vamg6792@UXA.CSO.UIUC.EDU (Vincent A Mazzarella) (03/24/91)
Sender: Hospital Computer Network Discussion Group and Data Base
<HSPNET-L@albnydh2.bitnet>
REPORT FROM ACR WORKSHOP
ON COMPUTER NETWORKS IN RADIOLOGY RESEARCH
MARCH 14 - 15, 1991
Bethesda, Maryland
This workshop was sponsored by the American College of
Radiology and supported by the National Science Foundation, the
National Library of Medicine and the National Cancer Institute. It
was held in the beautiful conference facilities at the Lister Hill
Centre on the campus of the NIH in Bethesda, Maryland. About 50-60
people from around the country attended. It became evident that
almost all of them were already familiar with the extensive
computer networks provided in the country and were present to learn
more about the further developments of high speed networks.
The meeting was opened by Dr. Dan Masys, the Director of the
Lister Hill Centre for Biomedical Communication and by Dr. Ron
Evens, who Chairs the ACR Commission on Technology Assessment and
Efficacy Studies. The first session was chaired by Dr. Ron
Arenson, Professor of Radiology at the University of Pennsylvania.
The first speaker was Dan VanBelleghem, Associate Program Director,
NSFNET. He described the historical development of the present
NSFNET starting with the early beginnings of BITNET and ARPANET,
through the present NSFNET to the projected NREN which stands for
National Research and Educational Network. The 1992 budget for
this facility is 638 million dollars of which 30% is provided by
ARPA and another 30% by the NSF. The remaining monies are provided
by other bodies. The primary purpose of the NSFNET is to provide
super computer services and by the summer of 1991, the major
network for this purpose will be running at 45 megabits/second.
Mr. VanBelleghem contrasted the various computers ranging from
CRAY Y-MP/8 running at 2,000 Mflops down to the VAX 780 at 0.15 and
the IBM PC-AT running at 0.033 Mflops. There are CRAYS based at
Cornell, the NCSA at the University of Illinois, at Pittsburgh and
in San Diego.
In 1984 the network used to use the super computer based at
Boeing and accessed this with a 9600 baud line. The speed of
communication has risen incredibly and the speeds in the 89/90 time
frame are 1.5 Mbits/second rising to 45 Mbits/second in 91/92 and
to 1 Gbit/second in 95/96. In contrast, the number of packets
transmitted in 89/90 was half a billion, and in 91/92 is expected
to exceed 5.5 billion packets. The connection to Europe runs at
128 Kbits/second. The actual physical connections are maintained
by commercial interests in the form of MCI, Sprint, etc. The use
made of NSFNET is 19% interactive sessions, 21% file transfers, 16%
name servers, 26% mail/bbs/conference and 18% for other activities.
The proposed uses of NREN are to access remote systems for
visualization and for large file transfers. A number of
researchers are also placing sensing devices in remote locations
and using networks to transmit the data back to a central base.
Software distribution is expected to be a large user of NREN and
other uses are for database and library activities, scholarly
collaboration and education.
A number of policy issues were discussed by Mr. VanBelleghem.
These included security, fair use and access, connection to
commercial networks, the transition of the government supported
NSFNET to the commercial sphere, methods of charging users,
standards for network interchange and international connections.
The next speaker was Dr. Dan Masys who is the Director of the
Lister Hill Centre for Biomedical Communications at the National
Library of Medicine. He discussed the biomedical networks and once
again, provided the historical background for this discussion. In
1964 Index Medicus was computerized and in 1971, MEDLINE was
established. In 1989, this was succeeded by MEDLARS and there are
now 39 databases which are accessed for over 5 millions searches
per year. Such searches are also available from the commercial
networks as well as a number of sites around the world. This
facility is to be augmented by Loansome Doc for ordering documents
for direct delivery or by FAX and Loansome Doc will become part of
GRATEFUL MED as of January, 1992.
Dr. Masys presented an interesting nomogram which related the
amount of data to be transmitted against the time which one would
feel appropriate for transmission in order to determine the data
rate necessary. This emphasized the fact that if one wishes to
display a medical image at fairly high resolution in a screen print
time of the order of one second, extremely high data rates are
required. On the other hand, if one is satisfied with the
transmission of fairly small files of ASCII data, and the time for
transmission is relatively immaterial, then data rates of the order
of 1200 board are perfectly acceptable.
Although the use of the network for e-mail and file transfer
is high, much of the super computing and database access activities
are associated with molecular genetics and the structure of DNA.
Dr. Masys suggested that the major challenge to date is bringing
the various databases into some form of standard format such that
search routines for one database were applicable in other
databases. In many senses, this is the database analogue of the
problem faced by medical imagers with the variety of formats in
which images are found in medical imaging devices.
Dr. Masys suggested that the current problems are a lack of
many local connections and the absence of organizational
commitment. Here he cited the case that when a large number of NCI
grantees were canvassed, they supported the concept of electronic
networks but emphasized that it should be done without any
reduction in the amount of money available for research grants.
Another problem which he cited was the fact that the biomedical
sociology tends to be based very much on the Flexner Report which
stipulates that medicine should be learned at the feet of the
greats and as a consequence of this, networks tend to be
psychologically shunned by the people involved in medical teaching.
Speaking to the future, Dr. Masys suggested that MEDLINE would
be further expanded to include genetic information for gene
matching under the human genome project and that providing the
standards could be established, it would be possible to start
consideration of transmission of biomedical images.
The National Library of Medicine is in the process of
transferring the many billions of pages of text to computer data
base due to the fact that the paper with an acid base is rapidly
crumbling to dust. Such transfer to computer data base would
facilitate the automated interlibrary loan system. 57,000 prints
and photos of medicine have been placed on video disk and the
National Library of Medicine is in the process of coding another
17,000 x-ray images onto optical disks. The visible human project
includes volumetric images of the complete human which it is hoped
would provide for library searches based on object classification
within such images. This, in a sense, would lead to the concept of
a virtual library. Dr. Masys concluded his presentation with the
quotation of a Chinese proverb, "tell me and I will forget, show me
and I may remember, involve me and I shall remember".
Dr. Elliot Fishman from the Department of Radiology, Johns
Hopkins Hospital, was the next speaker and he provided an overview
of the use of computers in radiology. Perhaps without intending to
be so, Dr. Fishman was very controversial in his suggestions that
radiologists should be reading studies from workstations and that
they should therefore be able to manipulate images in order to
achieve the best reading conditions. He was also very emphatic
that users should be able to access software in order to make
desirable changes. It seemed to me that to suggest that
radiologists would perform all of their film reading sessions at
workstations ignored the method of payment and would have a
significant reduction on radiologists salaries. Further, I
personally questioned whether the distribution of software to users
would be in the best interests of the users since I am sure the FDA
would have particular concern regarding the liability issue of
software errors.
Following the morning coffee break Dr. Robert Esterhay
reviewed the development of interactive workstations. He indicated
that most activity is text base with clinical trials, literature
searches, data collection and data uploads being the major
activities. Most physicians use computers for administrative
activities but do not use them for access to data such as that
available at the National Library of Medicine. The NCI is working
towards remote data collection and access facilities in order to
promote clinical trials into a wider community. This project is
planned in three phases with the third phase being that of decision
management tools associated with the data collected. A request for
proposals has gone out and the product is in the development stage
with a projection for the fall of 1992.
Also concerned with interactive workstations was a discussion
by Dr. Michael Merickel from the University of Virginia in
Charlottesville, who discussed the specific application of
workstations for pattern recognition of atherosclerosis and cancer.
The next two speakers addressed the problems associated with
standards. Dr. Thompson from the University of North Carolina in
Chapel Hill, suggested that the approach to standards needs to be
more holistic and that the ACR/NEMA standards should be extended
into the area of hospital information systems (HIS) to include
voice as well as both image and textual data. He suggested that
radiology is only one component of the complete information
available regarding patients. He had a vision of creating
communication links that would even avoid direct physician/patient
contact. There is a standard body concerned with HIS and this
includes delegates from Europe and Japan as well as those from
North America.
Dr. Steve Horii from Georgetown University, Washington,
described the history and mandates of the various working groups
within ACR/NEMA. Much of the information regarding ACR/NEMA is
familiar to the readers here so that I will not reiterate it.
The last speaker before lunch was the Senior Vice President of
Vortech Data, Mr. Lee Bryan, who observed that medicine in fact is
a very small marketplace and we should keep a constant eye on other
markets such as CAD/CAM and publishing. He envisaged networks
moving towards file servers and in the future, computer servers in
the form of large high capacity computing facilities. He mentioned
a new system that would be available from HP that would be running
at something like 72.2 megaflops.
Thursday afternoon commenced with a further description of the
NSFNET management by Susan Calcari of the Merit Corporation which
manages NSFNET. She described what NSFNET is and how Merit fits in
to the whole management of the network. She described the various
other networks such as MILNET and NASA which fit into INTERNET and
described how the network goes from the campus to the regional,
then national and finally to the world level. To give some
indication of the speeds of T1 and T3 networks, she suggested that
T1 network was capable of transmitting five pages of typed data per
second whereas a T3 network could transmit as many as fourteen
pages of such data per second. There are 2,417 networks associated
with NSFNET of which 717 are international networks. Merit adds
about two networks per week to the overall and once again, the
usage figures were very similar to those cited previously. The
usage by packets was 23% mail, 23% for file transfer, 19% for
interactive activities and 8% for name lookup. In the area of
actual data, in terms of bytes, she demonstrated that 27% is mail,
44% is for file transfer, only 7% for interactive work and 4% for
name transfer. These figures are growing at the rate of about 20%
per month. Ms. Calcari claimed very high reliability of the system
with a quoted figure of 99.89% up time. This is remarkable and
does not coincide with the experience of most people. However, as
Ms. Calcari pointed out, the figure she was quoting was for NSFNET,
not for the many nodes within smaller networks that are connected
to the INTERNET.
In response to a question regarding the connection of INTERNET
into Eastern block countries, Ms. Calcari pointed out that INTERNET
is dependant upon TCP/IP and this is a technology which cannot be
exported to East block countries. Only BITNET connections are
allowed to the Eastern block.
The next part of the afternoon session was occupied by some
demonstrations. Dr. Carl Jaffe from Yale University demonstrated
file transfer by logging into Yale and roaming his various accounts
there. He transferred some MRI images down to the Mac II at which
he was located. Dr. Jaffe is on sabbatical at the National Library
of Medicine and he described how he had been able to work with the
graphic artists at Yale to produce educational teaching files by
transferring the images back and forth and using the program IMAGE
for viewing on the Mac.
Dr. Lawrance from the National Centre for Super-computer
Applications described access to the CRAY and other high
performance computers in order to provide exceptional computer
power to various distributed users. His handout also gave
instructions for obtaining NCSA freeware using anonymous FTP. The
anonymous FTP address is FTP.NCSA.UIUC.EDU (128.174.20.50).
Following the coffee break Dr. Hahn from SURANET described the
midlevel network. He made one quotation which was a very telling
one and that is that hardware networks create social networks.
That is certainly true in many cases.
Mr. David Ely of the Corporation for National Research
Initiatives addressed the issues of digital library systems. He
described the growth of telecomputing in relation to the growth of
other infrastructures within society. He mentioned the development
of railroads followed by highways and then electricity distribution
followed by the telephone system and suggested that telecomputing
would be a further development of this form of infrastructure. The
Corporation for National Research Initiatives is extending MEDLARS
for the National Library of Medicine and there is a GRATEFUL MED
interface via INTERNET from Unix Systems. He suggested that there
would be future possible support for nontextual information. He
provided a demonstration of the use of GRATEFUL MED over INTERNET.
Mr. Ely also introduced the concept of KNOWBOTS which are in
essence, knowledge robots. One can build a KNOWBOT locally and
then send this to a larger machine, have the program run on that
larger machine, returning processed data to the local system.
GRATEFUL MED is a good example of this concept in which a search
routine is built at the local PC and is then sent to a large
database for a search.
The final speaker of the afternoon was Ms. Karen Roubicek of
BBN Systems and Technologies Corporation which is the information
resource for NSFNET. Ms. Roubicek and a colleague, John Curran,
demonstrated a product called SLATE which provides the facility for
multiple users to access multimedia documents simultaneously and
make changes while on line. This demonstration was without doubt
the most impressive of the day in which a colleague of theirs was
working in Boston on a document that was visible to all in the
audience. The users at either end were able to point to various
items within the document, both textual images and even sound so
that editing could take place at either end.
The first speaker on the Friday morning session was Dr.
Darleen Fisher who was the Associate Program Director for
networking research at the National Science Foundation. She
repeated much of the material that had been presented on the
previous day by describing the history of ARPA net through to NREN.
It should be recognized that Dr. Fisher was not present on the
Thursday, so had not been aware of the discussions that had
previously taken place.
The following session consisted of a panel of speakers who
described their own uses of networks in radiology research. This
session was Chaired by Dr. Ron Arenson and the first speaker was
Dr. Cradduck from the University of Western Ontario who described
the use of e-mail and file servers in facilitating the development
of INTERFILE for use by the COST-B2 project in Europe. The
development of INTERFILE could not have taken place as quickly and
expeditiously as it did, had it not been for the wide electronic
communication of the members of the working group within the COST-
B2 project.
The next speaker was Dr. Kuhn from Southern Illinois
University who described some collaborative activity in conjunction
with the NCSA on the reconstruction of MRI images using various
principle components. He presented images to demonstrate this
research activity. In response to a question from the audience,
Dr. Kuhn admitted that he preferred the high speed network to the
CRAY supercomputer over having his own Ardent computer on site.
Dr. Shannon who is the Director of the Radiological Services
for the Department of Veteran Affairs described the networks that
exist within the VA and that this is almost all textual in nature.
A new network is being developed for visual data. He felt that
fast workstations that would enable to radiologist to emulate film
manipulation is very necessary and that high speed communication
networks will allow for remote conferencing. Since the term
"virtual network" had been used on several occasions Dr. Shannon
described a truth table regarding the presence or absence of an
object and whether one sees it or not. If the object is there and
one sees it, then it can be described as real. If, on the other
hand, it is there but one cannot see it, then the object is
described as transparent. If, of course, the object is not there
and cannot be seen then object is absent yet on the other hand if
the object is not there yet can be seen, it is described as being
virtual.
Dr. Mike D'Alessandro from the University of Iowa described
ETNET which is a computer conferencing system operating from the
National Library of Medicine. Dr. D'Alessandro has a keen interest
in the development of the highway network throughout the United
States and he used this as an analogue to describe the development
of computer networks going from low speed links to the high speed
projected links of the mid 90's. He suggested that the present
NSFNET is equivalent to the Lincoln Highway of the 1920's.
Dr. D'Alessandro felt it was very important that the group
attending the workshop should be held together for future
discussions and suggested that ETNET should be used to that end.
Dr. Peter Santago of Bowman Gray School of Medicine described
the use of a 45 Mbit/second network in North Carolina to provide
for teleconsultation. He was followed by Captain Jerry Thomas from
the Uniformed Services University of Health Sciences in Washington,
DC. Dr. Thomas reported on the networks that had been set up for
imaging in association with the Gulf war. He also described the
use of local area networks for educational programs and many of the
radiology students at the US UHS are taught by this technique.
The session was concluded by Dr. Taaffe from Massachusetts
General Hospital who described the PACS system under development at
Massachusetts General Hospital. The MGH deals with close to
250,000 studies/year including 9,500 from nuclear medicine. This
gives a workload of something of the order of 1,200 studies/day for
a staff of 79 radiologists and the major problem is one of
overload. The PACS system is seen as a solution to this particular
problem and he described the system with its several links to
outlying centres.
Subsequent to the panel discussion there was some more general
discussion concerning the next steps to be taken with respect to
the development of a radiology network. There did not seem to be
any real conclusion to this discussion apart from the fact that Dr.
D'Alessandro reiterated the fact that ETNET should be used to
continue the discussion but in an electronic medium.
Mr. Gil Devey who had acted as the organizer of the workshop
indicated that he would be providing an executive summary of the
workshop to be published in the June issue of Investigative
Radiology.
Trevor D. Cradduck, PhD, FCCPM, ABMP
March 21, 1991
--
Trevor Cradduck, | Phone: (519) 667-6574
Department of Nuclear Medicine, | FAX: (519) 667-6734
Victoria Hospital, | E-mail:
University of Western Ontario, | BITNET: TREVORC@UWOVAX.BITNET
LONDON, Ontario, | INTERNET: TREVORC@UWOVAX.UWO.CA
Canada, N6A 4G5 | UUCP: TREVORC@RIA.UWO.CA
z HSPNET-L albnydh2 3/23/91
'DFP10@albnyvm1 Vincent Mazzarella 3/23/91 Medical Network Conference at