A SURGICAL PATHOLOGY SYSTEM FOR GROSS SPECIMEN
Domingos Cruz, MS 1,3, Mario Seixas, PhD 12
Institute ofMolecular Pathology and Immunology ofthe University ofPorto,
2. Medical Faculty of the University ofPorto, Portugal
3. Higher Education Institute ofMaia, Portugal
Rua Roberto Frias s/n Porto, Portugal
Tel: +351 2 5570700
The concepts used in the storage of still digital
images obtained during gross specimen examination
of tissues and organs in surgical pathology using a
described. We address
technical aspects related with the implementation of
a prototype tool to assist the pathologist during the
sampling process as well the logic archive support
to store the acquired images. We describe, also, the
hypermedia concepts that allow the navigation and
contained in the stored images. The advantages, the
technological and human limitations, and the effects
of using images in the documentation of a case are
KEYWORDS: Information, digital images, image-
based documentation, anatomic pathology, gross
electronic) of slides of tissue fragments of tissue
pieces, surgical specimens or biopsies submitted to
in Anatomic Pathology are obtained
In surgical pathology routine (gross examination),
several samples are taken from the tissue piece or
organ (sampling domain) and a macroscopical
morphologically described and precisely located
within the specimen. This report contains essential
information for ulterior microscopic analysis and
for the final diagnosis.
is elaborated; each sample
Sampling is done, trying to represent thoroughly
the whole surgical specimen. The sampling process
faces frequently difficulties that may lead to major
information losses. Often, it is difficult or even
impossible to describe the spatial location of the
fragments. This occurs mainly because the tissue
piece or the organ under examination looses its
1091-8280/99/$5.00©1999 AMIA, Inc.
references, along with the sampling process.
The knowledge of the fragment location is essential
to the diagnostic process. That is why the location
of the fragments
Generally, in surgical pathology laboratories some
guidelines for describing and sampling the most
experience or instruction manuals .
i ;Ssrbw hoaw
1 - Guidelines for a specimen sampling
Not all the specimens or situations can be covered
by a manual or a book since the guidelines represent
at least in most cases, general recommendations. In
fact many modifications to sampling procedures
concrete case are often necessary. Each specimen is
unique and thus requires variation in the dissection,
description and sampling procedures. That is why
each description is different from another and each
topography and topology.
The use of images appears to be the best way to
circumvent some of the problems posed by this
complex procedure and to document as precisely as
possible the sampling process . Furthermore,
images can substantially reduce the length of the
................................................................. ..... ........... .......
Figure 2-A print screen oftheprototype system during a uterus-hysterectomy. Note that thefragments
included in a cassette are also markedandgrouped before to be included
report and to avoid mistakes due to description
ambiguities. Finally, a modem information system
ofpathology should allow the acquisition of images
and their incorporation into the respective reports
With the purpose of avoiding information loss and
also to facilitate the correct knowledge ofthe spatial
relationship among fragments and their location in
original specimen, we designed an image
acquisition system which explores the advantages of
digital images over conventional photography  to
assist pathologists during the gross examination
process. This system provides simple graphic tools
to "annotate", on the acquired images, the precise
location of the samples. It also allows the creation
of hyperlinks of still images obtained at the key
moments ofthe sampling process.
In many cases, the acquisition of video movies or
video shots to document a process or a detail are not
good solutions. Often, images with
little or no
relevance are stored, increasing the redundancy of
the information and the amount of space needed for
The acquisition of still images at key moments
allows to synthesize the information, reduce the
redundancy and the amount of space needed for
storing and to decrease the access time to the
However, with just a separated sequence of images
we cannot represent clearly the sampling process.
This means that the information exists not only in
the image content but also, in the spatial and
temporal relationships between them. It is essential
to interconnect the acquired images in a temporal
accomplished by the establishing ofadequate links.
The use ofhypermedia to navigate between images
relationship between the images.
The main steps of the sampling process and the
correct spatial relationship among fragments and
represented using a hypermedia model.
Links in hypermedia models are defined as part of
relationships among object
logical connection between two (or more) end
via the use of anchors.
hypermedia systems allow the user to follow a link
as the basic form of interaction with the document
structure. The use of links in hypermedia allows the
users to make a choice of which presentations to
view and doing so the document structure becomes
apparent. This lead to the question of where, in our
particular case, links completely represent temporal
or spatial relationship between images.
. They specify a
Most of the
There are some theoretical approaches to solve
problems related with the use ofmultimedia objects
that require spatial and temporal relations. The
channels and synchronization arcs to solve the
problems of spatial and temporal relationships.
However in this case that kind ofapproach does not
seem to be the most appropriate, due to the
specificity ofthe application.
pathologist after the image acquisition just needs to
decide where to put the hotspots (the areas that
trigger links), creating the document by linking
areas to images. That kind of solution has the
advantage of working with graphically oriented
Web browsers. However image maps need to have
a separated image map file for presentation. Thus,
the pathologist will have to define the mapping
coordinates for each acquired image with a map
editor including these in a separated file.
Although, using a separated file to specify the
image maps is not the best solution for a system
that was designed in order to concentrate, as much
mapping coordinates and navigation scheme should
be stored inside of a single file.
To implement that concept, we defined a logical
structure representing images with regions linked to
other images in a single file.
We decided to use TIFF (Tag Image File Format) to
store images due to its ability to handle multiple
pages and store extra information in a single file.
TIFF allows, for each image, to store additional
information (text and objects) without damaging the
image contents. In fact, when we use an overlapping
layer it is possible to create objects (rectangles,
to the image (see
We create tags (annotations) to define regions of
interest (ROI). When a ROI is defined on an image,
one or more images can be linked to it, creating a
spatial relationship and a navigation link.
Every time an image is acquired it is also tagged
with the case number, timestamp and automatically
named and stored in a new page of the TIFF file.
When a ROI
immediately stored inside the file as an annotation
tag and appended to the ROI list. When a new
image is acquired, it must be linked to one region
already listed. The link ofeach region to an image is
stored by means of an annotation value. When the
pathologist wants to consult a case, the images and
all the annotations are loaded, so all the links are
is defined in the image,
Figure 3 - An example of hyperlinks between
images. Note that a region of interest it is linked
with an image or a set ofimages
During the gross specimen examination and image
information to the images. That information
stored inside the image file in an easy way using
graphical tools. It is possible to store an information
(for instance the outline of a tumor region) that due
to its complexity would be very difficult to define
and store in a database by other means.
An additional advantage of the utilization of this
format in this setting is the possibility of an easy
implementation of assisted morphometry along with
the process ofthe creation ofthe ROI.
Despite the simplicity of the concept it is possible,
implement a navigation scheme that enables the
pathologist to acquire the entire key steps of the
it can be seen in the Figures 2 and 3, to
sampling process. Using this approach it is possible
to review all the steps from the piece to a single
image or backtracking;
review different perspectives or views of a sample
and to pinpoint its precise origin.
it is indeed possible to
The browsing strategies depend on use and on the
type of user. The application can be used for gross
examination, to assist the final report, to teach and
We implemented two modes that can be regarded as
extreme examples of browsing strategies. In the
a tree-based browsing was
Relationships between ROI are represented by a tree
Figure 4 - An example of implicit browsing mode.
The selection ofa ROI in the main image window
selects the image linked to that ROI.
professional users that are already aware of the
relationships between images, and prefer a more
quick way to review them in detail. (Figure 4).
Other strategies could be conceived depending on
user needs and the problem nature.
Hardware and Software
The prototype was designed to be integrated in our
The program was implemented in Microsoft Visual
Basicg 6.0 with Wang(® OCX for Windows® 98
and Windows NT. The Microsoft Access® 97 was
used as the relational database and it runs either
under Windows® 98 and Windows NT® operating
to acquire images using digital photo cameras. In
this case a PolaroidTm Digital Camera was used to
at resolutions as high as 1600x1200
The images can be compressed with JPEG (Joint
Photographic Expert Group). The user can choose
between a low-compression to high-compression
ratio and low-quality up to high-quality images.
RESULTS AND DISCUSSION
The implemented acquisition prototype gives the
user the capability to easily create links between
images during the acquisition process with the sole
use ofa pointing device.
The graphical user interface allows the user to deal
with a graphic environment with input and output
screen that have windows, pop up boxes and the
range of mouse features seen today. Additionally
diagrams and images ofgross findings.
the form of
computer system to the cutting station. The usual
interface (keyboard and mouse) does not seem to
be a proper solution for a pathologist that works
with blooded hands. Using a camera in those
conditions is also very difficult. The lack of good
input/output and good photographic
could affect the usability of the system. The
motorization and the implementation of a hands
free control camera system (focus, aperture, and
zoom) would be a desirable solution.
is not easy to adapt a standard
Once acquired the information can be used in
several different scenarios. For instance, it can be
used by the pathologist to produce reports, to
perform quality control and to assist teaching and
training. All these tasks can occur locally or
remotely. That led us to take special care of the
usability of our system in the Web context.
TIFF allows us to store all the information in a
single file, in contrast to the navigation schemes
implemented, for instance, in HTML pages, where
the mapped regions are defined inside the HTML
files. With our strategy the image files can be
transferred between systems without the need of
transferring the linkage scheme.
However TIFF it is not supported by the current
Web browsers. In fact, the existing Web browsers
nowadays only support GIF, JPEG and PNG image
A TWAIN* acquisition interface was implemented
*TWAIN Developed by a consortium of imaging hardware and
software manufacturers, TWAIN is a cross-platform interface
for acquiring electronic images captured by TWAIN-compliant
scanners, digital cameras. and still-frame video capture boards.
We see three possible pathways to solve this
An application could generate a HMTL code,
perform special browsing needed in standard
configured to lunch an external application to
browse the images.
Finally, it is possible to extend a standard Web
browser and make it capable of use new data
type, by means ofa new protocol. For this, it is
necessary to make a content handler and a
protocol handler, and turn them available on
the Web server. Content and protocol handlers
could be downloaded over the Net, from the
same site that supplies the data.
Another drawback of TIFF
compress all the pages with JPEG turning a larger
TIFF file into a small one.
The prototype has been used in our institution for
three months in parallel with the traditional process.
So far the following conclusions were drawn:
Firstly, the graphic tools and the related tags need to
be specifically defined for the different types of
Secondly, it is necessary to improve the human-
machine interaction in the registry room in order to
turn the system fully operational.
Finally, it is clear that some organs, due to their
poorly defined anatomical limits (for instance breast
surgical specimens), do not benefit with the image
should be considered. Perhaps,
instead of images, standardized schematic drawings
should be used.
It seems that another approach
in these cases,
commom and important surgical specimens. In:
Ackerman's Surgical Pathology. 8th ed. New
York: Mosby-Year Book, Inc.; 1996. p. 2629-
Furness PN, The use of digital images in
pathology. Journal of Pathology 1997 Nov;
Computers in Surgical and Cytopathology. In:
Principles and Practice Surgical Pathology and
for handling of most
Cytopathology. 3rd ed. New York: Churchill
Livingstone Inc.; 1997. P. 113-25
O'Brien MJ, Sotnikov AV, Digital imaging in
Clinical Pathology 1996 Oct; 106(4 Suppl 1):
Hardman L., Bulterman D. and Guido van
Rossum, The Amsterdam Hypermedia Model:
Multimedia, Hypermedia Journal 5(1), 1993
July, 47 - 69.
pathology. In: Ackerman's Surgical Pathology.
8th ed. New York: Mosby-Year Book Inc.;
1996. p. 2727-32.