Nephropathology Consultation via Digitized Images" J. HELEN CRONENBERGER! HENRY HSIAO, RONALD J. FALK, AND J. CHARLES JENNETTE University of North Carolina at Chapel Hill School of Medicine Chapel Hill, North Carolina 27599
The Glomerular Disease Collaborative Network (GDCN) evolved in 1985 from a renal biopsy diagnostic service. GDCN was established to provide renal biopsy diagnostic service so that patients with renal disease could remain in their home communities and be treated by their own local physician. From a beginning with 15 associated nephrologists, GDCN has now grown to include more than 150 nephrologists practicing at approximately 53 sites. Most are located in North Carolina, Florida, Georgia, South Carolina, Tennessee, and Virginia, with a few in Oregon, Washington, Minnesota, and Ohio (FIG.1). For the past two years, GDCN has been networking via fax machines which are used to accrue patient data dealing primarily with pathologic phenotypes, clinical manifestations, and treatment responses of renal diseases.',? The referring nephrologists performed a renal biopsy at their local hospital and then mailed the specimen to GDCN Nephropathology Laboratory in Chapel Hill. The biopsy material was processed for light, immunofluorescence, and electron microscopy using standard procedures. The processed biopsy was evaluated by a nephropathologist who took representative photomicrographs and electronmicrographs. A diagnosis based upon biopsy evaluation plus clinical information supplied by the referring nephrologist was rendered. Representative photographs along with typed evaluation and diagnosis were mailed to the referring nephrologist. The routine of processing and evaluating the biopsy specimen and mailing photographs and reports to the referring nephrologist took an average of one week. THE PROBLEM The problem undertaken was to decrease the one-week time period required for the nephrologists to receive micrographs and reports, and to enhance the pathologistnephrologist exchange by developing computer technology for interactive clinical consultation that incorporated images and voice. The consultant nephropathologist and the referring nephrologist needed to be able to simultaneously view an image of light, immunofluorescence or electron microscopy results, to point to areas of the
"This project was supported by a grant from The Kate B. Reynolds Health Care Trust, Winston-Salem, North Carolina. Computers for this project were provided by the IBM Corporation, Raleigh, North Carolina. 'Address correspondence to Dr. J. H. Cronenberger, Department of Biomedical Engineering, CB No. 7575, University of North Carolina Medical School at Chapel Hill, Chapel Hill, NC 27599-7575. e-mail: [email protected] 28 1
ANNALS NEW YORK ACADEMY OF SCIENCES
282
THE GLOMERULAR DISEASE COLLABORATIVE NETWORK
B.
* ,
UNC - Central Office $r Academic Medical Centers 0 Private Practices
FIGURE 1. Geographical sites of the Glomerular Disease Collaborative Network. More than 150 nephrologists practicing at 53 sites in ten states form the network for which interactive computer communications providing image, annotation, and verbal exchange were developed.
image and have the other immediately see the pointer, and to talk with one another while viewing and discussing the images and case history. Initial constraints applied to the solution of the problem were that the system (1) must provide image resolution to allow review of biopsy tissue, (2) must be currently available, (3) must utilize existing lines of communication, (4) must easily and naturally integrate into the daily routine of the nephropathologist and nephrologist, and ( 5 ) must be economicallyjustifiable from the viewpoint of the practitioner. METHODS Development of the computer communications and resolution of the stated problem was a collaborative effort of GDCN and the Department of Biomedical Engineering’s Communications Research Facility (BMECRF). BMECRF already had a Distance Clinical Consultation and Education Communications (DCCEC) project in progress with the Area Health Education Centers (AHEC). This DCCEC project was already involved in developing the communications system that will be described in this paper. Both GDCN and AHEC required similar technology. Solution of the stated problem was, therefore, related to the North Carolina AHEC which comprises a series of nine establishments around the state (FIG.2). These nine units are directed from a central office at the University of North Carolina Medical School. AHEC was interested in retention of practicing health professionals in the rural areas of North Carolina. On-site accessibility to consultation and continuing education by the rural health care providers has been found to be a major factor in retention. The interactive computer communicationssystem described in this paper is, therefore, designed and developed with the applications of both GDCN and
CRONENBERGER el al.: NEPHROPATHOLOCY CONSULTATION
283
AHEC in mind, even though discussion and description will be of the GDCN application. Image Resolutiod
Several digitizing systems were used to capture images of varying horizontal and vertical resolutions. One was the International Business Machine (IBM) video capture board with their Audio Video Connection (AVC) software. A second was the FG-100-AT board (Imaging Technology Incorporated, Woburn, MA) with Imagepro software (Media Cybernetics, Silver Spring, MD). A third was Targa+ MCA board (Truevision, Indianapolis, IN) with their TargaTips software. Digitized images of various resolutions were displayed for the nephropathologist who determined the minimum resolution required for reviewing renal biopsies. Resolutions of 420 x 380 x 256 colors, 512 x 400 x unlimited colors, 640 x 480 X 196 colors, 640 x 480 X 256 colors, 512 X 496 x unlimited colors, 640 X 496 X innumerable colors, and 1024 x 768 x innumerable colors were evaluated. The first existing lines of communication that were investigated were the state of North Carolina’s X.25 LINCNET. This state-legislature-funded X.25 net interconnected some of the AHECs, most of the state’s two-year colleges, and most universities. AHEC was already utilizing the net; therefore, DCCEC initially selected the existing X.25 net as the basis for developing the GDCN communications. For reasons discussed in RESULTS,this line of communication was rapidly abandoned. Subsequently, normal telephone lines with modem connections were utilized. Two commercial software packages that were currently available and that supported telephone transmission were found: (1) Interact ($495.00) by Applied Communications Concepts, Inc. (Research Triangle Park, NC) and (2) SEND-IT ($695.00) by XENAS (Cincinnati, OH). Although neither product was developed for the medical environment, both vendors modified their software to support use in this area. Both of these modified products were used and are compared and discussed in RESULTS. In order to integrate the system into the daily routine of the nephropathologist and practicing nephrologist, both types of professionals were observed on several different occasions by different observers. Analysis of this observation plus limitations imposed by the physical nature of the equipment influenced design of the
W FIGURE 2. Geographical locations of the major area health education centers (AHEC) in North Carolina.
ANNALS NEW YORK ACADEMY OF SCIENCES
284
1 '1
_-
486/33MHz/TARGA+
I
FIGURE3. Image-originatingor -sending station. An IBM model 90 (486 CPU, 33-MHz, 8-Mb RAM, 320-Mb hard drive, VGA/XGA graphics, SCSI interface) with Targa+ MC board, a NEC 4FG multisync monitor, a SOW CPD-1302 multiscan monitor, and a JVC KYF30U three-chip RGB video camera composed the basic hardware.
software and hardware, placement of the hardware, and software design of the total system. Final hardware design is discussed in RESULTS.After installation of the computer system, integration into daily routine activities was and is continually enhanced by modifying the system according to observer and user assessments. Justifying cost of the system was important to substantiate purchase by the nephrologists who practiced mostly in small groups. The costs and benefits of the two different telephone/modem systems are discussed in RESULTS.Criteria for justifying the system and reasons for expending the initial investment are also presented. Return versus cost of investment is a real consideration to the business of the practicing nephrologists.
RESULTS Image Resolution
FIGURE3 diagrams the hardware of the image-originating or image-sending station. This particular system consisted of capabilities beyond those needed for this project; however, in order to be able to meet future needs that will require greater computer hardware capabilities, the following system was selected: IBM model 90 (486 CPU, 33-MHz clock, 8-Mb RAM, 320-Mb hard drive with 12-msec seek time, VGA/XGA graphics, and SCSI interface), National Enterprise Communications (NEC) 4FG multisync monitor for the VGA, Sony CPD-1302 multiscan monitor for the Targa/video camera output, Targa+ MC digitizer (or other image capture card), and Victor Company of Japan (JVC) professional three-chip RGB camera model KYF30U with 1:13 zoom lens. The basic image-receiving station, which was located at the practicing nephrologist's office, was exactly the same as the sending station except that it lacked the JVC camera and digitizer card. The sending station was used to capture images of the varying resolutions described in METHODS.After viewing these images, the nephropathologist determined that a 640 x 496 x unlimited color with Targa board noninterlaced output or a 640 x 480 x 256 color resolution with
CRONENBERGER el a/.: NEPHROPATHOLOGY CONSULTATION
285
VGA output was required for reviewing the renal biopsies with the practicing nephrologists. This image resolution was also found to be adequate for transmitting black and white electron micrographs and photomicrographs of immunofluorescence microscopy and text (e.g., typed diagnosis). Because the problem dictated setting up a system that was available, DCCEC approached the interactive communications design by utilizing currently available technology and pushing it to its limits. The philosophy was that by the time ISDN or some other more rapid means of communication becomes available, many factors concerning distant clinical consultation would have been learned by having the system in operation. Part of the design to confine costs and to begin immediately involved utilization of currently available lines of communication. AHEC was already using the state-funded X.25 LINCNET between some of the AHECs (FIG. 4), most of the state’s junior colleges, and most of the state’s universities. Because this net was in place and because it was funded by the state, GDCN could use it without cost. FIGURE 4 depicts the planned communications between the nephropathologist and the practicing nephrologist. Beginning with the image-sending station there was a connection to an ethernet (10 Mbs) that was connected to the campus broadband (Mbps). A link (an X.25 gateway) between the campus broadband and the X.25 LINCNET (56 Kbps) was not present; therefore, this project would have to purchase it and devclop appropriate software. Then the transmission would continue over LINCNET to a central receiving unit at an AHEC site. This site had an X.25 gateway to convert back from X.25 to Novel1 ethernet format. Then to get the image files to the practicing nephrologist’s office, several possibilities existed. One was to install a
FIGURE4. Preexisting LINCNET X.25net. Initially a state-funded X.25 net was selected to be the backbone of the GDCN net. The transmitted image would go from the consulting pathologist via an ethernet to a campus broadband through an X.25 gateway to LINCNET through the receiving X.25gateway to the AHEC receiving center.
ANNALS NEW YORK ACADEMY OF SCIENCES
286
multimodem access so that the associated office could access the images via either leased telephone lines (partial T1 of 56 Kbps) or normal telephone lines (19.2 Kbps). DCCEC began writing transmission software around this communications design using encapsulation and unencapsulation between the Transmission Control Protocol/ Internet Protocol (TCP/IP) of the campus ethernet, the AHEC units, and the X.25 of the LINCNET.s5 Initial elation over finding a state-funded net to be used by the GDCN, thereby drastically cutting communication costs plus eliminating any need for establishment of new lines of communication, was dashed when costs of the missing links and of hookup per practitioner plus speeds of transmission between receiving AHEC end and practicing nephrologist were investigated. The only economicallyfeasible method
t SENDER
RECEIVER
FIGURE 5. Final telephone/modem communicationsconfiguration.This configuration traces the line of communication (sending station through modem via normal telephone lines to receiving modem to receiving computer) that best fit this application. Compression and packetized format allowed rapid transmission (10 sec) over normal, nondedicated telephone lines.
to reach each practicing nephrologist was to use regular telephone lines rather than leased lines. If this approach were taken, the bottleneck for speed of transmission from nephropathologist to nephrologist would be the normal telephone line 19.2 Kbps. Therefore, if this was the limiting speed, why go to all the cost and trouble of using the 56-Kbps state-funded LINCNET? Instead, why not just begin with the normal telephone lines and go directly from the pathologist to each individual practitioner (FIG.5). At this time, DCCEC discovered two currently available software packages that supported capture and transmission of images of the desired resolution over normal telephone lines. These packages also provided point-to-point end-user interactivity and annotation during simultaneous viewing of the transmitted images. Both sender
CRONENBERGER el al.: NEPHROPATHOLOCY CONSULTATION
287
and receiver view and annotate the same image and both endsview identical monitor outputs. Interact was initially utilized. This package supported the 640 X 480 image resolution only on the AT platform. Although not currently available for the MC platform, Interact is being revised for IBM XGA graphics and will be available in approximately six months. The available package supported image capture via the Targa+ board. The image was converted and stored in a VGA format (640 x 480 x 256 color) before transmission. Depending upon computer speed, several minutcs were required to convert a Targa (TGA) image to a VGA format. The new VGA image file received a new name in the form of a number. Loss of original file name was unacceptable for the clinical consultation purposes. A big advantage of Interact was that it supported transmission with most brands of modems; however, there was difficulty in obtaining top modem performance over nondedicated telephone lines. Transmission time over normal (nondedicated) telephone lines was 3-4 min per image. With the Hayes Ultra 96 modem, normal telephone lines usually yielded a 2.4-Kbps rate. A 9.6-Kbps rate was never maintained for a series of images. An average biopsy review in the GDCN included approximately 9 to 12 images. This required about 36-48 min of telephone transmission time which was intolerable for this GDCN application. Serial boards with the newer 16550 Universal Asynchronous Receiver Transmitter (UART) chip were tried? No improvement was shown in performance in rate of image transmission on the dedicated computer system that was used. The UART converts parallel bits to serial bits and vice versa and is the hardware register that holds data between the RS232 interface and the processor. Although the faster chip did not improve performance in the system tested here, it might increase performance of communications in computers used in multitasking. The major advantageous feature for Interact was that the receiving station could be a normal 386-MHz computer with VGA to support the 640 x 480 display. There was no need to have an additional video card or Targa board on the receiving end. Another advantage of Interact was the manner in which transmitted batch images were thereafter displayed by the sending station. To display the images in sequence in the order in which they were saved to a folder or batch of images, the sender needed only to “mouse” over an arrow. A few other points of evaluation were worth noting. Minor drawbacks were that during image display the menu across the top of the screen remained on the screen and changed color with image palette changes. Interact does not provide for transmission of a batch of images to an unattended computer; however, once the computers have contacted each other, a batch of images may be transmitted. The user interface was very unfriendly compared to the SEND-IT software package described below. Therefore, even though the receiving station for Interact was attractively configured by being a normal VGA computer, the times required for image manipulation and the other minor drawbacks made use of Interact prohibitive for this project. Interact is currently being revised to shorten transmission time by incorporating compression via software and to correct most of the above disadvantages. The future revision that will incorporate software compression/decompression, as well as capability to run on the MC systems, will make Interact a valuable product. Again, the most attractive feature was that Interact ran on an average computer with VGA capabilities and used almost any brand of modem. No accessory video cards or compression boards were required. SEND-IT was the software package that best met the needs of this project. This package was a most user-friendly program. The menu consisted of nine icons. The
288
ANNALS NEW YORK ACADEMY OF SCIENCES
user only needed to mouse over an icon to perform some activity within the program, for example, capture an image or send an image. SEND-.IT provided image capture via the Targa+ or the Targa+MC card, and ran either on an AT or a MC platform. Images could be transmitted uncompressed or compressed. SEND+IT provided compression/decompression either with hardware (Alice MC board by Telephoto Communications, Inc., San Diego, CA) or with software (also by Telephoto Communications, Inc.). Compression at various ratios with lossy and lossless techniques following the Joint Photographic Experts Group (JPEG) recommendations was applied and evaluated for acceptability. Both Telephoto methods (hardware and software) employ algorithms according to the JPEG recommendations and consist of four steps: 1. Color conversion (RGB-YUV) with color component subsampling to remove color redundancy, 2. Division of image into 8 x 8 blocks and use of discrete cosine transform (DCT) to remove redundancy, 3. Quantization of the DCT coefficients using the psychovisual weighting functions, and 4. Encoding of data for minimum entropy by Huffman variable length method. Telephoto allows the user to select either a direct ratio for compression or a quality mode for compression. Actual compression ratios could be selected from 1:2 to 1:lOO.Image compression quality modes could be selected from (1) perfect, (2) best, (3) better, and (4) good. The resulting actual ratio of compression for each quality mode selection varied according to the complexity of the image. As the complexity increased, the compression ratio decreased. Images (640 x 496) before and after compression were evaluated by the nephropathologist. A compression ratio of 1:32 (JPEG lossy) was determined to have no discernible effect upon image quality. Decreasing compression ratios added nothing to image quality but did lengthen telephone transmission time. The quality mode of “better” compressed between 11:1 and 32:l depending upon complexity of the image. Most of the renal biopsy images were simple and therefore were compressed around 1:32. Because this “better” quality mode guaranteed the same resolution whether an image was of simple or of higher complexity, and because this mode accomplished a compression ratio of 1:32 on the majority of images, it was established as the compression criteria to use. In addition to compression of images, SEND-IT utilized a second feature, packetized ensemble protocol (PEP), on the Trailblazer T2500 modem (Telebit Corporation, Sunnyvale, CA) that speeded transmission over normal telephone lines. The modem assembled packets of data and added a 16-bit cyclic redundancy check (CRC) for error detection. If the receiving modem detected an error, it requested a retransmission. Each time a connection was made in PEP, the modem performed a line analysis and determined the operating parameters of the transmission. Line characteristics were constantly monitored by the modem during the transmission, and operation was adjusted as required for optimal data transmission. The PEP error detection was especially useful in the medical arena where many electrical devices may cause line interference. In the GDCN environments, transmission over normal, nondedicated, telephone lines routinely yielded transmission rates of 18.6 Kbps. Transmission time for a 640 x 496 image with the SEND+IT software varied between 10 and 20 sec depending upon image complexity. This transmission time was with the Telephoto board hardware compression/decompression. Although SEND+IT does support software compression/decompression,the GDCN elected to use the hardware method because of the time saved (about 6-9 sec per image).
CRONENBERGER er al.: NEPHROPATHOLOGY CONSULTATION
289
The average transmission time for the renal biopsy images was a fantastically short 12 sec, made possible by a combination of the Alice board hardware compression/ decompression and the PEP modem. The pathologist transmitted the 9 to 12 images in 90-144 sec. SEND-IT also provided the ability to send batch files to an unattended computer. Thus, the recipient nephrologist only needed to have the computer on and running SEND-IT. At any time thereafter, the sending pathologist could dial up the recipient modem and transmit a batch of images which was ultimately stored on the receiving computer. Then, at a later time convenient to both nephrologist and pathologist, the computers were connected and a live, interactive consultation session ensued. Either participant could point to a feature of the image and the other immediately vicwed it. The pathologist usually controlled which image was displayed from a directory of images displayed as 40 postage-stampsized reproductions per screen. Whenever the pathologist moused over a postage-stamp replica in the directory, that full-sized image was displayed on the monitors of both participants. Thereafter, either participant could draw or type notations on the images, and the other participant would immediately view the additions or changes. Major advantages of SEND-IT included the user-friendly interface, the short time for image transmission, the postage-stamp directory of images, and the batch file transmission to unattended computer. A major disadvantage was the pointer available during consultation which was a small square that was difficult to see and which did not really point to a discrete area. Another point for improvement is the need to refer to the directory or to type in an image name in order to display a new image rather than just mousing on an arrow (as in Interact) to display the next image in the batch file. In order to better meet the medical environment needs, SEND-IT is revising their software according to GDCN user input. Integration of the computer system into daily routine activities was accomplished Several observers visited both the pathologist and the after various forms of nephrologist at varying times during a day and noted normal activities of each. Clinical consultations reviewing renal biopsy results were observed. These noncomputerized sessions based on mailed photographs and telephone conversations lasted about 12-16 min per patient. Consultation was made during a time convenient for both participants. The referring nephrologist previewed the biopsy representative images before the consultation if they had been received through the mail; however, most consultations took place prior to receipt of the photographs by the nephrologists. Consultation occurred from the private offices of both participants. Based upon the above observations, a computer system was set up in the office of the pathologist so that it was convenient to digitize images during normal biopsy evaluation. The camera was mounted atop the microscope. Biopsy areas demonstrating particular features for future consultation were easily and rapidly digitized and stored on a hard disk. Image capture took about 2-4 sec including storage on hard drive. A most important factor for integrating the computer system into the natural routine of the pathologist and nephrologist was to locate it in physical spaces that they normally occupy. It soon bccame evident that another requirement for the system was to contain the physical size of the hardware. Size of hardware had to be small enough to fit into an office without inordinately adding bulky physical structures that would interfere with normal work flow. The original two-monitor system was changed to a onemonitor systcm. During observations and from previous experience with physician workstations, a wait period of inactivity for longer than 5 sec was found to be longer than practitioners would endure. Therefore, hardware and software for the actual consul-
290
ANNALS NEW YORK ACADEMY OF SCIENCES
tation were designed to limit wait periods to a maximum of 5 sec. This was accomplished by the pathologist continuing to discuss an image on the recipient’s screen while proceeding to view the image directory and proceeding to bring up a new image on the sending screen. The recipient viewed a continuous flow of images; however, the pathologist had to be versatile and comfortable with the computer manipulation. Although SEND-+IT supported switching from visual communication to verbal communication, both pathologist and nephrologist insisted on concomitant verbal and visual communication. Conference telephones were installed and provided this feature. Cost justification was proved in several aspects. First, because of the decrease in time between biopsy and consultation, patient treatment was begun sooner. Second, community-practice nephrologists also tended to perform their own biopsies rather than to send the patient to a regional medical center for a biopsy. Thus, cost to patient for travel to a distant medical center was reduced. In addition to consultation, other applications have been set up on the receiving station. Examples are word processing and medical search databases (MEDLARS). This allows the computer to function in other roles and adds to its utility. The computer could be used for any application when it is not in session for consultation. Most advantageous in regard to cost justification was the receiving of continuing education (CE) presentations within the office of nephrologists using the image communication system. Every nephrologist must substantiate a certain number of hours of CE per year, and the computer can be used to receive CE presentations from the GDCN. Average costs of $lo00 or more per meeting for CE can be applied toward the purchase of the computer. It should be mentioned that an initial effort was made to use cheaper, IBMcompatible computers. This proved to be a frustrating experience because of the many hardware breakdowns encountered. Hardware breakdowns included keyboard failure, drive failure, drive-card failure, and motherboard failure. Another cost factor of importance was the use of hardware compression versus cheaper software compression. Again, speed was essential in the busy routine of pathologists and nephrologists. The experience with this net design proved that a higher cost for computer hardware and software was justifiable if it provided a reliable, rapid and user-friendly system which was essential in this medical consulting application. SUMMARY AND CONCLUSIONS
Investigations into a digitized image communications system were prompted by a need to bring expert consultation to physicians in community practice. Pathologists desired the capability to concomitantly view, annotate, and discuss images with referring physicians at distant sites. Methods included evaluation of the human and procedural domain into which the system was to be integrated. The GDCN computer consultation system has the consultant nephropathologist first evaluate the processed biopsy slides, digitize representative images, transmit them with the diagnosis to referring nephrologist, and, finally, conduct an interactive consultation and review of the biopsy and case. Image resolution and compression variables must be set for each individual medical consulting application. For the GDCN, it was found that the 640 x 496 x unlimited color with compression ratios not exceeding 1:32 are acceptable. An obvious improvement of this computerized system over the noncomputerized review sessions is the ability to immediately share and discuss a new image that had not been previously sent. In the old noncomputerized consultation, only images that had been mailed could be discussed. The computerized sessions allow transmission
CRONENBERGER el
01.:
NEPHROPATHOLOGY CONSULTATION
291
(10 sec) of a new image that the consultation might demand. The computerized sessions also provide the ability to show the referring nephrologist an area of biopsy interest that the pathologist had not previously transmitted. Biopsy slides can be viewed during the consultation, an area digitized, and that image transmitted to the nephrologist during the consultation. Hardware and costs for the sending station were: IBM model 90 33-MHz, 320-Mb hard drive, 8-Mb RAM, SCSI interface NEC 4FG multisync monitor SONY C21 monitor XENAS communications package SEND-IT software, Targa+MC, Alice card, modem JVC 3 chip video camera with lens Total
$7,400 850 500 4,650 4,600 $18,000
This system far exceeds the requirements for this particular application; however, it is sufficient to support future, higher-technology computer applications. If necessary, this same system could be used with a less expensive computer, a less expensive camera, software compression, and a single monitor. These alterations could lessen the expenditures by some $8000 and result in a total cost of $lO,OOO. Hardware and costs for the receiving station were: IBM model 90 33-MHz, 320-Mb hard drive, 8-Mb RAM SCSI interface NEC 4FG multisync monitor SONY C21 monitor XENAS communications package SEND-IT software, Targa+ MC, Alice compression Total
$7,400 850 500 3,650 $12,400
Cost of the receiving station could be reduced by using a less expensive computer and a single monitor system, thereby saving up to $5000 and resulting in a total cost of $7,400. DCCEC and GDCN have elected to use the more expensive, user-friendly and more rapid image transmission system SEND-IT, rather than the less expensive system mainly because of experience with incorporating the system into the daily activities of the GDCN. SEND-IT best met the essentials for GDCN. Both systems had their pros and cons, and we are continuing to collaborate with both companies. After beginning to write a new communications software package, it was proved that certain existing software packages could provide a good starting point. These commercial vcndors were willing to work together with DCCEC to develop a package for the medical environment. In return, DCCEC was their partner in providing user critique for continued development. The venture with both companies is a cooperative one and is essential for continued support of any software during this rapidly changing communications technology.'"There are several areas of continued development for this system. Most immediately, a need exists for the consultation and continuing education sessions to simultaneously support recipients located at multipoints or more than one location. This capability is being developed and is essential for the GDCN continuing education presentations to be economically feasible. A need is always present to decrease time for communication of images and to increase image resolution. These aspects are most important when transforming the
292
ANNALS NEW YORK ACADEMY OF SCIENCES
system into a receive image and diagnosing solely via image situation. BMECRF is already conducting research with Integrated Service Digital Network (ISDN) and Fiberoptic Data Distributed Interface (FDDI) communications. Although these are in effect between some major institutions of the state, it will be some years before either will be economically feasible for use by GDCN. In the interim, it is most likely that the GDCN could switch over to Accupulse (56 Kbps) when it becomes cost efficient or when the current charges are reduced. A need to reduce cost of the total system always exists. This is happening with daily reductions in cost of computers and other hardware products that this system incorporates. The GDCN computer communications have been in operation for about six months. Feedback from participants is very enthusiastic. Spin-off advantages of the system are several. Nephrologists feel more in touch with the consulting pathologist. Some nephrologists have shared the digitized images with their patients who appreciated the opportunity to visualize their disease. Receiving continuing education in their office during their lunch hour has been one of the bigger advantages of this system. Easy access to medical databases and other information via the computer communications certainly will enhance the likelihood that the isolated nephrologist will use such facilities. This system, therefore, encourages rapid dissemination of medical data. Most of all, the GDCN computer communications promise a method of more effective consultation between medical specialists and the resultant improvement in patient care. REFERENCES 1. JENNEITE, J. C. 1991. Anti-neutrophil cytoplasmic autoantibody-associated disease. Am. J. Kidney Dis. 18: 164-170. 2. FALK,R. J., S. HOGAN,T. S. CAREY & J. C. JENNETIT.1990.The clinical course of patients with antineutrophil cytoplasmic autoantibody associated glomerulonephritis and systemic vasculitis. Ann. Intern. Med. 113: 656-663. 3. BLACK,U. 1991. The X Series Protocols for Data Communications Networks. McGrawHill, Inc. New York, NY. 4. COMER,D. E. 1991. Internetworking with TCP/IP. 2nd edit., Vol. 1. Prentice Hall. Englewood Cliffs, NJ. A. S. 1988. Computer Networks. 2nd edit. Prentice Hall. Englewood Cliffs, 5. TANENBAUM, NJ. 6. BLACK,U. 1991. The V Series Recommendations Protocols for Data Communications over the Telephone Network. McGraw-Hill, Inc. New York, NY. EDS. 1990. Medical Informatics. Addison-Wesley. E. H. & L. PERREAULT, 7. SHORTLIFFE, New York, NY. 8. SHULTZ,E. K. & R. W. BROWN. 1991. The pathologist’s workstation. Am. J. Clin. Pathol. 95(4): s50457. 9. MCNEELY,M. D. D. 1991. Advances in medical informatics. Am. J. Clin. Pathol. 96:s33439. 1992. An analysis of the relationship between a pathology B. & W. MITCHELL. 10. FRIEDMAN, department and its laboratory information system vendor. Am. J. Clin. Pathol. 97: 363-368.
The Glomerular Disease Collaborative Network (GDCN) evolved in 1985 from a renal biopsy diagnostic service. GDCN was established to provide renal biopsy diagnostic service so that patients with renal disease could remain in their home communities and be treated by their own local physician. From a beginning with 15 associated nephrologists, GDCN has now grown to include more than 150 nephrologists practicing at approximately 53 sites. Most are located in North Carolina, Florida, Georgia, South Carolina, Tennessee, and Virginia, with a few in Oregon, Washington, Minnesota, and Ohio (FIG.1). For the past two years, GDCN has been networking via fax machines which are used to accrue patient data dealing primarily with pathologic phenotypes, clinical manifestations, and treatment responses of renal diseases.',? The referring nephrologists performed a renal biopsy at their local hospital and then mailed the specimen to GDCN Nephropathology Laboratory in Chapel Hill. The biopsy material was processed for light, immunofluorescence, and electron microscopy using standard procedures. The processed biopsy was evaluated by a nephropathologist who took representative photomicrographs and electronmicrographs. A diagnosis based upon biopsy evaluation plus clinical information supplied by the referring nephrologist was rendered. Representative photographs along with typed evaluation and diagnosis were mailed to the referring nephrologist. The routine of processing and evaluating the biopsy specimen and mailing photographs and reports to the referring nephrologist took an average of one week. THE PROBLEM The problem undertaken was to decrease the one-week time period required for the nephrologists to receive micrographs and reports, and to enhance the pathologistnephrologist exchange by developing computer technology for interactive clinical consultation that incorporated images and voice. The consultant nephropathologist and the referring nephrologist needed to be able to simultaneously view an image of light, immunofluorescence or electron microscopy results, to point to areas of the
"This project was supported by a grant from The Kate B. Reynolds Health Care Trust, Winston-Salem, North Carolina. Computers for this project were provided by the IBM Corporation, Raleigh, North Carolina. 'Address correspondence to Dr. J. H. Cronenberger, Department of Biomedical Engineering, CB No. 7575, University of North Carolina Medical School at Chapel Hill, Chapel Hill, NC 27599-7575. e-mail: [email protected] 28 1
ANNALS NEW YORK ACADEMY OF SCIENCES
282
THE GLOMERULAR DISEASE COLLABORATIVE NETWORK
B.
* ,
UNC - Central Office $r Academic Medical Centers 0 Private Practices
FIGURE 1. Geographical sites of the Glomerular Disease Collaborative Network. More than 150 nephrologists practicing at 53 sites in ten states form the network for which interactive computer communications providing image, annotation, and verbal exchange were developed.
image and have the other immediately see the pointer, and to talk with one another while viewing and discussing the images and case history. Initial constraints applied to the solution of the problem were that the system (1) must provide image resolution to allow review of biopsy tissue, (2) must be currently available, (3) must utilize existing lines of communication, (4) must easily and naturally integrate into the daily routine of the nephropathologist and nephrologist, and ( 5 ) must be economicallyjustifiable from the viewpoint of the practitioner. METHODS Development of the computer communications and resolution of the stated problem was a collaborative effort of GDCN and the Department of Biomedical Engineering’s Communications Research Facility (BMECRF). BMECRF already had a Distance Clinical Consultation and Education Communications (DCCEC) project in progress with the Area Health Education Centers (AHEC). This DCCEC project was already involved in developing the communications system that will be described in this paper. Both GDCN and AHEC required similar technology. Solution of the stated problem was, therefore, related to the North Carolina AHEC which comprises a series of nine establishments around the state (FIG.2). These nine units are directed from a central office at the University of North Carolina Medical School. AHEC was interested in retention of practicing health professionals in the rural areas of North Carolina. On-site accessibility to consultation and continuing education by the rural health care providers has been found to be a major factor in retention. The interactive computer communicationssystem described in this paper is, therefore, designed and developed with the applications of both GDCN and
CRONENBERGER el al.: NEPHROPATHOLOCY CONSULTATION
283
AHEC in mind, even though discussion and description will be of the GDCN application. Image Resolutiod
Several digitizing systems were used to capture images of varying horizontal and vertical resolutions. One was the International Business Machine (IBM) video capture board with their Audio Video Connection (AVC) software. A second was the FG-100-AT board (Imaging Technology Incorporated, Woburn, MA) with Imagepro software (Media Cybernetics, Silver Spring, MD). A third was Targa+ MCA board (Truevision, Indianapolis, IN) with their TargaTips software. Digitized images of various resolutions were displayed for the nephropathologist who determined the minimum resolution required for reviewing renal biopsies. Resolutions of 420 x 380 x 256 colors, 512 x 400 x unlimited colors, 640 x 480 X 196 colors, 640 x 480 X 256 colors, 512 X 496 x unlimited colors, 640 X 496 X innumerable colors, and 1024 x 768 x innumerable colors were evaluated. The first existing lines of communication that were investigated were the state of North Carolina’s X.25 LINCNET. This state-legislature-funded X.25 net interconnected some of the AHECs, most of the state’s two-year colleges, and most universities. AHEC was already utilizing the net; therefore, DCCEC initially selected the existing X.25 net as the basis for developing the GDCN communications. For reasons discussed in RESULTS,this line of communication was rapidly abandoned. Subsequently, normal telephone lines with modem connections were utilized. Two commercial software packages that were currently available and that supported telephone transmission were found: (1) Interact ($495.00) by Applied Communications Concepts, Inc. (Research Triangle Park, NC) and (2) SEND-IT ($695.00) by XENAS (Cincinnati, OH). Although neither product was developed for the medical environment, both vendors modified their software to support use in this area. Both of these modified products were used and are compared and discussed in RESULTS. In order to integrate the system into the daily routine of the nephropathologist and practicing nephrologist, both types of professionals were observed on several different occasions by different observers. Analysis of this observation plus limitations imposed by the physical nature of the equipment influenced design of the
W FIGURE 2. Geographical locations of the major area health education centers (AHEC) in North Carolina.
ANNALS NEW YORK ACADEMY OF SCIENCES
284
1 '1
_-
486/33MHz/TARGA+
I
FIGURE3. Image-originatingor -sending station. An IBM model 90 (486 CPU, 33-MHz, 8-Mb RAM, 320-Mb hard drive, VGA/XGA graphics, SCSI interface) with Targa+ MC board, a NEC 4FG multisync monitor, a SOW CPD-1302 multiscan monitor, and a JVC KYF30U three-chip RGB video camera composed the basic hardware.
software and hardware, placement of the hardware, and software design of the total system. Final hardware design is discussed in RESULTS.After installation of the computer system, integration into daily routine activities was and is continually enhanced by modifying the system according to observer and user assessments. Justifying cost of the system was important to substantiate purchase by the nephrologists who practiced mostly in small groups. The costs and benefits of the two different telephone/modem systems are discussed in RESULTS.Criteria for justifying the system and reasons for expending the initial investment are also presented. Return versus cost of investment is a real consideration to the business of the practicing nephrologists.
RESULTS Image Resolution
FIGURE3 diagrams the hardware of the image-originating or image-sending station. This particular system consisted of capabilities beyond those needed for this project; however, in order to be able to meet future needs that will require greater computer hardware capabilities, the following system was selected: IBM model 90 (486 CPU, 33-MHz clock, 8-Mb RAM, 320-Mb hard drive with 12-msec seek time, VGA/XGA graphics, and SCSI interface), National Enterprise Communications (NEC) 4FG multisync monitor for the VGA, Sony CPD-1302 multiscan monitor for the Targa/video camera output, Targa+ MC digitizer (or other image capture card), and Victor Company of Japan (JVC) professional three-chip RGB camera model KYF30U with 1:13 zoom lens. The basic image-receiving station, which was located at the practicing nephrologist's office, was exactly the same as the sending station except that it lacked the JVC camera and digitizer card. The sending station was used to capture images of the varying resolutions described in METHODS.After viewing these images, the nephropathologist determined that a 640 x 496 x unlimited color with Targa board noninterlaced output or a 640 x 480 x 256 color resolution with
CRONENBERGER el a/.: NEPHROPATHOLOGY CONSULTATION
285
VGA output was required for reviewing the renal biopsies with the practicing nephrologists. This image resolution was also found to be adequate for transmitting black and white electron micrographs and photomicrographs of immunofluorescence microscopy and text (e.g., typed diagnosis). Because the problem dictated setting up a system that was available, DCCEC approached the interactive communications design by utilizing currently available technology and pushing it to its limits. The philosophy was that by the time ISDN or some other more rapid means of communication becomes available, many factors concerning distant clinical consultation would have been learned by having the system in operation. Part of the design to confine costs and to begin immediately involved utilization of currently available lines of communication. AHEC was already using the state-funded X.25 LINCNET between some of the AHECs (FIG. 4), most of the state’s junior colleges, and most of the state’s universities. Because this net was in place and because it was funded by the state, GDCN could use it without cost. FIGURE 4 depicts the planned communications between the nephropathologist and the practicing nephrologist. Beginning with the image-sending station there was a connection to an ethernet (10 Mbs) that was connected to the campus broadband (Mbps). A link (an X.25 gateway) between the campus broadband and the X.25 LINCNET (56 Kbps) was not present; therefore, this project would have to purchase it and devclop appropriate software. Then the transmission would continue over LINCNET to a central receiving unit at an AHEC site. This site had an X.25 gateway to convert back from X.25 to Novel1 ethernet format. Then to get the image files to the practicing nephrologist’s office, several possibilities existed. One was to install a
FIGURE4. Preexisting LINCNET X.25net. Initially a state-funded X.25 net was selected to be the backbone of the GDCN net. The transmitted image would go from the consulting pathologist via an ethernet to a campus broadband through an X.25 gateway to LINCNET through the receiving X.25gateway to the AHEC receiving center.
ANNALS NEW YORK ACADEMY OF SCIENCES
286
multimodem access so that the associated office could access the images via either leased telephone lines (partial T1 of 56 Kbps) or normal telephone lines (19.2 Kbps). DCCEC began writing transmission software around this communications design using encapsulation and unencapsulation between the Transmission Control Protocol/ Internet Protocol (TCP/IP) of the campus ethernet, the AHEC units, and the X.25 of the LINCNET.s5 Initial elation over finding a state-funded net to be used by the GDCN, thereby drastically cutting communication costs plus eliminating any need for establishment of new lines of communication, was dashed when costs of the missing links and of hookup per practitioner plus speeds of transmission between receiving AHEC end and practicing nephrologist were investigated. The only economicallyfeasible method
t SENDER
RECEIVER
FIGURE 5. Final telephone/modem communicationsconfiguration.This configuration traces the line of communication (sending station through modem via normal telephone lines to receiving modem to receiving computer) that best fit this application. Compression and packetized format allowed rapid transmission (10 sec) over normal, nondedicated telephone lines.
to reach each practicing nephrologist was to use regular telephone lines rather than leased lines. If this approach were taken, the bottleneck for speed of transmission from nephropathologist to nephrologist would be the normal telephone line 19.2 Kbps. Therefore, if this was the limiting speed, why go to all the cost and trouble of using the 56-Kbps state-funded LINCNET? Instead, why not just begin with the normal telephone lines and go directly from the pathologist to each individual practitioner (FIG.5). At this time, DCCEC discovered two currently available software packages that supported capture and transmission of images of the desired resolution over normal telephone lines. These packages also provided point-to-point end-user interactivity and annotation during simultaneous viewing of the transmitted images. Both sender
CRONENBERGER el al.: NEPHROPATHOLOCY CONSULTATION
287
and receiver view and annotate the same image and both endsview identical monitor outputs. Interact was initially utilized. This package supported the 640 X 480 image resolution only on the AT platform. Although not currently available for the MC platform, Interact is being revised for IBM XGA graphics and will be available in approximately six months. The available package supported image capture via the Targa+ board. The image was converted and stored in a VGA format (640 x 480 x 256 color) before transmission. Depending upon computer speed, several minutcs were required to convert a Targa (TGA) image to a VGA format. The new VGA image file received a new name in the form of a number. Loss of original file name was unacceptable for the clinical consultation purposes. A big advantage of Interact was that it supported transmission with most brands of modems; however, there was difficulty in obtaining top modem performance over nondedicated telephone lines. Transmission time over normal (nondedicated) telephone lines was 3-4 min per image. With the Hayes Ultra 96 modem, normal telephone lines usually yielded a 2.4-Kbps rate. A 9.6-Kbps rate was never maintained for a series of images. An average biopsy review in the GDCN included approximately 9 to 12 images. This required about 36-48 min of telephone transmission time which was intolerable for this GDCN application. Serial boards with the newer 16550 Universal Asynchronous Receiver Transmitter (UART) chip were tried? No improvement was shown in performance in rate of image transmission on the dedicated computer system that was used. The UART converts parallel bits to serial bits and vice versa and is the hardware register that holds data between the RS232 interface and the processor. Although the faster chip did not improve performance in the system tested here, it might increase performance of communications in computers used in multitasking. The major advantageous feature for Interact was that the receiving station could be a normal 386-MHz computer with VGA to support the 640 x 480 display. There was no need to have an additional video card or Targa board on the receiving end. Another advantage of Interact was the manner in which transmitted batch images were thereafter displayed by the sending station. To display the images in sequence in the order in which they were saved to a folder or batch of images, the sender needed only to “mouse” over an arrow. A few other points of evaluation were worth noting. Minor drawbacks were that during image display the menu across the top of the screen remained on the screen and changed color with image palette changes. Interact does not provide for transmission of a batch of images to an unattended computer; however, once the computers have contacted each other, a batch of images may be transmitted. The user interface was very unfriendly compared to the SEND-IT software package described below. Therefore, even though the receiving station for Interact was attractively configured by being a normal VGA computer, the times required for image manipulation and the other minor drawbacks made use of Interact prohibitive for this project. Interact is currently being revised to shorten transmission time by incorporating compression via software and to correct most of the above disadvantages. The future revision that will incorporate software compression/decompression, as well as capability to run on the MC systems, will make Interact a valuable product. Again, the most attractive feature was that Interact ran on an average computer with VGA capabilities and used almost any brand of modem. No accessory video cards or compression boards were required. SEND-IT was the software package that best met the needs of this project. This package was a most user-friendly program. The menu consisted of nine icons. The
288
ANNALS NEW YORK ACADEMY OF SCIENCES
user only needed to mouse over an icon to perform some activity within the program, for example, capture an image or send an image. SEND-.IT provided image capture via the Targa+ or the Targa+MC card, and ran either on an AT or a MC platform. Images could be transmitted uncompressed or compressed. SEND+IT provided compression/decompression either with hardware (Alice MC board by Telephoto Communications, Inc., San Diego, CA) or with software (also by Telephoto Communications, Inc.). Compression at various ratios with lossy and lossless techniques following the Joint Photographic Experts Group (JPEG) recommendations was applied and evaluated for acceptability. Both Telephoto methods (hardware and software) employ algorithms according to the JPEG recommendations and consist of four steps: 1. Color conversion (RGB-YUV) with color component subsampling to remove color redundancy, 2. Division of image into 8 x 8 blocks and use of discrete cosine transform (DCT) to remove redundancy, 3. Quantization of the DCT coefficients using the psychovisual weighting functions, and 4. Encoding of data for minimum entropy by Huffman variable length method. Telephoto allows the user to select either a direct ratio for compression or a quality mode for compression. Actual compression ratios could be selected from 1:2 to 1:lOO.Image compression quality modes could be selected from (1) perfect, (2) best, (3) better, and (4) good. The resulting actual ratio of compression for each quality mode selection varied according to the complexity of the image. As the complexity increased, the compression ratio decreased. Images (640 x 496) before and after compression were evaluated by the nephropathologist. A compression ratio of 1:32 (JPEG lossy) was determined to have no discernible effect upon image quality. Decreasing compression ratios added nothing to image quality but did lengthen telephone transmission time. The quality mode of “better” compressed between 11:1 and 32:l depending upon complexity of the image. Most of the renal biopsy images were simple and therefore were compressed around 1:32. Because this “better” quality mode guaranteed the same resolution whether an image was of simple or of higher complexity, and because this mode accomplished a compression ratio of 1:32 on the majority of images, it was established as the compression criteria to use. In addition to compression of images, SEND-IT utilized a second feature, packetized ensemble protocol (PEP), on the Trailblazer T2500 modem (Telebit Corporation, Sunnyvale, CA) that speeded transmission over normal telephone lines. The modem assembled packets of data and added a 16-bit cyclic redundancy check (CRC) for error detection. If the receiving modem detected an error, it requested a retransmission. Each time a connection was made in PEP, the modem performed a line analysis and determined the operating parameters of the transmission. Line characteristics were constantly monitored by the modem during the transmission, and operation was adjusted as required for optimal data transmission. The PEP error detection was especially useful in the medical arena where many electrical devices may cause line interference. In the GDCN environments, transmission over normal, nondedicated, telephone lines routinely yielded transmission rates of 18.6 Kbps. Transmission time for a 640 x 496 image with the SEND+IT software varied between 10 and 20 sec depending upon image complexity. This transmission time was with the Telephoto board hardware compression/decompression. Although SEND+IT does support software compression/decompression,the GDCN elected to use the hardware method because of the time saved (about 6-9 sec per image).
CRONENBERGER er al.: NEPHROPATHOLOGY CONSULTATION
289
The average transmission time for the renal biopsy images was a fantastically short 12 sec, made possible by a combination of the Alice board hardware compression/ decompression and the PEP modem. The pathologist transmitted the 9 to 12 images in 90-144 sec. SEND-IT also provided the ability to send batch files to an unattended computer. Thus, the recipient nephrologist only needed to have the computer on and running SEND-IT. At any time thereafter, the sending pathologist could dial up the recipient modem and transmit a batch of images which was ultimately stored on the receiving computer. Then, at a later time convenient to both nephrologist and pathologist, the computers were connected and a live, interactive consultation session ensued. Either participant could point to a feature of the image and the other immediately vicwed it. The pathologist usually controlled which image was displayed from a directory of images displayed as 40 postage-stampsized reproductions per screen. Whenever the pathologist moused over a postage-stamp replica in the directory, that full-sized image was displayed on the monitors of both participants. Thereafter, either participant could draw or type notations on the images, and the other participant would immediately view the additions or changes. Major advantages of SEND-IT included the user-friendly interface, the short time for image transmission, the postage-stamp directory of images, and the batch file transmission to unattended computer. A major disadvantage was the pointer available during consultation which was a small square that was difficult to see and which did not really point to a discrete area. Another point for improvement is the need to refer to the directory or to type in an image name in order to display a new image rather than just mousing on an arrow (as in Interact) to display the next image in the batch file. In order to better meet the medical environment needs, SEND-IT is revising their software according to GDCN user input. Integration of the computer system into daily routine activities was accomplished Several observers visited both the pathologist and the after various forms of nephrologist at varying times during a day and noted normal activities of each. Clinical consultations reviewing renal biopsy results were observed. These noncomputerized sessions based on mailed photographs and telephone conversations lasted about 12-16 min per patient. Consultation was made during a time convenient for both participants. The referring nephrologist previewed the biopsy representative images before the consultation if they had been received through the mail; however, most consultations took place prior to receipt of the photographs by the nephrologists. Consultation occurred from the private offices of both participants. Based upon the above observations, a computer system was set up in the office of the pathologist so that it was convenient to digitize images during normal biopsy evaluation. The camera was mounted atop the microscope. Biopsy areas demonstrating particular features for future consultation were easily and rapidly digitized and stored on a hard disk. Image capture took about 2-4 sec including storage on hard drive. A most important factor for integrating the computer system into the natural routine of the pathologist and nephrologist was to locate it in physical spaces that they normally occupy. It soon bccame evident that another requirement for the system was to contain the physical size of the hardware. Size of hardware had to be small enough to fit into an office without inordinately adding bulky physical structures that would interfere with normal work flow. The original two-monitor system was changed to a onemonitor systcm. During observations and from previous experience with physician workstations, a wait period of inactivity for longer than 5 sec was found to be longer than practitioners would endure. Therefore, hardware and software for the actual consul-
290
ANNALS NEW YORK ACADEMY OF SCIENCES
tation were designed to limit wait periods to a maximum of 5 sec. This was accomplished by the pathologist continuing to discuss an image on the recipient’s screen while proceeding to view the image directory and proceeding to bring up a new image on the sending screen. The recipient viewed a continuous flow of images; however, the pathologist had to be versatile and comfortable with the computer manipulation. Although SEND-+IT supported switching from visual communication to verbal communication, both pathologist and nephrologist insisted on concomitant verbal and visual communication. Conference telephones were installed and provided this feature. Cost justification was proved in several aspects. First, because of the decrease in time between biopsy and consultation, patient treatment was begun sooner. Second, community-practice nephrologists also tended to perform their own biopsies rather than to send the patient to a regional medical center for a biopsy. Thus, cost to patient for travel to a distant medical center was reduced. In addition to consultation, other applications have been set up on the receiving station. Examples are word processing and medical search databases (MEDLARS). This allows the computer to function in other roles and adds to its utility. The computer could be used for any application when it is not in session for consultation. Most advantageous in regard to cost justification was the receiving of continuing education (CE) presentations within the office of nephrologists using the image communication system. Every nephrologist must substantiate a certain number of hours of CE per year, and the computer can be used to receive CE presentations from the GDCN. Average costs of $lo00 or more per meeting for CE can be applied toward the purchase of the computer. It should be mentioned that an initial effort was made to use cheaper, IBMcompatible computers. This proved to be a frustrating experience because of the many hardware breakdowns encountered. Hardware breakdowns included keyboard failure, drive failure, drive-card failure, and motherboard failure. Another cost factor of importance was the use of hardware compression versus cheaper software compression. Again, speed was essential in the busy routine of pathologists and nephrologists. The experience with this net design proved that a higher cost for computer hardware and software was justifiable if it provided a reliable, rapid and user-friendly system which was essential in this medical consulting application. SUMMARY AND CONCLUSIONS
Investigations into a digitized image communications system were prompted by a need to bring expert consultation to physicians in community practice. Pathologists desired the capability to concomitantly view, annotate, and discuss images with referring physicians at distant sites. Methods included evaluation of the human and procedural domain into which the system was to be integrated. The GDCN computer consultation system has the consultant nephropathologist first evaluate the processed biopsy slides, digitize representative images, transmit them with the diagnosis to referring nephrologist, and, finally, conduct an interactive consultation and review of the biopsy and case. Image resolution and compression variables must be set for each individual medical consulting application. For the GDCN, it was found that the 640 x 496 x unlimited color with compression ratios not exceeding 1:32 are acceptable. An obvious improvement of this computerized system over the noncomputerized review sessions is the ability to immediately share and discuss a new image that had not been previously sent. In the old noncomputerized consultation, only images that had been mailed could be discussed. The computerized sessions allow transmission
CRONENBERGER el
01.:
NEPHROPATHOLOGY CONSULTATION
291
(10 sec) of a new image that the consultation might demand. The computerized sessions also provide the ability to show the referring nephrologist an area of biopsy interest that the pathologist had not previously transmitted. Biopsy slides can be viewed during the consultation, an area digitized, and that image transmitted to the nephrologist during the consultation. Hardware and costs for the sending station were: IBM model 90 33-MHz, 320-Mb hard drive, 8-Mb RAM, SCSI interface NEC 4FG multisync monitor SONY C21 monitor XENAS communications package SEND-IT software, Targa+MC, Alice card, modem JVC 3 chip video camera with lens Total
$7,400 850 500 4,650 4,600 $18,000
This system far exceeds the requirements for this particular application; however, it is sufficient to support future, higher-technology computer applications. If necessary, this same system could be used with a less expensive computer, a less expensive camera, software compression, and a single monitor. These alterations could lessen the expenditures by some $8000 and result in a total cost of $lO,OOO. Hardware and costs for the receiving station were: IBM model 90 33-MHz, 320-Mb hard drive, 8-Mb RAM SCSI interface NEC 4FG multisync monitor SONY C21 monitor XENAS communications package SEND-IT software, Targa+ MC, Alice compression Total
$7,400 850 500 3,650 $12,400
Cost of the receiving station could be reduced by using a less expensive computer and a single monitor system, thereby saving up to $5000 and resulting in a total cost of $7,400. DCCEC and GDCN have elected to use the more expensive, user-friendly and more rapid image transmission system SEND-IT, rather than the less expensive system mainly because of experience with incorporating the system into the daily activities of the GDCN. SEND-IT best met the essentials for GDCN. Both systems had their pros and cons, and we are continuing to collaborate with both companies. After beginning to write a new communications software package, it was proved that certain existing software packages could provide a good starting point. These commercial vcndors were willing to work together with DCCEC to develop a package for the medical environment. In return, DCCEC was their partner in providing user critique for continued development. The venture with both companies is a cooperative one and is essential for continued support of any software during this rapidly changing communications technology.'"There are several areas of continued development for this system. Most immediately, a need exists for the consultation and continuing education sessions to simultaneously support recipients located at multipoints or more than one location. This capability is being developed and is essential for the GDCN continuing education presentations to be economically feasible. A need is always present to decrease time for communication of images and to increase image resolution. These aspects are most important when transforming the
292
ANNALS NEW YORK ACADEMY OF SCIENCES
system into a receive image and diagnosing solely via image situation. BMECRF is already conducting research with Integrated Service Digital Network (ISDN) and Fiberoptic Data Distributed Interface (FDDI) communications. Although these are in effect between some major institutions of the state, it will be some years before either will be economically feasible for use by GDCN. In the interim, it is most likely that the GDCN could switch over to Accupulse (56 Kbps) when it becomes cost efficient or when the current charges are reduced. A need to reduce cost of the total system always exists. This is happening with daily reductions in cost of computers and other hardware products that this system incorporates. The GDCN computer communications have been in operation for about six months. Feedback from participants is very enthusiastic. Spin-off advantages of the system are several. Nephrologists feel more in touch with the consulting pathologist. Some nephrologists have shared the digitized images with their patients who appreciated the opportunity to visualize their disease. Receiving continuing education in their office during their lunch hour has been one of the bigger advantages of this system. Easy access to medical databases and other information via the computer communications certainly will enhance the likelihood that the isolated nephrologist will use such facilities. This system, therefore, encourages rapid dissemination of medical data. Most of all, the GDCN computer communications promise a method of more effective consultation between medical specialists and the resultant improvement in patient care. REFERENCES 1. JENNEITE, J. C. 1991. Anti-neutrophil cytoplasmic autoantibody-associated disease. Am. J. Kidney Dis. 18: 164-170. 2. FALK,R. J., S. HOGAN,T. S. CAREY & J. C. JENNETIT.1990.The clinical course of patients with antineutrophil cytoplasmic autoantibody associated glomerulonephritis and systemic vasculitis. Ann. Intern. Med. 113: 656-663. 3. BLACK,U. 1991. The X Series Protocols for Data Communications Networks. McGrawHill, Inc. New York, NY. 4. COMER,D. E. 1991. Internetworking with TCP/IP. 2nd edit., Vol. 1. Prentice Hall. Englewood Cliffs, NJ. A. S. 1988. Computer Networks. 2nd edit. Prentice Hall. Englewood Cliffs, 5. TANENBAUM, NJ. 6. BLACK,U. 1991. The V Series Recommendations Protocols for Data Communications over the Telephone Network. McGraw-Hill, Inc. New York, NY. EDS. 1990. Medical Informatics. Addison-Wesley. E. H. & L. PERREAULT, 7. SHORTLIFFE, New York, NY. 8. SHULTZ,E. K. & R. W. BROWN. 1991. The pathologist’s workstation. Am. J. Clin. Pathol. 95(4): s50457. 9. MCNEELY,M. D. D. 1991. Advances in medical informatics. Am. J. Clin. Pathol. 96:s33439. 1992. An analysis of the relationship between a pathology B. & W. MITCHELL. 10. FRIEDMAN, department and its laboratory information system vendor. Am. J. Clin. Pathol. 97: 363-368.
