Stephen P. Yanek 1 Introduction 759
Quentin E. Dolecek 2 Compression and Decompression 761
Robert L. Holland 2.1 Joint Photographic Experts Group (JPEG) Compression • 2.2 Moving Picture Experts Group
Joan E Fetter (MPEG) Compression • 2.3 Wavelet Compression • 2.4 Fractal Compression fohns Hopkins University 3 Telecommunications 766
3.1 Signal Hierarchy and Transfer Rates • 3.2 Network Interoperability • 3.3 Telemedicine Applications Compatibility
4 Conclusion 769
The principal objectives of this chapter are to introduce basic concepts and standards for data compression and communications in telemedicine systems, and to lay a foundation for further study in this field. According to the Institute of Medicine, telemedicine is the use of electronic information and communications technology to provide and support health care when distance separates the participants . Viewing a system as a collection of hardware, software, people, facilities, data, and procedures organized to accomplish a common objective offers insight into the wide range of issues that affect the design, deployment, and use of telemedicine systems. The interest in telemedicine systems has created significant demand for standards that ensure compatibility between heterogeneous systems that process and display health care data, improved compression and decompression algorithms, and economical and reliable ways for sending and receiving, storing, and retrieving multimedia data. Multimedia is a general term used for documents, presentations, and other means of disseminating information in the form of text, voice, and graphics, as well as moving and still pictures in color, gray scale, or black and white .
Telemedicine applications involve mainly image transmission within and among health care organizations. In an earlier time, the term "telemedicine" was used interchangeably with video teleconferencing, or specific clinical applications such as teleradiology or telepathology. Then, as now, a significant number of consultations between participants involved audio teleconferencing and exchange of facsimiles containing text, numerical data, or waveforms such as the electrocardiogram. Several steps are typically involved in transferring multimedia data from one site to another, including scanning and digitizing film images, and incorporating demographic and other patient information. Then may come compressing the volume of data to allow images to be sent more economically and quickly, followed by reconstruction of images at the receiving end for viewing and interpretation. In general, transactions used in telemedicine applications may be placed into three general categories: dynamic (e.g., interactive television); static (e.g., teleradiology); and a combination involving static data and interactive or dynamic features (e.g., telesurgery and telementoring) .
As depicted in Fig. 1, telemedicine systems comprise a variety of component technologies and services. Figure 1 illustrates technology and services that perform several basic functions, for example, image or data acquisition, digitization, compression, storage, data manipulation and computation, display, and communication. The software that enables many of the technologies and services is not apparent in the figure. Telemedicine applications have demanding networking requirements when large volumes of data are required. Since
Copyright © 2000 by Academic Press.
All rights of reproduction in any form reserved.
Patient and Communications Care-Giver Location Options
use of images by participants in telemedicine applications places great demands on hardware and software technology associated with each telemedicine system, we will focus here on image processing and telecommunications features.
In general, an image is a representation of an object, organ, or tissue made visible through physical and computational processes. Images used in telemedicine applications are frequently reproductions of a clinical image initially obtained using various modalities such as conventional projection X-rays, computed radiography (CR), computed tomography (CT), nuclear medicine (NM), magnetic resonance imaging (MRI) or ultrasound (US). A digital image processing system, for the purposes of this discussion, will consist of a set of hardware and software modules that perform basic functions such as acquisition, display, and communication.
As an example, a picture archive and communications system (PACS), discussed in Chapters 47 and 48, provides storage, retrieval, manipulation, and communication of digital images and related data. PACS manage image records that come from a variety of imaging sources. A PACS is designed to provide archiving as well as rapid access to images, and to integrate images of different modalities. The large volume of information produced by these modalities requires the use of appropriate storage systems or media for both recent images and archives for older images. Furthermore, reduction in the time and cost of transmission of images and related data requires effective and efficient compression and communications standards.
In some imaging modalities such as photography or projection X-ray film, the original image is the first analog counterpart of the imaged object. The photographic camera forms an image on a light-sensitive film or plate and the radiograph is a negative image on photographic film made by exposure to X-rays or gamma rays that have passed through matter or tissue. For computerized analysis this two-dimensional continuous function f(x, y) has to be discretized in space, and intensity or color. In the resulting digital image g(i, j) each pixel is represented by one or more bits of data depending on whether the image is binary, gray scale, or color. Typically, gray-scale medical images require 8, 10, or 12 bits per pixel and full color images require 24 bits per pixel. In other imaging techniques such as MRI, the original image is in digital form. Bits representing pixels are the essential elements that need to be stored, retrieved, transmitted, and displayed. For the purposes of compression we will consider the initial digital image as the original image.
Digital images often require large amounts of storage that depends upon the resolution of the imaging, digitization, or scanning processes. As the resolution or size requirement of the image increases, the amount of data increases. For example, an image that covers an area of 1 square inch with a density or resolution of 400 dots per inch (dpi) consists of 160,000 pixels. Therefore, an 8-by-10 inch image of the same resolution requires 12,800,000 pixels. As the volume of pixel data storage grows, the time required to access, retrieve, and display the data increases. The location (i.e., on-line or off-line), type, and efficiency of storage media affect access and retrieval times. Images are frequently compressed to reduce the file size for efficient storage and transmission. Once compressed, the image requires less space for storage and less time for transmission over a network and between devices.
An important question in the maturation of telemedicine concerns the quality and utility of the clinical images provided for interpretation. For example, dermatologists are trained to provide an accurate diagnosis and/or differential diagnosis using photographs and photographic slides during residency training, intramural exams, in-service exams, board certification preparation, and continuing medical education. Although dermatologists are accustomed to evaluating skin conditions using 2-by-2 inch photographic slides, there is reluctance to accept digital images because they inherently provide less visual information (resolution) than conventional clinical photography. In addition, color is an important factor in diagnosis of dermatological images. At present the rendition of colors in a telemedicine image is often determined by adjusting the red-green-blue mix in the display.
Table 1 sets the stage for discussions about data compression techniques by specifying the typical size of digital images that are generated by modalities just cited.
Compression and decompression are essential aspects of data management, display, and communication in telemedicine systems. The International Consultative Committee for
Telegraph and Telephone (CCITT) defines the standards for image compression and decompression. Initially defined for facsimile transmissions, the standards now contain recommendations that accommodate higher-resolution images. The main objective of compression is removal of redundancies that occur in three different types: coding redundancy, interpixel redundancy, and psychovisual redundancy. Coding refers to the numerical representation of the intensity or color of each pixel, and several techniques exist to optimize the choice of code primarily using the histogram of image intensity or color . Interpixel redundancy is related to the correlation among consecutive pixels in the image. If a black pixel is followed by 17 white pixels, the latters can be represented and stored in a more efficient manner than storing 17 pixels with the same values, for example, with a mechanism that indicates the start and length of the white-pixel run such as in run length coding . Psychovisual redundancy results from the fact that human perception of image information does not rely on specific analysis of individual pixels. Typically, local edges and texture structures are evaluated and compared to known information to interpret the image visually. The level and location of psychovisual redundancy can be determined only with feedback from a human operator and tends to be relatively subjective. While removal of coding and interpixel redundancies does not eliminate information from the image, the removal of psychovisual redundancies decreases the information in the information theoretic sense. The effect of this reduction depends on the image, observer, and application. Compression techniques that remove only coding and interpixel redundancies are lossless, compression techniques that also remove psychovisual redundancies are lossy. The evaluation of the quality of images compressed with lossy techniques is addressed in Chapters 49, 50, and 51.
Compression techniques generally attempt to achieve a compromise between two undesirable outcomes: potentially deleting critical information and insufficient reduction of the image file size. Furthermore, compression and decompression should not introduce errors or artifacts. Performance and cost are important factors in choosing between lossy and lossless compression. Higher performance (i.e., closer to lossless) compression and decompression typically requires more storage. In other words, an identical copy of the original image costs more than a "reasonable facsimile." In general,
Image acquisition modality
Image size (number of pixels)
Pixel value (number of bits)
Scanned conventional radiography
Was this article helpful?