the display processor, and in particular upon the ability of the processor to randomly access image data. If the processor is constrained to sequential or near-sequential access to the voxel data, the choice of algorithms and the utilization of parallel processing is severely limited.

Determination of the relative visibility thresholds for basis functions of the Frazier-Jawerth transform

A new method for spatial-frequency analysis, the Frazier-Javierth transform (FJT), has recently been introduced which allows the simultaneous encoding of image properties in both the space and frequency domains. This technique appears to have applicability to the problem of image data compression, in a way similar to the Fourier transform or discrete cosine transform. A necessary step in the development of an FJT compression algorithm is to produce a method for appropriately quantizing the FiT coefficients. The current work undertakes the measurement of the thresholds for the visual detection of the FJT basis functions. These results will be used in the development of a psychovisually-based quantizer. The thresholds were determined by the use of psychovisual observer experiments. It is shown that the detectability of the basis functions is in accord with current knowledge of the spatial frequency sensitivity of the human visual system. However, these results are not predictable from simple psychovisual models, and because of the frequency domain behavior of the FJT, the measured values must be corrected before they can be used in a quantizer. The corrected values imply that the FJT quantizer may not match the properties of ordinary images as well as the psychovisual quantizers used for such methods as the discrete cosine transform.