Pixel-perfect
Why digital imaging leaders forge ahead
September 2003 Eric Skjei
If there
isn’t already a digital-imaging system in your laboratory,
there probably should be one soon. This technology is steadily making
the transition from an infrequent add-on used here and there by
pathologist enthusiasts and early adopters to a powerful tool for
not only clinical diagnosis and consultation but also education,
documentation, quality assurance, and cost control. Digital imaging
even includes digital microscopes (Nikon’s CoolScope, for
example), electronic “staining” of slides, and digital
analysis of the specimen, including its DNA distribution.
“We now
know that we can create digital images that can be better than chemical-based
photography and that offer significant advantages in terms of the
ability to manipulate, store, retrieve, and analyze information,
including information that isn’t available to the naked eye,”
says Mark Tuthill, MD, division head of pathology informatics at
Henry Ford Hospital, Detroit. “This is a huge step forward.”
Right now, digital
imaging is probably used most often to take still pictures of gross
specimens. “It’s cost-effective, user-friendly, and
much more efficient for all parties involved,” says Michael
Becich, MD, PhD, director of the Center for Pathology Informatics
at the University of Pittsburgh Medical Center, Shadyside. Dr. Becich,
a pioneer in whole-slide digital imaging, is a founder of Interscope
Technologies, which manufactures a whole-slide digital-imaging system.
The company recently won an Air Force contract to demonstrate and
implement a telepathology system. “The Air Force can’t
recruit and retain pathologists at all of their locations,”
Dr. Becich explains. “Obviously, they can process the tissues
and provide the slides, but they cannot keep enough high-quality
expertise in all locations where it’s needed.” Interscope’s
digital-imaging system will fill that gap.
Digital imaging
is also slowly making its way into pathology reports, though it’s
still the exception rather than the rule. In most labs where digital
imaging has made its debut, especially in community hospital settings,
it has been driven by the lab’s desire to make reports more
attractive and effective. In these cases, digital imaging’s
introduction is often fueled by a few pathologists and other laboratorians
with a strong interest in the technology.
One fan is Eric
Schubert, MD, a pathologist at Memorial Hospital, Chattanooga, Tenn.
“I actually enjoy making reports with images,” Dr. Schubert
says. “I enjoy the process, and I’m not the only one.
There are others here who have the same attitude.” He estimates
that perhaps five percent of his lab’s reports now include
images, but he expects that percentage to rise. The technology offers
the lab a way to sell its services to existing clients and new prospects,
Dr. Schubert points out.
“I think
the big bang for your buck in imaging is in marketing. We’re
a hospital-based group, but we do actually try to get a lot of outside
accounts, serving community-based docs.”
Memorial Hospital
is one of the many places where digital imaging is actively used
at the gross stages of the laboratory process. “We do a lot
of gross images, and in fact we go to the OR fairly often and take
pictures of the organs in situ,” Dr. Schubert says. This includes
taking pictures of the radiographs that are typically hanging up
in the OR during the operation. “We’ve had surgeons
say it’s really neat when they get a report from us and they
can see not only the diagnosis but also a picture of the radiology,
a picture of the gross, and a picture of the microscopic views,”
he says. “It’s great to be able to put all those modalities
into one report.”
Commercial
products are steadily infiltrating the imaging market.
Tamtron, a division of Impath, began offering this year an imaging
module that is integrated into the company’s AP PowerPath
product. Tamtron has been working closely with the Department of
Pathology at the University of Washington, among others, for about
the past 18 months to develop and implement the module, which is
scheduled to go live this fall as part of a version upgrade.
“We’ve
been using it a fair bit in the test mode and really like it a lot,”
says Rodney Schmidt, MD, PhD, associate professor of pathology and
director of medical informatics at the university. “In fact,
it’s so nice that that’s one of my prime motivators
for getting the whole package into production soon.”
The imaging module is a tightly integrated extension of the core
product, he says. “Basically there’s one window that
has several tabs for different types of information relating to
a case, and what they did with imaging is just add another tab onto
that,” he says. “So it’s still a very case-centric
approach to design.”
If the pathologist
has already taken images and associated them with the case, they
show up on the imaging tab. If not, he or she can click on a camera
icon above the tab to bring up a window with a live feed from the
camera. The pathologist frames the shot and, with the click of a
button, acquires the image and associates it with the case. There’s
also an opportunity to add information, such as a legend or a different
name for the image.
The image is
stored in the database at high resolution. “You then have
the option of including a particular image in the report,”
says Dr. Schmidt. “If you choose to include it, the software
in the background automatically makes a lower-resolution image that
is appropriate for printed output and puts that lower-resolution
copy in the report.” The high-resolution image is retained
in the database for later use in, say, a PowerPoint presentation
or manuscript.
The use of images
in reports, Dr. Schmidt says, is likely to be “very much dependent
on the particular pathologist and the particular type of case.”
Some want to collect pictures for conferencing or for additional
documentation purposes, but not necessarily for the report. “At
the time that you do the gross you can be taking pictures and they’re
saved right with the case so they’re available for reference,
but they wouldn’t necessarily go into the report,” he
says.
The system accommodates
different cameras. The method that is least bound to a specific
camera uses a Twain image-capture interface for Windows operating
systems. This interface accommodates any camera that is fully Twain-compliant.
Alternatively, the system allows importing and exporting of images
to and from the file system. Regardless of the method, image acquisition
usually involves only a few clicks of the mouse.
Dr. Schmidt
himself prefers “one of the consumer-level Olympus cameras”
for gross photography.
“Those
cameras are nice because they produce the equivalent of a video
output through a separate cable to the video card in the computer.
You can see the real-time view right on the computer monitor to
frame everything. You click the button and the camera automatically
takes the picture and downloads it into the database.” The
integration is smooth, Dr. Schmidt says, and the cameras are relatively
inexpensive and becoming high resolution.
The system takes
gross and microscopic pictures. “Both types appear on the
tab at the time you take the picture, and you can specify whether
it’s a gross photo or a photomicrograph,” Dr. Schmidt
says. “If you specify it as a photomicrograph and then click
the check box to include it in the report, it automatically lands
in the microscopic description section of the report.” The
user doesn’t have to manipulate the image in any way—that’s
handled automatically. “From my point of view, that is the
real key to this product because it’s all about workflow,”
says Dr. Schmidt. “The easier you make it for the user to
get a good-looking report, the more usable it is.”
How quickly
will digital imaging replace film? Dr. Schmidt is optimistic: “I
expect us to be eliminating film for gross photography probably
within two months of going live.” The timeline for photomicroscopy
will depend in part on how many cameras Dr. Schmidt deploys. “If
we have only one in the department, the transition to full digital
will be pretty slow, because no one wants to get out of their chair
and wander over to the imaging station,” he says. “But
we’re going to be deploying cameras on most pathologists’
desks, so I’m expecting the transition time to digital to
be relatively fast.”
In other environments,
digital imaging has already become a mainstream workflow tool to
help pathologists handle large volumes of cases. “Just the
other day I did 200 cases, and I imaged every one and looked at
the gross images of every one,” says Steve McClain, MD, director
of dermatopathology and director of pathology informatics at the
Albert Einstein College of Medicine/Montefiore Medical Center, NY.
“So we’re talking about getting this technology to a
very fast level, to a level where it doesn’t impede the process—it
enhances it.” Images, he says, are a new form of pathology
documentation. “We can now document the entire process, from
the time the requisition and the specimen enter the system to the
time when it’s grossed and cut, to slide histology, surgical
pathology, cytology, second opinions, and an array of other consulting
and educational uses, including autopsies,” he says. “It’s
limitless.”
What makes this
kind of digital imaging possible? Dr. McClain points above all to
the use of bar codes. “We bar code everything, each requisition,
each specimen bottle, every slide,” he says. “The use
of bar codes to make imaging, image recording reliable in terms
of identification is a critical step because putting the wrong image
in the report doesn’t help anyone and may well lead to a disaster
or an error,” he says. Without bar coding or some other form
of automated, reliable identification system, he adds, digital imaging
is going to have limited application in health care.
Henry
Ford Hospital, too, uses digital photos for some documentation
and hopes to do more. Dr. Tuthill is now shaping a next-generation
system. “We want to create a digital-imaging system that will
not only do the photo documentation function we’ve already
been doing but will provide for images that are stored in such a
way that we can do ad hoc and post-acquisition analysis on them,”
he says.
“If you’re
going to do this type of work effectively, these images can’t
be just stored as files sitting on a single hard drive, times 20
pathologists,” Dr. Tuthill adds.
Think about
searching the Internet, he suggests. If you undertook a search and
all you got back was the result from one server, then found you
had to search again on a second server, then a third, and so on,
you’d soon give up. “We need to have an integrated imaging
database that supports not only the photo documentation, education,
and research components but also supports the diagnostic components
of using digital imaging,” Dr. Tuthill says.
Such a database
has to be integrated with the other clinical information systems
from which clinicians and others draw information. For example,
the image database needs to be integrated with the anatomic pathology
information system, Dr. Tuthill says. Otherwise, users have to find
the case, search for the image, tie the two together, and send them
out for review. The process must be smoother and simpler than that.
Integration
with other clinical information systems also means reducing the
number of times images have to be replicated, which saves money.
“When the surgeon goes to look at the pathology report and
he wants to see the image of this surgery, we want him or her to
be pulling from the same image database to see that image,”
Dr. Tuthill says. “We’re really looking for an integrated
imaging database that works across all our clinical information
systems and allows the pathology images, once they’re acquired
and released, to be viewed by those we have decided should have
access to them.”
Along with bar
codes and image databases, digital imaging relies on “fast
networks and good interconnections to information systems,”
says Yukako Yagi, PhD, director of technology management at University
of Pittsburgh Medical Center, Shadyside. UPMC is developing a decision-support
system to help pathologists interpret image information, and the
center has already implemented a DVD jukebox storage system for
its images. In short, this is a process that requires a complex
web of tools.
To become
truly mainstream digital imaging will have to prove itself
capable of helping control quality and costs. And it has the potential
to do just that.
“I’m
going to tell you a secret that’s only about a hundred years
old,” Dr. McClain says. “And that is that the key to
reducing errors is to be able to correlate the gross with the microscopic.”
Dr. McClain uses a system developed at Montefiore Medical Center
that allows him to scan a bar-coded slide at the microscope and
instantly view the relevant digital gross images. This could, of
course, also be accomplished with traditional film-based photography,
but the process would be prohibitively cumbersome.
Comparing gross
findings with microscopic findings, says Dr. McClain, provides answers
to basic questions such as: Do we have the lesion? Have we cut enough
slides? Should we go back and cut more? Are the margins free? “Sometimes
the margins are clearly involved in the gross, but that’s
not reflected in the report text,” Dr. McClain notes. “A
single glance oftentimes at my gross images shows that the margins
are free—it’s pretty obvious.”
This kind of
simple comparison also highlights other errors, such as a mislabeled
slide or specimen. And the potential for error reduction is significant.
“When we talk about reducing errors by an order of magnitude,
by a factor of 10, so your accuracy goes from 99 percent to 99.9
percent, we’re talking about very significant kinds of quality
steps,” Dr. McClain says. And it is efficient: “Reviewing
gross images adds maybe four seconds to my process, so we do it
in every case now,” he says.
“Quality
is going to be a big driver,” Dr. Becich agrees, “because
there are too many medical errors out there, and reducing them is
going to require everyone’s best efforts.” But within
pathology, quality assurance too often depends on self-regulation,
Dr. Becich says. “You’ve got your own pathologists reviewing
their own material, dictating whether something is right or wrong,”
Dr. Becich points out. “There’s a natural conflict of
interest.”
“The right
way to do quality assurance,” Dr. Becich continues, “would
be for institution A to be the quality assurance officer for institution
B.” But think about how that happens now: “You have
to grab a bunch of glass slides, bundle up a bunch of pathology
reports, and send them to another place. We’re nervous enough
about doing that internally because we lose slides every time we
pull them from the files and put them on somebody’s desk,”
he says. “Moreover, after the diagnostic stage the slides
become of secondary importance because everyone is focused on signing
out the stuff that’s coming through the door. And that’s
why we lose things.”
On average,
he says, if you could examine major medical centers with large slide
archives that run many conferences and tumor boards and provide
consultations, you would find about “20 percent of the slides
missing from the files.” And when the slides can’t be
found, what happens? “We go back and do another cut, and there’s
a cost associated with that.” And that’s just for biopsies
and large resections. “You lose a cytology slide and it’s
gone forever, so in these cases, where you only have one precious
sample, a digital imaging quality assurance step becomes even more
important,” he says.
In contrast,
in a system in which the quality assurance is driven by whole-slide
imaging, the original slide doesn’t have to be removed from
its file. “You take the digital image, you distribute it to
two or three places, and they can take two months, three months
with it, and nothing ever gets lost,” Dr. Becich says. “You
still get a complete report, and the slide image and the report
stay together.”
Dr. Becich’s
group sees 27,000 surgical specimens a year at UPMC. Of those, 9,000
are shipped to two other hospitals in the UPMC system. Keeping track
of all that traffic is a massive job. “Wouldn’t it be
nice to be able to share those 9,000 cases just by clicking and
sending, by creating an automatic rule in the LIS that says, ‘Take
these 50 cases, go out to the database, grab the digital images
and the reports, and drop them into these pathologists’ workflow
queues?’” says Dr. Becich. “That’s the direction
we’re headed in in QA.”
If eliminating
slide transfer and unnecessary recuts saves time, and if time is
money, what is the return on investment?
“Our own studies, involving Interscope and interns from Carnegie
Mellon University, found that the ROI on a digital-imaging system—for
post-diagnostic use only, meaning conference support, quality assurance,
slide sendouts, internal consultative activities—was 18 months,”
Dr. Becich says.
Not much time at all.
Eric Skjei is a writer in Stinson Beach, Calif.
|
|
|