Image Management at FMI - Imagic Bildverarbeitung AG

Multimodal Imaging in Life Science Requires
Intelligent Digital Data Management
Abstract
Within the last decade, digital imaging
techniques replaced the classical negatives
and dark rooms. Thus, in many
microscopic facilities a tremendous
amount of digital data piles up in various
data formats and has to be stored, administered
and documented. Moreover, due to
technical improvements, i.e. the increased
resolution of images, automated image
acquisition tools and the growing number
of 3D and 4D applications, the data is
overwhelming the storage and administration
capacities. Only by using dedicated
image management systems the whole
process of image acquisition, documentation
and storage can be automated, standardised
and, most important, will keep
the value of your image over a long period
of time. In addition to loss-free and effective
documentation, the stored knowledge
can be further reinvestigated under different
new aspects and kept in use even
after the operator left the laboratory.
However, the introduction of image
databases is the first step towards a final
solution for data handling, since there is
a lot more information that has to be
handled, e.g. sample specifications,
preparation protocols, analytical data
and measurements. Future laboratory
data management systems must provide
a convincing solution for storing all essential
data that arise during the process
flow of laboratory investigation. These
systems might even replace the old-fashioned
hand-written laboratory journal.
From Negative to Digital Imaging: A
Change of Paradigm
The most prominent change that microscopy
facilities have undergone in the
last 10 years is the replacement of
darkroom technology by digital imaging.
This change does not only affect the
photographers, who had to switch from
classical handcraft to computer work,
but also the microscopist.
Keywords :
image databases, image processing, image
server, multimodal imaging, data management,
digital imaging
Reprint from
Today, the number of images that can be
recorded is virtually unlimited. Moreover,
the broad range of image processing tools
leads to even more images with modified
information content. This flood of digital
data, however, poses challenges for the organisation
and documentation of the image
pool. Requirements for a data management
system following the work-flow in an
imaging facility are…
� Each respective image file has to be
uniquely named
� Experimental information as well as a
brief description of the motif have to
be documented together with the image
files and location on data carriers
� The image files must be linked to
other analogue and/or digital experimental
data from the same project
� Raw image data must be clearly
distinguishable from modified or
processed images
� The storage of digital data on local
drives with personalised folder
architecture has to be replaced by a
standardized storage process on a
central file server.
� Image data must be stored on a
common server, accessible from all
workstations
� To facilitate viewing and sharing of
images, the image file format should
be standardised and independent
from the image source
� A backup routine must guarantee
long-term availability of all important
image data independent from file
format or data carrier technology
� The transfer of complete information
has to be assured when personnel
changes
In general, there are two distinct levels
of information flow: the classical laboratory
workflow, usually documented in
the laboratory journal, and the data
backbone, necessary for storage and
retrieval. These levels have to be relationally
linked (see also Fig. 2).
Image Databases and Their Surplus Value
for Microscopy Units: A Field Report
The generation of digital image data today
ranges from simple digital photography
to the highly sophisticated recording
of light and electron microscopic data in
3D, 4D and 5D. Many of the digital
recording devices in high-end microscopy
are using non-standardised file
formats, depending on the manufacturer
and the complexity of the image data
(e.g. *.lsm, *.zvi, *.lei, *.ics, *.pic, *.ome).
For each of these systems, recording information
should be digitally linked to
the image files, either within the image
file or within an additional text file (e.g.
*.lei, *.txt, *.xml), in order to facilitate
the restoration of the imaging conditions
later on. Such image/info packages are
transferred to a central image server, secured
by regular backups. This usually
leads to a complex process-architecture
as shown in Fig. 1.
Although at first glance this architecture
seems to solve most problems in handling
digital image data, at closer inspection
it does not. Even if data storage and
backup might assure data safety, this
system is hardly to search through for a
specific information. Just imagine how
long it takes to find this information,
from screening the laboratory files for
the name and searching the file of interest
on the image server or within the
backup files? How is the retrieval of relevant
information from the laboratory
files guaranteed? Who other than the author
is able to find information within his
laboratory journal? Even worse: what, if
one searches not for information from
one defined experiment but looking for
general image data, e.g. on the distribution
of cell organelles within a certain tissue
type? Is it possible to search for relevant
image files in running and finished
projects by screening the lab files? At this
point, at latest, image databases really
become a mandatory tool for data management
within a microscopic facility.
Image management systems, e.g.
ImageAccess provide complete solutions
for handling images and the corresponding
information, from image acquisition
and storage through image evaluation
Reprint from Imaging & Microscopy 04/2005, pp 45-47 GIT VERLAG GmbH & Co. KG, Darmstadt, Germany, www.gitverlag.com

Fig. 1: Digital imaging in life sciences
Each digital imaging device is usually linked to a computer-based image management software (based on e.g.
Windows, Mac OS or Linux), especially adapted to the recording device. For a proper info management, i.e. documentation
of the raw image data, the image files must be linked to the (handwritten) laboratory journal, and this link has
also to be established on the image server. Finally, for long term storage a regular server backup must be established
to guarantee full data recovery to the image server after system crashes; this backup regime should also cover the information
from the laboratory journals.
Fig. 2: Process and information flow in a microscopy unit
For optimal process control, every experimental data produced for a certain sample has to be stored in the image database.
Standardised sample information must be linked to every image of the respective sample within the database by
the operator. For each image, microscope and camera metadata (e.g. instrument settings, calibration information)
should be automatically stored to the database. All stored information is afterwards used for further processing of the
images, and the newly generated information, e.g. measurements, is again automatically stored to the database.
Finally, all information can be directly used for the semi-automated generation of reports using standardised report
forms based on customisable templates.
and analysis up to the semi-automated
generation of reports and presentations
(Fig. 2). All data generators in the lab are
ideally linked to the database via a grabbing
interface. The image files are stored
in a standardised, uncompressed file format,
and all acquisition information
stored in the header of the image file or
in a separate file is now available as
search fields within the image database
(e.g. magnification, objective, calibration).
Moreover, unambiguous naming of
folders and image files is guaranteed and
administered by the database.
On top of these automatically stored image
data, the database can handle additional
information from the laboratory
files such as sample data, preparation
protocols or specific comments on sample
and/or images. For essential information,
the input should be made mandatory
– i.e. image recording is not possible
without entering this information into
the database. Customisable database
layouts allow a tailor-made organisation
of the images together with all accompanying
information in order to cover the
laboratory needs in both, university or
industry. In the end, even the semi-automated
compilation of experimental reports
is possible directly from the
database, using customised templates.
Benefits of Using Image Data
Management Tools
Besides the evident advantages, i.e. (1)
the automation and standardisation of
image acquisition and information storage
and (2) the reduction of complexity
regarding the number of necessary software
packages for image recording, evaluation
and measurement, the most important
gain is the searching capability of
such databases. This leads to reliable and
fast results. Moreover, the search result is
independent from individual bias, as long
as common rules for data input are defined
and obeyed by the staff. Although
the resources needed for implementing
such a system in our labs were formidable
in both aspects, finances as well as
manpower, we learned that the advantages
we now see are tremendous:
� All image data accessible through the
intranet or even internet
� Independent and safe access through
flexible user management
� Image archiving directly at acquisition
or processing workstation
� One point of entry for all different file
formats
� No loss of files anymore
� Data mining across all different project
and experiments possible

Fig. 3: Integrated laboratory data management
Today, the digitalisation rate of experimental data has already reached almost 100%. This is true for the generation of
data (e.g. images, spectra) as well as for data processing (e.g. measurements, statistics) and the final publication in internal
reports or scientific journals.
Thus, the traditionally handwritten, i.e. non-digital laboratory journal is no longer the adequate tool to overlook this
huge amount of information, and it might soon be replaced by software-based solutions providing an Integrated Laboratory
Management System that will go along with the complete experimental project flow and integrates all digital
data generators and processors, e.g. image databases (red frame).
� Massively reduced resources for data
maintenance and retrieval
We feel that image databases will evolve
from “nice-to-have but expensive addons”
to absolutely essential tools for
information management and process
control in integrated microscopy facilities.
Future of Image Databases: A New Era
of Viewing and Retrieving Images and
Metadata
Although the standardisation of the whole
imaging work flow as shown in Fig. 2 is an
important step towards an optimal information
handling, this is rather the beginning
than the endpoint of an evolution. A
lot of additional digital information has to
be stored, e.g. Excel sheets with sample
lists, Word documents with preparation
protocols or Adobe Acrobat files with related
literature. Moreover, images are just
one class of digital information that is generated
during experiments, and only the
combination of microscopic information
with results from e.g. molecular biology
approaches or spectroscopic examinations
might finally lead to satisfying results.
The most prominent link between all experimental
data that may contribute to
the final report still is the laboratory
journal. This leads to the idea that a
management system for all available laboratory
data should simply be replaced
by a digital laboratory journal and avoid
multiple input procedures of the same
data at different working stations (Fig.
3). Similar to the carry-around book,
such a software package has to be present
at every data source in the laboratory
as well as on every desktop, and it
should automatically provide relational
links to the respective data generator.
For future applications, also the distinction
between available data and essential
data might become more and more important.
The declining costs of disk space
together with the increasing computer
performance enable the storage of
literally all information available, while
eventually only a small portion of this
information is necessary for a future eval-
uation. So the final question might be: do
we really need all the information we can
store and, if not, how do we know in the
beginning what information we need in
the future? Until we find an answer to this
question, semi-automated storage of all
available data in qualified database systems
seems to be the only solution to
avoid loss of information and data. This
turns data mining and reallocation of old
data into a real scientific work field in order
to provide a knowledge generation
environment for future challenges in various
application fields.
Roger Wepf, Stefan Biel, Patrick Schwarb
Stefan Biel
Advanced Development Deo/AP
Tel.: +49 40 4909 6671
stefan.biel@beiersdorf.com
Roger Wepf
GM Research Microscopy
Tel: +49 40 4909 2588
roger.wepf@beiersdorf.com
Beiersdorf AG
Unnastraße 48
20245 Hamburg, Germany
Patrick Schwarb
Friedrich Miescher Institute for Biomedical
Research
Microscopy and Imaging Facility
Maulbeerstr. 66
4058 Basel, Switzerland
Tel : +41 79 353 49 65
patrick.schwarb@fmi.ch

Image Management at FMI
(Friedrich Miescher Institut, Part of the Novartis Research Foundation, Basel)
Founded in 1970, the Friedrich Miescher
Institute is devoted to fundamental
biomedical research. It employs new
technologies to explore basic molecular
mechanisms of cells and organisms in
health and disease. The current research
focuses on the fields of epigenetic,
growth control and neurobiology. The
FMI is an internationally recognized research
center that has initiated key developments
in molecular biology over the
years. It also provides young scientists
from all over the world with an opportunity
to participate in scientific research.
FMI employs 22 research groups and a
core of technicians and support staff, and
is home to 90 PhD students and 75 postdoctoral
fellows from almost 40 different
countries.
In May 2005 the Image Management System
ImageAccess by Imagic Bildverarbeitung
AG, Switzerland, has been introduced
at the FMI.
The goal was a complete solution to handle
the fast increasing amount of image
related data at the institute. In order to
cope with the wide variety of image ac-
quisition systems (wide field, confocal
and two-photon microscopes) and to
meet the requirements of other imaging
systems (Blot, Gel and Scanner) and
desktop users, the system was designed
to provide several different data views
and keyword lists.
ImageAccess, as a well established Image
Management System in general microscopy
and one of the leaders in handling
microscope operating parameters
and metadata, was extended with experiment
and project information in order to
achieve a long term value of each picture.
All meta data are automatically stored
within the database, and the linked image
and document files are kept on a
dedicated file server. This allows instant
access from any work place to images,
documents and meta data throughout
the whole FMI intranet for further processing
and presentation. Due to the variety
of image types (2D, 3D, 4D, movies),
they are linked to specific external viewers
(e.g. IMARIS) or can be dropped to
Figure 1: ImageAcess user interface showing thumbnails, project and experimental data
any other imaging software (e.g. Image-
Pro, Metamoroph, Photoshop) for visualization.
After a year of experience using the system,
some important facts are to be considered:
� Advanced automated microscopy requires
individual setups to maintain
speed and resolution. This customisation
process may take quite some time
and involves local IT
� Dedicated user trainings for the different
tasks like image acquisition,
processing and presentation are key
requirements of a successful system
implementation
� Due to the typically large fluctuations
of students and post docs, it is essential
to maintain and build up system
know how by assigning in-house super
users within the various groups
ImageAccess by Imagic Bildverarbeitung
AG is one of the only Imaging
Systems on the market not only to manage
2D images, but also to cope with the
fast growing amount of multidimensional
datasets and metadata. The user interface
reflects nicely the scientist’s work
flow, increases productivity and is a
great help to achieve the real scientific
goals.
Imagic Bildverarbeitung AG
Europastrasse 27
CH-8152 Glattbrugg
Tel. +41 (0)44 809 40 60
Fax +41 (0)44 809 40 61
www.imagic-imaging.com