Evaluation of a combined two- and three-dimensional compression method using human visual characteristics to yield high-quality 10:1 compression of cranial computed tomography scans.

RATIONALE AND OBJECTIVES: The compression of cranial computed tomography scans was improved by using independent intra- and interframe compression techniques. METHODS: For intraframe compression, an image was decomposed into four subimages, one subimage was chosen as a reference subimage, and three of the subimages were predicted from the reference subimage. The prediction error was encoded with a classified vector quantizer (CVQ) based on human visual perception characteristics. Interframe redundancy is exploited by a displacement estimated interslice (DEI) algorithm that encodes the differences between reference subimages from adjacent slices. This combined DEI/CVQ method was subjectively evaluated by 13 radiologists under a blinded protocol, and was compared to the CVQ method alone, the DEI method alone, the original images, and to a standard intraframe discrete cosine transform (DCT) compression method. RESULTS: Only the combined DEI/CVQ method at 10:1 compression was not scored significantly different from the original images. At 15:1 compression, the DEI/CVQ method was scored significantly better than the 10:1 DCT and any other 15:1 compression methods. CONCLUSIONS: Compressed image quality is enhanced by exploiting inter- and intraframe redundancy, and by modeling some characteristics of human visual perception. The DEI/CVQ method is well-suited for progressive transmission, and thus, holds potential in teleradiology as well as picture archiving and communications systems.

A new method for computed tomography image compression using adjacent slice data.

RATIONALE AND OBJECTIVES: The authors developed and subjectively evaluated an interslice compression algorithm that explores the redundancy among adjacent slices of an x-ray computed tomography (CT) scan. This algorithm has been compared to an intraslice compression algorithm based on the two-dimensional discrete cosine transform. METHODS: Nine x-ray CT head images from three patients were compressed with this interslice method at compression ratios of 5:1, 10:1, and 15:1. The same images were also compressed with the intraslice method at the same ratios. Six radiologists judged quality of randomly selected compressed and decompressed images compared to that of the originals. The evaluation data were analyzed statistically with the analysis of variance and Tukey's multiple comparison. Kappa-like statistics (Williams index and O'Connell and Dobson indexes) were also calculated to measure the agreement among readers beyond the amount expected by chance. RESULTS: The interslice coding algorithm showed significantly better quality than the intraslice method at significance level 0.05, even though there was no difference in the objective distortion measure (signal-to-noise ratio). Also, the quality of 10:1 compressed images with the interslice coding algorithm was not significantly different from that of the originals at level 0.05. While large variations in agreement occurred among readers, the overall agreement was statistically significant. CONCLUSIONS: By using adjacent slice information in compressing x-ray CT images, significantly better quality in compressed and decompressed images was achieved. While 10:1 compressed images with the interslice algorithm were not significantly different from the originals in quality at level 0.05, effect on diagnostic accuracy remains to be investigated.

Defective lipid disposal mechanisms during bacterial infection in rhesus monkeys.

Mechanisms producing hypertriglyceridemia during bacterial sepsis have not been well defined. In this study lipid disposal mechanisms were assessed in 76 infected and 19 control male rhesus monkeys by the ability to dispose of triglycerides after: (1) oral lipid loading; (2) intravenous lipid loading; and (3) by lipolytic enzyme activity tests as measured by postheparin lipolytic activity (PHLA). Studies were performed both before and 48 hr after intravenous inoculation with either Salmonella typhimurium or Diplococcus pneumoniae when illness was uniformly severe and fasting serum triglyceride elevations were increased maximally. S. typhimurium-infected monkeys demonstrated significant fasting hypertriglyceridemia (p is less than 0.001), reduced clearance of orally and intravenously administered lipid and markedly reduced PHLA. During this gram-negative sepsis, mild lethargy, slight diarrhea, and a 2% mortality were observed. During D. pneumoniae sepsis, average fasting triglyceride concentrations were slightly, but not significantly elevated. While oral lipid clearance was impaired, intravenous lipid clearance was unimpaired, and PHLA was slightly reduced. Marked lethargy, agitation, and a 20% mortality were present during this gram-positive infection. Results of this study support the concept that an impairment of lipid disposal mechanisms, particularly during gram-negative sepsis with S. typhimurium, may significantly contribute to the observed hypertriglyceridemia.

A computed tomography-radiation therapy treatment planning system utilizing a whole body CT scanner.

An external beam radiation therapy treatment planning system has been developed to run on the GE CT/T whole body scanner. The system interactively obtains treatment planning information directly from CT scans, including relative density conversion of user input inhomogeneity regions. The program generates beams isodose tables from input TAR-SAR data using the Cunningham model. Inhomogeneity corrections are applied using the power law TAR method at low energies and the TAR ratio method at high energies. Beam data are generated on the central axis and at off-axis locations coplanar with each CT scan, and isodose distributions are displayed on any transverse, coronal or sagittal plane. Examples of plans and initial verification results are discussed.

Statistical distributions of DCT coefficients and their application to an interframe compression algorithm for 3-D medical images.

Displacement estimated interframe (DEI) coding, a coding scheme for 3-D medical image data sets such as X-ray computed tomography (CT) or magnetic resonance (MR) images, is presented. To take advantage of the correlation between contiguous slices, a displacement-compensated difference image based on the previous image is encoded. The best fitting distribution functions for the discrete cosine transform (DCT) coefficients obtained from displacement compensated difference images are determined and used in allocating bits and optimizing quantizers for the coefficients. The DEI scheme is compared with 2-D block discrete cosine transform (DCT) as well as a full-frame DCT using the bit allocation technique of S. Lo and H.K. Huang (1985). For X-ray CT head images, the present bit allocation and quantizer design, using an appropriate distribution model, resulted in a 13-dB improvement in the SNR compared to the full-frame DCT using the bit allocation technique. For an image set with 5-mm slice thickness, the DEI method gave about 5% improvement in the compression ratio on average and less blockiness at the same distortion. The performance gain increases to about 10% when the slice thickness decreases to 3 mm.

A predictive classified vector quantizer and its subjective quality evaluation for X-ray CT images.

The authors have developed a new classified vector quantizer (CVQ) using decomposition and prediction which does not need to store or transmit any side information. To obtain better quality in the compressed images, human visual perception characteristics are applied to the classification and bit allocation. This CVQ has been subjectively evaluated for a sequence of X-ray CT images and compared to a DCT coding method. Nine X-ray CT head images from three patients are compressed at 10:1 and 15:1 compression ratios and are evaluated by 13 radiologists. The evaluation data are analyzed statistically with analysis of variance and Tukey's multiple comparison. Even though there are large variations in judging image quality among readers, the proposed algorithm has shown significantly better quality than the DCT at a statistical, significance level of 0.05. Only an interframe CVQ can reproduce the quality of the originals at 10:1 compression at the same significance level. While the CVQ can reproduce compressed images that are not statistically different from the originals in quality, the effect on diagnostic accuracy remains to be investigated.

Stereotactic software for the GE 8800 CT scanner.

A computer software program (Seapit) was developed for precise determination of intracranial targets identified by stereotactic computed tomography (CT). This program was added to the software of a GE 8800 CT scanner to perform the following operations: millimetre precise calculation and display of the rectilinear coordinates of a target identified on axial CT images; preplotting of phantom target trajectories on the CT images or electronic radiographs; calculation of probe angles required to achieve various trajectories; display of a coordinate scale on each CT image to allow direct target determination without mathematical calculations; calculation of the intercommissural plane for functional neurosurgery. In a series of 100 patients undergoing stereotactic surgery, the Seapit program proved to be a superior and accurate method of target coordinate calculation. Preview display on the CT images of 'phantom' probes significantly enhanced the safety of stereotactic intervention.

Preliminary experience with portable digital imaging for intensive care radiography.

A digital radiography system based on reusable, photostimulable phosphor technology was evaluated in approximately 3,500 portable chest radiographs of patients in an intensive care unit. The system functioned well in this application. No major problems were encountered in the visualization of tubes or catheters or in the detection of pneumothoraces. Assessment of fluid volume status or the presence of small pleural effusions, especially when these were bilateral, was initially somewhat difficult but became easier as investigators became familiar with the system. Radiologists were quicker than nonradiologists to accept the minimized two-on-one display format. Critical evaluation of the overall performance of digital systems such as this one is needed for a better definition of the system's strengths and weaknesses. Specifically, statistical analyses of the ability to detect disease states such as pneumothoraces, interstitial lung disease, lung nodules, and pleural abnormalities need to be performed.

Digital hardware in CT and NMR.

How much does one need to know about the computer part of a computed tomographic or nuclear magnetic resonance system to select the best system and to use it effectively? Enough to comprehend how features or options relate to individual applications and needs. Besides acquiring some basic knowledge, one needs to consider upgradability, word size, and features of the image display controller, graphic input, array processor, and other hardware.

Authentication and management of radiologic reports: value of a computer workstation integrated with a radiology information system.

An increasing number of radiology departments are using computers to facilitate management of radiologic information. Transcription, storage, and printing of radiologic reports are among the primary functions of a radiology information system. Consequently, manual signature of radiologic reports is being replaced by on-line electronic authentication. However, the utilities provided on most radiology information systems to review, edit, and otherwise manipulate radiologic reports are relatively crude compared with commercial word-processing and data-base management software available for personal computers. We have developed a personal computer software system that is integrated with our existing radiology information system; expedites the radiologist's task of reviewing, editing, and authenticating (i.e., signing) radiologic reports; provides a teaching file data base on the workstation into which radiologic reports and patients' demographics can be instantly transferred; and provides a similar data base to facilitate follow-up on those examinations deemed appropriate for quality-assurance procedures. These features improve the radiologist's efficiency and increase his or her willingness to more fully exploit the intended purposes of a radiology information system.

Data-base management system on a CT scanner computer: application to teaching file and procedure records.

A prototype relational data-base management system was installed on computed tomographic (CT) scanner computers at two hospitals. This was used to create computerized indices for a teaching file and a record of CT procedures. Several problems commonly encountered when maintaining and using a radiologic teaching file were solved. Interesting cases were easily retrieved for teaching, conferences, or publication because the system permits rapid search on the basis of patient name, identification number, date, diagnosis, special description, or a combination of these data. The procedure record index contains these data as well as administrative and technical data on all CT examinations. These data are entered into the data base semiautomatically. The result is an extensive set of records that is easily accessible and requires a minimum of manpower to maintain.

Radiological images on personal computers: introduction and fundamental principles of digital images.

This series of articles will explore the issue related to displaying, manipulating, and analyzing radiological images on personal computers (PC). This first article discusses the digital image data file, standard PC graphic file formats, and various methods for importing radiological images into the PC.

Displaying radiologic images on personal computers: image storage and compression: Part 1.

This is the third article of our series for radiologists and imaging scientists on displaying, manipulating, and analyzing radiologic images on personal computers. Part 1 of this article discusses image storage and reviews the basic concepts of information theory and image compression; part 2 will discuss specific methods of image compression. There are a wide variety of removable storage devices available to users who need to archive radiologic images on their personal computers. Tape drives have potentially very large storage capacity but slow performance. Removable SyQuest (SyQuest Technology, Femont, CA) and Bernoulli disks have near hard disk performance and can store from 100 to 150 Mbytes. Magneto-optical drives can store nearly 1 Gb on a 5.25" disk, with somewhat slower performance. Selecting the most appropriate storage solution requires a careful balance of the user's requirements, including performance, storage needs, cost and compatibility with other users. Despite the advances in low cost high capacity storage technology, image compression remains a crucial technology for modern diagnostic radiology because digital images require such large amounts of storage. Image compression is possible because radiologic images have relatively low entropy (high information content) compared with random noise. Image compression is classified as lossless (nondestructive) or lossy (destructive). Lossless image compression commonly achieve compression ratios of 1.5:1 to 3:1 (33% to 67%), whereas lossy compression can compresses images from 3:1 to 30:1 (67% to 97%).(ABSTRACT TRUNCATED AT 250 WORDS)

Displaying radiologic images on personal computers.

This is the second article of our series for radiologists and imaging scientists on displaying, manipulating, and analyzing radiologic images on personal computers (PCs). The first article discussed the digital image data file, standard PC graphic file formats, and various methods for importing radiologic images into the PC. This article discusses the hardware, software, and user interface issues related to displaying gray scale images on PCs. In particular, this segment focuses on the process of converting the digital image into gray shades on a color monitor. A method for displaying and interactively setting the window width and window level parameters of 16-bit radiologic images on PCs with standard red green blue graphic hardware is illustrated in a sample application.

The need and user requirements for integrating images with radiology reports.

Radiology reports are likely to be more useful if they contain appropriate graphic material. Diagnostic conclusions and recommendations become more convincing and useful when the clinician personally can review the image on which these are based. Modern desk-top publishing techniques make it possible to incorporate radiographic images, appropriately selected and annotated, as part of the radiology report. It is believed that such illustrated reports would be preferred by referring physicians, notwithstanding a significant loss of image detail. A survey of these referring physicians was carried out to determine whether this hypothesis was correct.

Developing a framework for worldwide image communication.

The increasing mobility of the population and frequent changes in healthcare coverage, in both the government and private sectors, require integration of medical records not only longitudinally, but also across a variety of healthcare providers. Early in 1998, the federal government decided to solve this problem by constructing a framework for access to medical records by all of the government's health care facilities, called the Government Computer-Based Patient Record (GCPR). The government consortium chose a proposal by Litton PRC, a partnership of 11 companies with complementary areas of expertise. The framework is based on open systems, which use publicly available standards, and includes a Master Patient Information Locator that allows access to medical information from remote facilities, based on creating a unique identifier for each and every individual patient. PRC will use the Digital Imaging and Communications in Medicine (DICOM) imaging standard for radiology, supplemented by Health Level Seven (HL7).

Dr. Browse, a digital image file format Browser.

The emerging widespread adoption of the Digital Imaging Communications in Medicine (DICOM) standard will increase the demand for radiologic image transfer between radiologic image acquisition, archive, display and printing devices. Unfortunately, there are and will continue to be many devices that do not and will not support this standard, especially older radiologic equipment and devices from nonradiologic vendors. Determining the image file format characteristics of images from such equipment is often difficult, and done on an ad hoc basis. We have developed a software tool that assists users in determining the image file format parameters of unknown radiologic images.

Displaying radiologic images on personal computers: image processing and analysis.

This is the fourth article of our series for radiologists and imaging scientists on displaying, manipulating, and analyzing radiologic images on personal computers. Classic image processing is divided into point, area, frame, and geometric processes. Point processes change image pixel values based on the value of the pixel of interest. Histogram equalization adjusts the pixel values in the image based on the distribution of pixel values. Area processes change the pixel of interest based on the values of the surrounding pixels, known as the neighborhood. Area processes using a convolution kernel are often used as image filters. Common convolution kernels include low-frequency, high-frequency, and edge-enhancement filters. Edge enhancement can be performed with convolution kernels such as shift and difference, gradient-directional and Laplacian filters, or with nonlinear methods such as Sobel's algorithm. Frame processes mathematically combine two or more images, often for noise reduction and background subtraction. Geometric processes alter the location of pixels within the image, but usually not the pixel values. Common radiologic applications of image processing include window width and window level adjustments (point process), adaptive histogram equalization (area process), unsharp masking (area process), computed radiography image processing (combined area and point processes), digital subtraction angiography (frame and geometric processes), region of interest analysis (area process), and image rotation (geometric process). As digital imaging becomes more widespread, radiologists need to understand the image processing that is fundamental to these modalities.

Displaying radiologic images on personal computers: image storage and compression--Part 2.

This is part 2 of our article on image storage and compression, the third article of our series for radiologists and imaging scientists on displaying, manipulating, and analyzing radiologic images on personal computers. Image compression is classified as lossless (nondestructive) or lossy (destructive). Common lossless compression algorithms include variable-length bit codes (Huffman codes and variants), dictionary-based compression (Lempel-Ziv variants), and arithmetic coding. Huffman codes and the Lempel-Ziv-Welch (LZW) algorithm are commonly used for image compression. All of these compression methods are enhanced if the image has been transformed into a differential image based on a differential pulse-code modulation (DPCM) algorithm. The LZW compression after the DPCM image transformation performed the best on our example images, and performed almost as well as the best of the three commercial compression programs tested. Lossy compression techniques are capable of much higher data compression, but reduced image quality and compression artifacts may be noticeable. Lossy compression is comprised of three steps: transformation, quantization, and coding. Two commonly used transformation methods are the discrete cosine transformation and discrete wavelet transformation. In both methods, most of the image information is contained in a relatively few of the transformation coefficients. The quantization step reduces many of the lower order coefficients to 0, which greatly improves the efficiency of the coding (compression) step. In fractal-based image compression, image patterns are stored as equations that can be reconstructed at different levels of resolution.