infoRAD
`
`1495
`
`A New Approach to
`Teleconferencing with
`Intravascular US and
`Cardiac Angiography
`in a Low-Bandwidth
`Environment1
`
`Johannes N. Stahl, MD • Jianguo Zhang, PhD • Chris Zellner, MD
`Eugene V. Pomerantsev, MD, PhD • Tony N. Chou, MD • H. K. Huang,
`DSc
`
`A common problem in radiology teleconferencing is the difficulty of trans-
`mitting a large volume of data over communication channels with a relatively
`low bandwidth. Although videoconferencing systems are easily implemented,
`they generally require lossy image compression, which can lead to signifi-
`cantly altered findings. A teleconsultation and teleconferencing system was
`developed that uses a store-and-forward approach with high-quality dynamic
`medical images obtained with intravascular ultrasonography and cardiac an-
`giography. The system allows use of high-resolution dynamic images while
`preserving their original quality and can be adapted to different clinical appli-
`cations with varying requirements. The system involves a standard prepara-
`tion procedure to transmit images from one location to another prior to a
`conference; once the conference starts, however, the system becomes fully
`automatic and synchronizes the display and manipulation of images in both
`locations without further image data transmission. In general radiologic ap-
`plications, the system is superior to videoconferencing systems in that it does
`not require specialized hardware and dedicated high-bandwidth communica-
`tion links. Further investigation with large-scale studies will be required to
`determine whether these benefits can lead to more widespread acceptance of
`such a system in routine clinical practice and whether teleconferencing itself
`can enhance the effectiveness of clinical procedures.
`
`Abbreviations: DICOM = Digital Imaging and Communications in Medicine, LAN = local area network, LCD = liquid crystal display, MPEG
`= Moving Picture Expert Group, PC = personal computer, VHS = Video Home System
`
`Index terms: Computers • Computers, diagnostic aid • Computers, multimedia • Digital imaging and communications in medicine (DICOM)
`Radiology and radiologists, design of radiological facilities • Teleradiology
`
`RadioGraphics 2000; 20:1495–1503
`
`1From the Laboratory for Radiological Informatics, Department of Radiology (J.N.S., J.Z., H.K.H.), and the Cardiology Division, Department
`of Medicine (C.Z., T.N.C.), University of California, San Francisco; and the Department of Cardiology, Stanford University Medical Center,
`Stanford, Calif (E.V.P.). Recipient of a Cum Laude award for an infoRAD exhibit at the 1998 RSNA scientific assembly. Received August 17,
`IVI LLC EXHIBIT 2005
`1999; revision requested September 20 and received January 10, 2000; accepted January 21. Supported in part by the National Library of Medi-
`XILINX V. IVI LLC
`cine and the Deutsche Forschungsgemeinschaft. Address correspondence to J.N.S., Radiologische Klinik, Abt Radiodiagnostik, Universi-
`taetskliniken, Geb 49, 66421 Homburg/Saar, Germany (e-mail: jnstahl@yahoo.com).
`Inter Partes Review Case 2013-00112
`
`©RSNA, 2000
`
`

`

`1496 September-October 2000
`
`RG ■ Volume 20 • Number 5
`
`Introduction
`The American College of Radiology broadly de-
`fines the term teleradiology as “the transmission of
`radiological images from one location to another
`for the purpose of interpretation and/or consulta-
`tion” (1). In this article, we use the term telecon-
`sultation to describe a process during which two
`or more physicians hold a consultation regarding
`doubtful or problematic cases over a certain dis-
`tance by means of telecommunication. Telecon-
`sultation consists of two components: the ex-
`change of multimedia documents (eg, images,
`video, text) and the interactive verbal and graphi-
`cal discussion of the related findings and their
`consequences. In addition, we use the term tele-
`conferencing to describe a consultation in which
`both parties present cases as peers, in contrast to
`the asymmetric physician-consultant relationship
`usually seen in consultation. The system de-
`scribed in this article was originally designed for
`teleconsultation but has been extended to in-
`clude teleconferencing.
`We use the term dynamic images to describe a
`sequence of images displayed in rapid succession
`so that the human eye perceives continuous mo-
`tion. The most commonly used dynamic imaging
`modalities in radiology are ultrasonography (US)
`and angiography. We prefer the term dynamic im-
`ages to the term video because the latter is very
`closely associated with the analog video format
`used in television and videocassette recording
`equipment. Dynamic angiographic images, for
`example, are not necessarily compatible with this
`format.
`Conferences and consultations are a natural
`part of clinical practice. For consultations requir-
`ing the presentation of materials to a specialist in
`a more or less distant location, it has been tradi-
`tional to use regular mail or couriers or to travel
`to the conference in person. Teleconferencing
`and teleconsultation might help reduce the time
`and cost associated with these procedures. How-
`ever, although teleconferencing and teleradiology
`systems have been available for a number of
`years, they are still not as widely accepted by
`general radiologists for high-quality dynamic im-
`aging as was anticipated at their inception.
`A common problem in radiology teleconfer-
`encing is that a large amount of data must be
`transmitted, but the available communication
`channels have a relatively low bandwidth. At
`
`present, increasing the bandwidth (eg, using an
`asynchronous transfer mode network) is often
`not an option due to high costs and limited avail-
`ability of high-bandwidth connections. Two basic
`strategies exist to avoid this bottleneck. If the im-
`ages need to be transmitted in or near real time,
`image compression––possibly resulting in a loss
`in quality––must be applied. If a time interval be-
`tween image transmission and display is accept-
`able (eg, if the conference is scheduled to take
`place after the examination), the images can be
`transmitted in advance (eg, overnight) without
`compression. This procedure is often referred to
`as store-and-forward teleradiology.
`For static radiologic images, store-and-forward
`teleradiology is a well-established method. For
`dynamic images, most solutions published to
`date have used real-time transmission by means
`of videoconferencing systems, which digitize ana-
`log video signals and send them digitally over the
`network. These systems are easy to set up and
`use; however, they invariably require lossy image
`compression, unless costly high-bandwidth net-
`works such as asynchronous transfer mode are
`used. This compression can result in significant
`alteration of the findings (2). Concerns about
`compromised image quality might be one of the
`reasons why teleconferencing with dynamic im-
`ages is not yet widely accepted for clinical proce-
`dures.
`In this article, we present a prototypical sys-
`tem for store-and-forward teleconferencing with
`dynamic medical images. We outline implemen-
`tation details, present preliminary clinical experi-
`ence, and discuss the advantages and disadvan-
`tages of such a system.
`
`Background and Objectives
`The authors’ institutions operate two major aca-
`demic medical centers 30 miles apart. Cardiolo-
`gists in both centers use intravascular US, a rela-
`tively new imaging technique for the planning
`and assessment of coronary intervention (3). As
`part of a cooperative effort, educational confer-
`ences involving participants from both centers
`are held on an irregular basis. During these con-
`ferences, both sides present recent interventional
`cases studied with intravascular US and cardiac
`angiography and discuss possible strategies, in-
`terpretations, and outcomes.
`Thus far, these conferences have been held us-
`ing conventional methods. Cardiologists travel
`the 30 miles between the centers by car and
`
`

`

`RG ■ Volume 20 • Number 5
`
`Stahl et al 1497
`
`transport the materials on Video Home System
`(VHS) tapes and 35-mm celluloid film. Our goal
`was to develop a high-quality teleconferencing
`system that would not only reduce travel time
`and facilitate the conferencing process but also
`allow a level of quality and interactivity similar to
`that possible with conventional methods. Ini-
`tially, the installation of a videoconferencing sys-
`tem was discussed. However, it became evident
`during these discussions that the image quality
`requested by the cardiologists could not be
`achieved with the available network bandwidth
`(two T1 lines [1.5 Mbits/sec each]).
`As we looked for alternative solutions, it be-
`came obvious that a store-and-forward solution
`would be ideal because the schedule of planned
`conferences allowed sufficient time between im-
`age acquisition and the conference. Because we
`could not locate a system for store-and-forward
`teleconferencing with dynamic images, we de-
`cided to design our own system based on an ex-
`isting teleconsultation system for static radiologic
`images (4).
`
`Materials and Methods
`
`Images
`We used images obtained with two modalities:
`cardiac angiography and intravascular US. The
`intravascular US studies were always recorded on
`S-VHS videotape in National Television System
`Committee format. The cardiac angiography stud-
`ies were recorded digitally in the Digital Imaging
`and Communications in Medicine (DICOM) for-
`mat (5) (512 ´ 512 pixels, 8-bit gray scale, lossless
`compression). DICOM-compatible cardiac an-
`giography studies were stored on recordable com-
`pact disks or could be retrieved electronically
`through a local area network (LAN).
`For materials to be used in a digital confer-
`encing system, they must be available in digital
`format. Although the DICOM cardiac angiogra-
`phy studies were already recorded digitally, we
`had to convert the analogous intravascular US
`sequences. The video signal was digitized and
`compressed into Moving Picture Expert Group
`(MPEG)–1 format (352 ´ 240 pixels, 24-bit
`color, lossy compression) (6) using a personal
`computer (PC) with a video digitizer board (Data
`Translation, Marlboro, Mass).
`
`Hardware and
`Network Architecture
`The conferencing system consists of two PCs
`with Intel Pentium II 400-MHz processors, 128
`Mbytes of RAM, and 10 Gbytes of disk storage
`(Dell, Round Rock, Tex). The system at one lo-
`cation is equipped with a liquid crystal display
`(LCD) projector (1,024 ´ 768 pixels) (Proxima,
`San Diego, Calif), whereas the system at the
`other location includes a regular 21-inch cathode
`ray tube monitor for viewing. Both systems are
`equipped with Matrox Millennium II graphic
`adapters (Matrox Graphics, Toronto, Ontario,
`Canada) that provide real-time video scaling
`functions. For fast, lossless decompression of
`DICOM images, we used a commercial high-per-
`formance decompression library (Pegasus Imag-
`ing, Tampa, Fla).
`The conferencing systems are installed in the
`conference rooms of the two cardiology depart-
`ments. The systems are connected to the LAN
`(100 Mbits/sec) of their institutions, which are
`linked by two T1 lines (1.5 Mbits/sec each). The
`LAN also provides a connection to the DICOM
`catheterization laboratories and to the PC used
`for image digitization. For voice communication,
`we simply use a standard conference telephone
`equipped with a speaker. Digital voice-over-
`Internet solutions that we tested as possible alter-
`natives were difficult to configure and provided
`inferior sound quality.
`
`Software Architecture
`A large number of technologies and software
`products are available for displaying dynamic im-
`ages. With regard to medical images, these prod-
`ucts may be regarded as either general-purpose
`or specialized. General-purpose video software
`can display a large variety of digital video formats
`(eg, MPEG, AVI [Microsoft, Redmond, Wash],
`QuickTime [Apple Computer, Cupertino, Calif])
`(6). In contrast, specialized medical viewers with
`enhanced functionality are available for DICOM
`images. Architectural differences between the
`two groups make it difficult to integrate them in
`a single application.
`
`

`

`1498 September-October 2000
`
`RG ■ Volume 20 • Number 5
`
`For our project, however, it was desirable to
`use a single display architecture that could be used
`for both general-purpose video and DICOM
`multiframe images. Because the DICOM format
`is relatively simple compared with advanced digi-
`tal video formats, we decided to start with a gen-
`eral-purpose video tool kit and then add DICOM
`support, rather than vice versa.
`Because our system was designed for the Win-
`dows NT platform (Microsoft), we used the
`Microsoft DirectShow technology. DirectShow is
`a general-purpose multimedia tool kit that uses a
`filter graph model to build various kinds of multi-
`media applications (7). We found DirectShow to
`be ideally suited for our purpose because (a) it
`utilizes PC hardware resources very well and pro-
`vides high-quality images, (b) it supports a large
`number of common video formats and can easily
`be extended for specialized medical formats (eg,
`DICOM), and (c) it provides a high level of con-
`trol over the playback process (eg, frame-accurate
`positioning, slow motion, zooming/panning).
`We added the dynamic image capability to the
`teleconsultation system described earlier. This
`system provides basic teleradiology services such
`as store-and-forward file transmission and a re-
`mote control component with real-time dual cur-
`sor functionality that are compatible with the dy-
`namic image component. The system uses the
`DICOM protocol for image transmission. We
`implemented a private service-object-pair class
`that allows transmission of the MPEG video files,
`which are not DICOM-compatible.
`The dynamic image component provides the
`user interface for the DirectShow pipeline (Fig
`1). The pipeline is dynamically assembled from
`the available filters, depending on the type of me-
`dia to be presented (Fig 2). This is transparent to
`the application, so that for the dynamic image
`component there is no difference between
`DICOM angiography and MPEG US. DICOM
`sequence files are handled within the Direct-
`Show pipeline with a custom filter that we devel-
`oped.
`During the conference, the dynamic image
`component communicates with its counterpart at
`the remote site so that both can display the same
`image sequence in synchronicity. Start/pause
`commands, display parameters, playback speed,
`and sequence position are transmitted over the
`network and used to generate the same output on
`
`Figure 1. User interface for the DirectShow pipeline.
`The controls include play/pause (buttons in lower left-
`hand corner), playback speed, sequence position,
`brightness (B), contrast (C), gamma correction (G),
`and edge enhancement (E).
`
`both computers. This works in both directions,
`so that either side can assume control at any
`time.
`The software package can be set up and con-
`figured on any PC within minutes if sufficient
`hardware resources are available.
`
`Teleconferencing Procedure
`The teleconferencing procedure consists of three
`phases: data preparation, data transmission, and
`conferencing.
`
`Data Preparation.––DICOM sequences are
`transmitted over the LAN to the teleconferencing
`workstation, and analog videotapes are digitized.
`Because digitization of analog materials relies on
`time-consuming manual operation, DICOM-
`compatible modalities are to be preferred because
`they greatly simplify and automate the process.
`
`Data Transmission.––Prior to the conference,
`the necessary documents are exchanged between
`the two locations over the wide area network.
`This process may take a considerable amount of
`time depending on the available network band-
`width; however, it is fully automated in unat-
`tended mode and may, for example, be performed
`overnight. The selection of materials and the start
`of transmission are achieved in a matter of min-
`utes. At the end of the data transmission phase,
`the workstations in both locations possess identi-
`cal copies of the images from the two hospitals.
`
`

`

`RG ■ Volume 20 • Number 5
`
`Stahl et al 1499
`
`Figure 2. DirectShow pipelines. The filter graph for MPEG-encoded video sequences (top) consists of an MPEG
`splitter, a decoder filter, and a video renderer (Microsoft). The filter graph for the DICOM angiography files (bot-
`tom) includes the DICOM multiframe filter and the JPEG decompressor filter provided by the authors as well as an-
`other custom filter (cursor filter), which embeds the remote pointer into the video data.
`
`Conferencing.––During the conference, the
`workstations operate on identical sets of locally
`stored image data so that only the control infor-
`mation has to be exchanged. Thus, the band-
`width requirements during the conference are
`very low. Even a dial-up modem (33.6 Kbits/sec)
`would be sufficient to maintain real-time com-
`munication.
`It is assumed that during a conference session,
`the time available for image presentation is very
`limited. Therefore, it is necessary to spend some
`time in preparation to ensure that all images and
`image sequences are immediately available at the
`time of the conference. This is very similar to
`preparation for traditional clinical conferences,
`during which all needed materials are to be col-
`lected. The preparation time required with our
`system is comparable to that required for a con-
`ventional conference.
`
`Results
`
`Laboratory Results
`We performed a number of laboratory tests to
`objectively assess image quality and system per-
`formance. For the DICOM images, we subjec-
`tively compared the image quality produced with
`the teleconferencing system with that produced
`with the commercial DICOM viewing software
`running on the same computer. As expected,
`there was no noticeable difference in image
`quality because the DICOM images are lossless
`compressed. However, there was quality degrada-
`tion between the DICOM images and the origi-
`
`nal images because the modality scans the images
`at 1,024 ´ 1,024 resolution and subsamples to
`512 ´ 512 resolution for export to DICOM. This
`is a limitation of the DICOM interchange format
`rather than the teleconferencing system.
`The MPEG video sequences were compared
`side by side with the original S-VHS tapes played
`on a professional videocassette recorder. There
`was noticeable quality degradation caused by re-
`duction of resolution from about 800 ´ 420 to
`352 ´ 240 and compression-induced artifacts.
`However, because the spatial resolution of the in-
`travascular US technique itself is rather poor, the
`cardiologists found that the visibility of anatomic
`details and pathologic findings such as calcifica-
`tion and soft plaque was still adequate for con-
`ferencing purposes. We did not verify these re-
`sults with a formal quality evaluation because
`they are consistent with the results from other
`studies that compared MPEG-1 with S-VHS for
`echocardiography and found major discrepancies
`in only 2.7% of cases (8).
`The frame rate for the MPEG-1 video was
`29.97 frames/sec, which was identical to that for
`the original sequence. For the DICOM images,
`we could match the 30-frames/sec frame rate of
`the original sequence only when the optional im-
`age edge–enhancement algorithm was deacti-
`vated. With active edge-enhancement, the maxi-
`mum frame rate was 22 frames/sec, which could
`
`

`

`1500 September-October 2000
`
`RG ■ Volume 20 • Number 5
`
`be compensated for either by playing the se-
`quence at two-thirds speed or by skipping every
`third frame. We expect that the new generation
`of 550-MHz processors has sufficient processing
`power to perform edge-enhancement at full
`frame rate.
`However, the synchronization mechanism is
`independent of central processing unit clock
`speed and frame rate because absolute posi-
`tions (expressed in milliseconds from start of
`sequence) are used for positioning and synchro-
`nization. In the laboratory test, the difference in
`synchronization between the two systems con-
`nected through a LAN was less than 300 msec,
`even if the systems had different clock speeds
`(266 MHz vs 400 MHz). For wide area connec-
`tions, the difference in synchronization could not
`be measured exactly due to lack of an absolute
`reference time source. However, subjective tests
`did not reveal a noticeable difference. The new
`global positioning system absolute reference time
`system (Advanced Network & Services, Armonk,
`NY) being installed in our laboratory will allow
`more precise measurements.
`
`Clinical Results
`
`Local Conferences.––We documented 12 clini-
`cal conferences in “local mode” at one institution
`to subjectively assess image quality and perfor-
`mance under clinical conditions. In the local con-
`ferences, we found that overall the system fulfills
`the requirements of clinical conferencing: high
`image quality, ease of use, and fast operation.
`Some modifications, such as the addition of
`gamma correction and edge-enhancement filters,
`were necessary to improve the quality of the im-
`ages displayed with the LCD projector. Direct
`comparisons between the LCD projector and a
`high-quality cathode ray tube monitor revealed
`that the latter provided superior contrast and
`gray-scale resolution. However, the LCD projec-
`tor provided a larger viewing angle of the pro-
`jected image for a relatively large audience of
`eight to 12 physicians.
`Regarding ease of use, the main criterion was
`whether there were significant delays during the
`conference that could be attributed to the opera-
`tion of the conferencing system. Initial handling
`and reliability problems could be resolved with
`
`improvements in the software. With the updated
`software, efficiency was superior to that of con-
`ventional equipment (videocassette recorder
`and 35-mm film projector) because the need
`to change film rolls and tape cassettes was elim-
`inated.
`Regarding performance, we found that one of
`the requirements during conferencing was the
`ability to display the angiographic series in a
`DICOM cardiac angiography study (typically
`10–15 series) in very rapid succession. With our
`system, one can switch to the next series in a
`study by pressing a single button, with playback
`starting immediately. This mode of operation ob-
`viates prefetching of the series from the com-
`puter’s hard disk into main memory, which
`would increase the possible frame rate but would
`also increase the time until playback could be
`started. For this reason, we found that high-
`speed image processing algorithms and high-per-
`formance PCs are necessary for efficient clinical
`conferencing.
`We also found that up to 2 hours is required
`to prepare for a conference, depending on the
`number of analog studies that require digitiza-
`tion. The automatic transmission of DICOM
`digital angiograms over the LAN took about 1
`hour for six to eight patients (about 15 angiogra-
`phic sequences [300 Mbytes of image data] per
`patient) but could be performed without user in-
`tervention. Including the digitized intravascular
`US studies, the amount of data per patient was
`typically 400 Mbytes. The total amount of data
`required for a single conference averaged 3
`Gbytes. Up to 4 hours was required to transmit
`the materials over the dual T1 connection be-
`tween the two hospitals, also without user inter-
`vention. This is more time than what would be
`expected with a 3-Mbits/sec connection, because
`the lines are shared with other campus users.
`Overall, we found it reasonable to plan a telecon-
`ference at least 24 hours in advance, whereas a
`local conference could be set up in less than 3
`hours. The Table shows a comparison of the
`time requirements for classical conferencing ver-
`sus teleconferencing.
`
`Teleconferences.––We performed eight telecon-
`ferences between the two medical centers. The
`most important finding was that communication
`was possible and effective, not only in a technical
`sense but also in terms of human interaction. At
`
`

`

`RG ■ Volume 20 • Number 5
`
`Stahl et al 1501
`
`Time Requirements for Classical Conferencing versus Teleconferencing
`
`Classical
`Conference
`
`Teleconference
`
`Future Teleconference
`(anticipated)
`
`Time
`Required
`
`Task
`
`Time
`Required
`
`Task
`
`Time
` Required
`
`Task
`
`2 h
`
`4 h
`
`Collect and
`digitize films
`and videotapes
`Transmit images
`over network
`(unattended)
`Discuss cases
`1 h
`1 min Delete files
`
`10 min
`
`Retrieve DICOM
`sequences from
`archive
`15 min* Transmit images
`over high-speed
`network
`Discuss cases
`Delete files
`
`1 h
`1 min
`
`Phase
`
`Data preparation
`
`1 h Collect films,
`copy videotapes
`
`Data transmission
`
`1 h Drive car to
`conference
`
`Conference
`Cleanup
`
`1 h Discuss cases
`1 h Return by car
`
`*Assumes availability of the proposed Next-Generation Internet.
`
`both centers, the system was operated primarily
`by a cardiologist who was trained to use the sys-
`tem. The other team members presented per-
`tinent clinical information over the speaker-
`equipped telephone and used the remote mouse
`cursor to point out interesting findings or create
`image annotations.
`Our evaluation criteria for the teleconference
`were (a) whether all of the six to eight scheduled
`cases could be discussed within the time allotted
`for the conference, and (b) whether significant
`communication problems (eg, misunderstand-
`ings, need to repeat sentences, need to replay im-
`ages) occurred. Surprisingly, the system evaluat-
`ed positively for both criteria, even for the first
`teleconference. A possible explanation is that, ac-
`cording to our observations, the physicians fo-
`cused their attention mostly on the images, even
`when talking to each other during the traditional
`conferences. Therefore, they could adapt rela-
`tively easily to the fact that their communication
`partners were only virtually present in the tele-
`conferences.
`The cardiologists could easily subjectively re-
`construct the descriptions and provide meaning-
`ful input regarding the presented cases. The con-
`versations and discussions about the cases were
`very natural.
`
`Discussion
`The combination of off-line transmission of dy-
`namic images and real-time synchronized play-
`back offers a number of advantages over video-
`conferencing with real-time video transmission.
`First and most important, the store-and-for-
`ward technology allows use of high-resolution
`dynamic images with their original quality pre-
`served, whereas videoconferencing systems allow
`use of only low-quality compressed images due to
`bandwidth limitations. In many clinical applica-
`tions similar to cardiology conferencing, low im-
`age quality is not acceptable, and teleconferenc-
`ing would not be possible with a videoconferenc-
`ing system.
`Second, the DirectShow technology allows our
`system to use many different digital video for-
`mats, with images varying from original quality
`(DICOM) to low quality (highly compressed)
`(MPEG, AVI [Microsoft]). Thus, with this sys-
`tem it is possible to balance image quality, trans-
`mission time delay, and bandwidth requirements
`for optimal cost-effectiveness. Figure 3 shows the
`relationship between these three parameters. No
`conferencing system can be optimal in more than
`two parameters. A fast, high-quality system will
`
`

`

`1502 September-October 2000
`
`RG ■ Volume 20 • Number 5
`
`Figure 3. Relationship between image quality, immediacy, and cost-effectiveness. A location close to the
`edge of the triangle represents a high value for the corresponding parameter. With real-time videoconfer-
`encing, the transmission delay is near zero by definition and immediacy is high (gray area). Thus, improving
`image quality means higher costs owing to increased bandwidth requirements. By allowing a variable delay, a
`store-and-forward system can use the entire range of combinations (white area) to find the best compromise
`between the three factors depending on clinical requirements.
`
`be costly, whereas a low-cost system will be ei-
`ther slow or low-quality. Because a real-time
`videoconference is fast by definition, any im-
`provement in quality will result in higher costs
`owing to increased bandwidth. With store-and-
`forward technology, transmission delay becomes
`the third variable in the equation. For example,
`increasing the delay between acquisition and
`conferencing can improve image quality without
`affecting the cost of the system. The system can
`therefore be adapted to different clinical applica-
`tions with varying requirements (Fig 3).
`Third, a dynamic sequence is usually very
`short but will be reviewed several times during a
`conference. Our system will transmit the image
`data only once but can play the data back many
`times. In contrast, a videoconferencing system
`does not store the video once it is displayed; it
`must retransmit the entire sequence for each re-
`play, effectively wasting bandwidth.
`
`Fourth, our consultation system allows parties
`at both sites to control the playback process (eg,
`pause, replay, zoom, look-up table adjustment).
`This allows true interactive discussion of the
`findings. A videoconferencing system allows only
`one-way communication, with one site passively
`receiving the images transmitted by the other.
`There are, of course, possible drawbacks to
`the system. Certain clinical applications can be
`implemented only with real-time transmission of
`dynamic images. For example, remote monitor-
`ing of examinations in progress or assistance dur-
`ing interventional procedures requires immediate
`feedback from the remote party. In this particular
`clinical application, the dynamic images can be
`transmitted directly to our system with DICOM,
`thus bypassing the data preparation phase for im-
`mediate consultation. Depending on the band-
`width, the delay required for this transmission
`can be as short as a few seconds; unlike with a
`videoconferencing system, however, it cannot ap-
`proach zero. Whether this is acceptable depends
`
`

`

`RG ■ Volume 20 • Number 5
`
`Stahl et al 1503
`
`entirely on the circumstances of the clinical ap-
`plication. From our experience, most consulta-
`tion and conferencing will take place after the ex-
`amination is completed, in which case the delay
`created by the store-and-forward process is ac-
`ceptable and will result in better image quality
`and increased interactivity.
`Another possible problem involves the effort
`required for conference preparation. In our expe-
`rience, most of this effort involved digitization of
`the analog intravascular US tapes. We expect
`that the amount of effort required can be reduced
`significantly in the very near future with the tran-
`sition to digital, DICOM-compatible US imag-
`ers. Of course, any clinical conference that in-
`volves several physicians and patients requires
`some preparation effort (eg, collecting and sort-
`ing materials), regardless of the mode of confer-
`encing. The goal should always be to streamline
`the presentation of materials so that time spent in
`preparation can be regained during the confer-
`ence.
`Our system can support a number of practical
`conferencing applications. For example, in cardi-
`ology it is common practice for patients with cor-
`onary artery disease to undergo cardiac catheter-
`ization in community hospitals or imaging cen-
`ters. Some of these patients may be eligible for
`cardiac surgery, which can be performed in spe-
`cialized heart surgery centers. Because this is
`usually elective surgery, consultations between
`cardiologists and cardiac surgeons are routinely
`held to determine therapeutic strategy (ie, con-
`servative, interventional, surgical). Traditionally,
`this is done either by traveling to a conference or
`by mailing the images. In this setting, teleconfer-
`encing could be used to save travel time or in-
`crease interactivity while preserving diagnostic
`image quality.
`In radiology, similar consultations are being
`held between general radiologists and their coun-
`terparts at specialized interventional centers (eg,
`for embolizations). Typically, these consultations
`are less frequent than those related to cardiology
`due to the lower prevalence of the diseases in
`question. In this setting, our teleconferencing
`system has an additional advantage over video-
`conferencing systems in that it does not require
`specialized hardware and dedicated high-band-
`width communication links.
`
`In both applications, the topics of discussion
`are elective procedures that need to be performed
`in a specialized hospital. Store-and-forward tele-
`conferencing is ideal in these situations because
`image quality and ease of use are more important
`than immediacy.
`
`Conclusions
`Compared with commercially available video-
`conferencing systems, the teleconferencing sys-
`tem we developed can significantly improve im-
`age quality and ease of use for certain clinical ap-
`plications. Whether these benefits can lead to
`more widespread acceptance of such a system in
`routine clinical practice and whether teleconfer-
`encing itself can enhance the effectiveness of
`clinical procedures must be the subject of further
`investigation with large-scale studies.
`
`References
`1. American College of Radiology. ACR Standard for
`Teleradiology: ACR Standa