Access to a Digital Xray Archive over Internet

Abstract

This paper describes a project involving the creation of an electronic archive of xray images, and the development of geographically dispersed workstations that access the image store, retrieve the image files over Internet, and allow viewers to display, manipulate, enhance and read the images.

1. INTRODUCTION AND BACKGROUND

The main goal of this project, DXPNET or Digital Xray Prototype workstations linked via InterNET, is to investigate the technical feasibility of developing, maintaining and operating an archive of digitized radiographs, and providing local and remote access to the archive over a wideband packet-switched wide area network such as Internet.

Specific objectives include:

1. Develop an archive for digitized xrays implemented via an optical disk jukebox, and interface this archive to the Internet.
2. Develop image retrieval and display workstations suitable for radiologists and other users to remotely access the archive and retrieve stored images over the Internet.
3. Evaluate technical aspects of system performance such as response time, compression ratio and use patterns.

The impetus for the DXPNET project is a periodic nationwide survey of public health conditions, the National Health and Nutrition Examination Survey (NHANES). The second such survey, NHANES II, yielded a broad spectrum of information on each of approximately 20,000 participants such as age, sex, eating habits and blood chemistry. A subset of the participants received a detailed examination that included radiographs of the cervical and lumbar spine. This resulted in a collection of approximately 17,000 films. The third survey, NHANES III, currently in progress, is expected to produce an estimated additional 10,000 films of hands, wrists and knees.

The problems that DXPNET addresses are: preservation of the radiographic collection from these surveys, extracting information from this collection, and providing remotely located researchers with access to it. To date, the project has succeeded in developing an affordable, PC based workstation for the quality control of the xrays that are digitized by UCLA. These image files are to be stored in an Electronic Xray Archive (EXA) comprising a high capacity optical disk jukebox controlled by a UNIX machine. To extract information from the radiographs, the project aims to build Standardized Readings Workstations (SRW) that will be used by geographically remote radiologists to access the images in the EXA. The radiologists may then retrieve them, view them, and generate readings that are then entered into the system. The availability of the existing Internet network and the promise of its eventual upgrade to gigabit/second speeds, offers the opportunity to fill these needs with a network solution. Hence, the solution proposed builds on the existing network infrastructure with a view toward utilizing future improvements.

2. SYSTEM AND NETWORK DESCRIPTION

In the overall DXPNET system, the resources are geographically distributed. They include systems for image capture, storage, and viewing. A schematic representation appears in Figure 1. The following is a functional description of each component, including critical features:

Figure 1.

Capture Workstation: Currently located at UCLA, this resource is capable of digitizing xray films as large as 14x17 inches at a minimum of 146 pixels per inch and 12 bits per pixel. The scanning operator monitors the digitizing process with an electronic display. Each image is sent to local storage as a DOS data file with no image header. Large films (lumbar spine) result in 10 Megabyte files and small films (cervical spine) result in 5 Megabyte files. Each pixel is stored in Intel standard format, i.e., the least significant byte appearing first. The most significant hex digit is set to zero and the maximum value denotes black while 0 denotes white. The image line beginning in the upper left hand corner appears first in the file. The digitized images are currently delivered on 5 1/4 inch WORM optical disks, but delivery via Internet is a future possibility.

QC Workstation: This resource, currently located at NCHS, is capable of preliminary quality checking of the digitization process by a technician. The technician verifies that all images are present, have the correct file name, are oriented properly, have labels obliterated, and are free of pencilled annotations. The workstation is able to rapidly display reduced spatial resolution versions of the images at 1Kx1Kx8. The technician can also pan over full spatial resolution versions of the images, viewing a 1Kx1Kx8 segment at a time. In order to format archived images for display, software hosted on the QC workstation scales contrast resolution from 12 to 8 bits and inverts black and white. Figure 2 shows a schematic of this workstation.

Figure 2.

Standardized Readings Workstation: This resource, currently under development, will be capable of efficiently displaying diagnostic quality images. The display subsystem is capable of displaying a 2Kx2.5K image in its entirety. The user will be able to interactively map the 12 bits of contrast per pixel in the captured image, to the display's 8 bits per pixel capability. Other functionality such as decompression, zoom, pan and local image buffering will also be included. This workstation will provide windowed access to Internet so that remote access to mainframe-based NHANES data (alphanumeric data including demographic information and blood chemistry lab tests, among others), access to images in the EXA, and communications among workstations can be achieved. Locally hosted software will enable a radiologist to record standardized radiological readings, and will pre-fetch images for rapid access. The performance parameters of this workstation are similar to those of workstations currently being successfully used in clinical practice at UCLA. They are also consistent with recommendations found in the literature [1], [2], [3]. See Figure 3.

Figure 3.

Electronic Xray Archive: This resource, currently under development, will provide access to the image collection stored in compressed form. It will be implemented incrementally with one or more optical disk jukeboxes. The archive will be centrally located at NLM and will be expandable to a capacity capable of hosting the entire collection of about 27,000 images, about 224 Gigabytes as computed below. With the reasonable expectation that a standardized technique, such as JPEG, will become available in hardware, compression will be implemented. If it is judged that a lossless compression technique is the only practical approach, the required capacity will thus be reduced by about 50% to 112 Gigabytes. Total capacity is estimated as follows:

Uncompressed NHANES II xray images
Cervical spine, 1463x1755x16/8 bits/byte =  5,135,130 bytes/image
Lumbar spine,   2048x2487x16/8 bits/byte = 10,186,752 bytes/image
5,100 cervical spine images x  5,135,130 =  26,189 Megabytes
11,900 lumbar  spine  images x 10,186,752 = 121,222 Megabytes
Total NHANES II:      147,411 Megabytes = 147.4 Gigabytes

Uncompressed NHANES III xray images
Hand and wrist, 1463x1755x16/8 =  5,135,130 bytes/image
Knees           2048x2487x16/8 = 10,186,752 bytes/image
5,000 hand/wrist x  5,135,130 =  25,676 Megabytes
5,000 knees  x  10,186,752    =  50,934 Megabytes
Total NHANES III: 76,610 Megabytes =  76.6 Gigabytes

Total estimated capacity required = 224 Gigabytes

1. It should have a maximum local access time of 10 seconds to any file on any volume (optical disk) once the file has been identified.
2. It should be modular and expandable so that it can be implemented incrementally as the collection increases in size.
3. It should support ANSI standard media, UNIX file format, SCSI 2.0 connectivity and NFS.
4. The archive software should be capable of handling both online media (within the jukebox) as well as offline media (stored on an adjacent shelf).

3. SYSTEM DESIGN CONSIDERATIONS

From the system architecture point of view, the archive will be a server in a client - server structure. The Standardized Readings Workstation will be the client. Performance of the FTP protocol which operates on the application level of the ISO Open System Model will be evaluated, as well as an application level protocol based on the Berkeley sockets mechanism. This latter protocol is being developed inhouse and is customized for the xray application. The evaluation will consist of a comparison of access time, ease of use and level of security. Initial loading of the archive will possibly require byte swapping, contrast inversion and image rotation about the horizontal center axis. The feasibility of using existing UNIX facilities such as dd (for byte swapping) to accomplish this formatting will be studied.

Since standardization is desirable, we will evaluate the role of ACR/NEMA, in terms of conversion difficulty, time and benefits. Managing access to the images on the archive will initially require a simple mapping from NHANES image identifiers to UNIX file names for each desired image. Entering requests for images in a batch mode will permit rapid access through pre- fetching. This procedure is consistent with the needs of current NHANES users and radiologists who will provide the initial standardized readings. After the standardized readings become available, the use of a suitable database product to manage the readings and image file pointers will be evaluated. This DBMS, such as Postgres or some commercial alternative, will allow mapping image identifiers from the NHANES data set to UNIX file names, and the addition of standardized readings so that images may be accessed on the basis of radiologist findings, independent of NHANES data. Long range goals include the indexing of images based on their content possibly by automatically quantifying features [4] and storing results in a relational table or other structure such as a feature list.

4. SYSTEM AND TECHNICAL ISSUES

Evaluation of the system will include technical issues as well as issues concerned with system usage. An evaluation of how well the system meets the needs of the radiologists doing the standardized readings will start with querying the users for their assessment of response time, ease of use, and utility of image processing modules for contrast manipulation, magnification, edge enhancement, and data recording.

A major objective of the technical evaluation is to identify the design factors for a migration path to a larger, more capable system. Data on image access time will be acquired for both uncompressed and compressed images, from optical disk and magnetic cache disk, using a local ethernet for delivery. The effects of local network loading and multiple simultaneous access requests will be tabulated and studied using UNIX network monitoring tools.

General technical questions include the following:

1. What are the design considerations for software which will efficiently and reliably convert an image collection to ACR/NEMA format?
2. What are the design considerations to efficiently and reliably manage an optical disk jukebox image archive as an Internet file server?
3. What are the design considerations for software which will efficiently and reliably deliver images from an optical disk archive to remote users via Internet?
4. What are the design considerations for workstation hardware and software which will be acceptable to a radiologist for viewing and commenting on images from a remote archive?

1. For each of N trials, how many seconds does it take to move a cervical spine image from archive to display? What is the mean and variance of this measurement and how does it vary with network loading?
2. For each of N trials, how many seconds does it take to move a lumbar spine image from archive to display. What is the mean and variance of this measurement and how does it vary with network loading?
3. What are the system throughput constraints and how can they be alleviated?
4. How many images of each kind can be cached in local storage and what is their access time?

1. For each of N trials, how many seconds does it take to compress a cervical spine image using these compression techniques? What is the compression ratio consistent with required image quality? What is the mean and variance of these measurements?
2. For each of N trials, how many seconds does it take to compress a lumbar spine image using these compression techniques? What is the compression ratio consistent with required image quality? What is the mean and variance of these measurements?
3. What are the system throughput constraints, and how can they be alleviated?

The most salient outcome will be the establishment and exercise of a prototype image database, accessible via Internet, and the development of workstation hardware and software to use the database. An evaluation plan will be developed that will define the data and analysis procedures to be used to measure technical performance. The technical specifications of this testbed system will be described and a migration path to a more capable system will be identified.

6. ACKNOWLEDGEMENT

The origination and early development of this project is in large measure due to the efforts of Mr. John Cookson, formerly an electronics engineer with the National Library of Medicine, and now with the Library of Congress.

7. REFERENCES

1. Arenson, Chakraborty, Seshadri and Kundel, "The Digital Imaging Workstation", Radiology, 176(2), August 1990.

2. Beard, DV, "Computer Human Interaction for Image Information Systems", Journal of the American Society for Information Science, 42(8):600-608, 1991.

3. Fisher, Grover, Brauer and Ritchie, "Digital Image Display Station Performance Requirements Based on Physician Experience with a Prototype System" Journal of Digital Imaging 1989 Aug;2(3):150-55.

4. Chwialkowski, Shile, Pfeifer, Parkey and Peshock, "Automated Localization and Identification of Lower Spinal Anatomy in Magnetic Resonance Images", Computers and Biomedical Research, 1991 Apr;24(2): 99-117.