Int J CARS (2012) 7 (Suppl 1):S97–S102
DOI 10.1007/s11548-012-0698-1
TELERADIOLOGY AND ELEARNING
European teleradiology now and in the future: results of a survey
E. Ranschaert1, F. Barneveld Binkhuysen2
1
Jeroen Bosch Ziekenhuis, Radiology, Den Bosch, The Netherlands
2
Meander Medisch Centrum, Amersfoort, The Netherlands
Keywords Teleradiology
PACS
Future
Survey
Purpose
The purpose of this survey was to attempt to map the current use of
teleradiology in Europe and to evaluate the respondents’ current and
future vision about this technique. In addition, analysis of the results
should allow the authors to give some recommendations about the
further usage and deployment of teleradiology services within
Europe.
Methods
An online electronic survey was made using the SurveyMonkey
web-based survey tool [1], consisting of 32 multiple-choice and 3
open questions. Questions were asked about demographics of the
participants, types of services used, security of data transmission,
communication of information, quality assurance and legal regula-
tions. Furthermore questions were posed concerning the potential
advantages, disadvantages, threats and opportunities of teleradiology
services. In the open questions the participants were asked to express
their opinion about the current and future role of teleradiology. The
target group was defined as radiologists, radiology residents and
radiology managers with a professional activity in a European
country. Advanced logic features were used, as provided by Sur-
veyMonkey, directing participants to the right questions and
excluding participants not belonging to the target group. The survey
was published online with collaboration of the ESR. Participation was
encouraged on the ESR homepage under ‘‘Highlights’’ [2]. A message
about the survey was published on the ESR Facebook page [3] and
also distributed via Twitter [4]. A brief article was published in the
September 2011 issue of the ESR Science Report. The authors sent an
electronic invitation to the National Societies of Radiology of Euro-
pean countries of which the addresses can be found on the ESR
Institutional Member Societies webpage. The LinkedIn virtual net-
work was used to distribute personal invitations within the authors’
network [5]. Finally some publicity was made during the MIR 2011
Scientific Meeting [6] and the Dutch Radiology Days [7]. The survey
remained online between September 14th and October 10th. For
analysis of the quantitative data the SurveyMonkey statistical tools
and SPSS statistical software (SPSS Statistics 20, IBM Corporation,
NY, USA) were used. For the qualitative analysis of data the thematic
coding technique was used as described by Uwe Flick [8].
Results
A total of 368 persons participated in the survey, of which 70.7 %
radiologists, 13.6 % residents and 6 % radiology managers. The
participants represented 35 European countries, the two largest groups
coming from Austria (22 %) and Belgium (22 %). The third largest
group came from the UK (6.5 %), followed by the Netherlands
(5.7 %), Germany (5 %), Italy (3.7 %), Estonia (4 %), Spain (3.7 %),
France (3.4 %), Greece (2.8 %), and Sweden (2.8 %). Of the total,
64.8 % were already using teleradiology and 36.1 % planned to use it
in the near future. 17.6 % did not use or plan to use this technique.
Distribution of images within the same institution is the most fre-
quently used form of teleradiology (70.9 %) followed by reading
images from home during nighttime and weekends (44.3 %). Out-
sourcing of examinations is done by 35.5 % of the participants,
mostly to obtain an expert opinion (20.2 %) or to solve capacity
problems (19.7 %). Teleradiology services are provided by 31.6 % of
the respondents on a non-commercial basis and by 26 % on a com-
mercial basis. In most cases it concerns primary reads (76.6 %) and
subspecialty interpretations (50.5 %). The examinations mostly come
from other hospitals within same country (86.9 %). DICOM is the
preferred image format (91.6 %). A VPN-type connection is present
in 41.6 % of cases, a regional network in 37.4 %. Additional infor-
mation about the patient is usually obtained through connection with
the PACS/RIS (54.1 %), 29.7 % uses telephone and 23.8 % fax. Only
16.2 % is using a dedicated teleradiology platform or software. The
possibility to discuss cases in a collaborative network is considered as
the greatest advantage of teleradiology (46.1 %), followed by the
ability to efficiently distribute the workload (37.8 %). The major
advantages of outsourcing with teleradiology are: the possibility to
ask for second opinion/subspecialty advises (81.9 %) and the greater
availability of radiologists for on-call services (71.3 %). According to
68.9 % the main disadvantage of outsourcing image interpretations
lies in the insufficient integration of patient history and/or previous
studies. Limited communication with clinicians is regarded as the
second most important disadvantage (67.7 %). When asking about
possible threats and dangers of teleradiology, 60.2 % thinks that
radiologists might lose control over their business, 54.4 % fears that it
might diminish the quality of radiology services, 48.8 % thinks that it
could create instability in the job market and the radiologists’ income,
and another 48.4 % sees it as a threat for good patient care. According
to 78.3 % the quality of teleradiology services could be improved by
developing an obligatory quality assurance system and specific tele-
radiology training programme (70 %). The provision of double
readings is also seen as a valuable solution (56.9 %). Regarding cross-
border teleradiology, 64.8 % agrees that pan-European image sharing
should be allowed under specific regulations, however the majority of
participants who addressed this topic thinks it has no future. 6.55 %
mentions the need for further standardisation including a better
financial and legal regulation. According to 59.9 % an adapted
reimbursement scheme should be agreed upon. A positive vision
regarding the future of teleradiology was expressed by 79.9 %, even
46.1 % of the responders believed that its importance would grow.
The mentioned principal reasons for teleradiology to stay were:
shortage of radiologists and need for sub-specialisation. More
opportunities should be found in emergency reading services, flexible
support of small practices and in the creation of platforms to enhance
collaboration between radiologists.
Conclusion
Teleradiology has become a technique that is widely used throughout
Europe, mostly for distribution of work within the same institution
and for on-call purposes. Usage of commercial teleradiology services
is still relatively limited, mostly for solving capacity problems during
night calls. Regarding cross-border image distribution within Europe
there is still a great demand for a clear pan-European legislation, price
regulation and quality assurance system. When the market for
123
S98
teleradiology was only seen as limited, expert opinions and second
opinions were thought to have the biggest potential, followed by
emergency services, support of small practices and hospitals, and
creation of platforms that may enhance collaboration between
radiologists.
References
[1] SurveyMonkey website: http://www.surveymonkey.com
[2] ESR website, Highlights: http://www.myesr.org
[3] myESR facebook webpage: http://www.facebook.com/myESR
[4] myESR Twitter webpage: http://twitter.com/myESR
[5] LinkedIn website: http://www.linkedin.com
[6] MIR Annual Scientific Meeting 2011, Sept. 29–30, Nice, France
[7] Dutch Radiology Days 2011, Sept. 29–30, Maastricht, the
Netherlands
[8] Flick U, An Introduction to Qualitative Research. Sage Publi-
cations Ltd, 2009, p. 528.
Using control messages for administrative tasks and constancy
testing of DICOM e-mail based teleradiology networks
F. Schwind1, H. Münch1, A. Schröter1, G. Weisser2, H.-P. Meinzer1,3,
U. Engelmann1,3
1
CHILI GmbH, Heidelberg, Germany
2
Universitätsklinikum, Institut für Klinische Radiologie und
Nuklearmedizin, Mannheim, Germany
3
German Cancer Research Center, Division Medical and Biological
Informatics, Heidelberg, Germany
Keywords Teleradiology
DICOM e-mail
PACS

Constancy testing
Purpose
DICOM e-mail is a wide used standard application in German tele-
radiology networks [1]. It is easy to admit new partners to the
network, because all communication is done via dedicated servers for
all participants. Only the e-mail connection must be configured and
keys for encryption exchanged between partners [2].
One major drawback is the asynchronous communication, which
makes it hard to trigger automated constancy tests between partners
and generate protocols, which are required by the German authorities
for legal reasons [3]. Furthermore it was up to now not possible to
update changes in addresses or encryption keys automatically.
This paper presents a solution for automated constancy tests and
the exchange of administrative data via e-mail, which was developed
with the Working Group on Information Technology (Arbeitsgeme-
inschaft Informationstechnologie, @GIT) of the German Radiology
Society (Deutsche Röntgengesellschaft, DRG) [4].
Methods
The newly developed DICOM e-mail service parts contain an XML
body part describing the action, which should be interpreted and
executed by the receiver. To ensure privacy, completeness and
validity of the transferred e-mails, all communication must be
encrypted PGP/GnuPG compatible and signed by the sender. The
recipient has to decrypt the e-mail and check the validity of the sig-
nature. It is recommended to configure a whitelist with trusted
partners for each allowed operation.
All participating hosts must support the message disposition
notification (MDN) described in the @GIT Whitepaper v1.5 and
RFC3798, RFC3462. While only complete messages can be notified
by the receiver with the basic MDN, only one service part per e-mail
is allowed at the moment.
To enable filtering of received messages, each service part message
is identified by an unencrypted X-TELEMEDICINE-SERVICEPART
tag, describing the action contained in the encrypted mail body.
The tag value is either an ADDRESS- or KEYUPDATE or TEST-
TRANSFER for constancy test requests respectively PROTOCOL for
123
Int J CARS (2012) 7 (Suppl 1):S97–S102
Fig. 1 Example of an XML message describing an address update
constancy test results. The XML body, which is interpreted by the
receiver, describes each action in detail.
Address and key updates
Figure 1 shows an example of an XML message for an address update
where a new e-mail address is transmitted to all partners. Each
XML service part is identified by a name (ADDRESSUPDATE)
and an optional action (SET or REMOVE). The body of the ser-
vice part is processed differently based on the found name and
action. In case of an address update a new connection is added or
updated if the given id already exists. If the action value REMOVE
is set, the connection with the given id will be removed or dis-
abled, depending on the implementation of the service part
receiver.
After completing or denying the action, a MDN is send to the
given e-mail address. For this purpose the existing @GIT error codes
were extended with new messages matching the needs of service part
messages including parser or update errors.
The service part KEYUPDATE works like the described AD-
DRESSUPDATE while only the complete PGP/GnuPG keyblock
must be transmitted to add a new key. To remove an existing PGP/
GnuPG key the key id must be transferred in a service part KEY-
UPDATE message using the action REMOVE.
Constancy tests
Service parts can also be used to trigger constancy tests between
DICOM e-mail partners. The administrator of the network sends an
e-mail containing a TESTTRANSFER request to a communication
partner. The service part must contain the address and PGP/GnuPG
key of the testdata receiver as well as the address and key of the
protocol receiver. Additionally a test dataset id and an error timeout
can be specified.
At the moment 17 datasets, matching common radiological stud-
ies, are predefined. It is possible to add new dataset as long as all
partners use the same naming convention and support the requested
imagetype.
After receiving the TESTTRANSFER request, the receiver sends
the specified test data to the given connection partner. After all data is
transferred correctly (which is ensured by using the described MDN
mechanism) the test data sender generates a service part protocol and
delivers it to the specified recipient using the service part type
PROTOCOL (Fig. 2).
Each service part PROTOCOL is send in an XML format con-
taining the transmission status (either COMPLETED or ABORTED if
the given timeout was reached), the used test dataset id and mail count
as well as the size and transfer time of each transmitted object. In case
of errors during the constancy test all error codes and descriptions are
included as well. Using this information, the protocol receiver is able
to compile a complete constancy test protocol which is compliant
with the German standard for quality assurance in teleradiology (DIN
6868-159).
Int J CARS (2012) 7 (Suppl 1):S97–S102
Fig. 2 Flowchart for constancy tests
Results
The authors implemented the described service part e-mails and tested
them in a connect-a-thon with other members of the @GIT. The
newly developed features extend the already existing ‘‘Guidelines for
a standardized teleradiology data exchange v1.6’’ (Empfehlung für
ein standardisiertes Teleradiologie Übertragungsformat) [5] and
worked between multiple teleradiology clients and servers of different
vendors. It was possible to trigger constancy tests between DICOM
e-mail partners and propagate changes in address or encryption keys
automatically between the participants.
Discussion
The newly defined and implemented extensions of the DICOM e-mail
protocol simplify administrative tasks, such as public key updates,
changes of e-mail addresses and quality measurements.
Teleradiology users in DICOM e-mail based networks can now
easily analyze the quality of service by triggering constancy tests via
DICOM e-mail while requesting the generated protocols simulta-
neously. It is now even possible to administrate such networks solely
via DICOM e-mail, once the initial connection is setup correctly.
While only four service part messages are defined at the moment,
these are easy extendable in the future to match the needs of
upcoming DICOM e-mail teleradiology networks.
Conclusion
The implementation of the described methods enable automated time-
and quality test for teleradiology networks, which communicate via
DICOM e-mail. In addition it is possible to send administrative
messages between DICOM e-mail partners. At the moment this
guidelines are only used in Germany but might be useful in other
countries as well, where the image communication is done via DI-
COM e-mail.
References
[1] Weisser G, Engelmann U, Ruggiero S, Runa A, Schröter, Baur
S, Walz M. Teleradiology applications with DICOM-e-mail.
Euro Radiol. 17(5) (2007) 1331–1340. doi 10.1007/s00330-006-
0450-8.
S99
[2] Engelmann U, Schroeter A, Schweitzer T, Meinzer HP. An
architecture for secure exhange of medical images and reports
based on mail, DICOM and public key encryption. Eur J Med
Res 7 (Suppl. I) (2002) 22.
[3] Engelmann U, Seidel B. DIN 6868-159: A German standard for
quality assurance in teleradiology. Int J CARS (2009) 4 (Suppl 1):
S158–S159.
[4] Weisser G, Walz M, Ruggiero S, Kämmerer M, Schröter A, Runa
A, Mildenberger P, Engelmann U, Standardization of teleradi-
ology using Dicom e-mail: recommendations of the German
Radiology Society. Eur Radiol. 2006 Mar;16(3):753–758.
[5] @GIT. Empfehlung für ein standardisiertes Teleradiologie
Übertragungsformat: Version 1.6 (2010) www.tele-x-standard.
de (Last access: 08.01.2012)
Telemedicine applied to the management of Ischemic Stroke
patients
C. Dario1, G. Scannapieco2, M.G. Carraro3, B. Giometto4,
C. Saccavini1, A. Rebeschini1, E. Martignago1, E. Vio1
1
Arsenàl.IT, Veneto’s Research Centre for Innovation in e-Health,
Treviso, Italy
2,4
Local Health Authority no.9, Treviso, Italy
3
Local Health Authority no.7, Pieve di Soligo, Treviso, Italy
Keywords Telecounselling
Stroke
Telemedicine
Purpose
Ischemic Stroke represents the third cause of death and the leading
cause of chronic disability in Europe: Stroke causes the 10–12 % of
death every year.
The only therapy scientifically validated for the Treatment of
Ischemic Stroke is thrombolysis [1]. This type of therapy has to be
used only in specific and organized centres because there is a high risk
of haemorrhage. Only 2 % of these patients receive thrombolytic
treatment although 20 % of users are potentially treatable. Since
2008, Veneto Region, placed on N-E of Italy, aims to set up a
structured network (according to an Hub and Spoke approach) to treat
the ischemic stroke, optimizing the use of specialized resources
among hospitals and increasing the continuity of care within the
regional territory, even in remote areas.
In the wake of HEALTH OPTIMUM Project [2], Arsenàl.IT [3]
developed a system and analysed the appropriate organization that
allows the evaluation of stroke on patients thanks to the support of a
telecounselling system that can avoid patient transfer towards a spe-
cialized structure. In this way, the telemedicine service fulfills
Regional Resolution and allows to the neurologist to prescribe
thrombolytic treatment even if he is not physically present at the
peripheral hospital.
The key purpose of these applications is to increment the number
of patients treated with thrombolysis therapy within 3 h from the
beginning of the symptoms, with the consequent decrease of number
of disabled persons, and to avoid useless patients transfer.
Methods
With some evolutions, the telecounselling system for neurology
developed by Arsenàl.IT adopts the same interoperability infrastruc-
ture that was set up during HEALTH OPTIMUM Project.
This platform, that allows sharing documents and images among
different LHAs, is based on IHE Integration profiles [4] as:
• XDS-Cross-Enterprise Document Sharing,
• XDS-I-Cross-Enterprise Document Sharing for Imaging,
• XCA-Cross-Community Access,
• etc.
123
S100
The adherence to IHE Guidelines allows using, in a coordinate and
efficient way, all the most authoritative ICT Standards that support
eHealth field, such as:
• CDAR2 Structured Documents,
• LOINC Coding System,
• PKCS#7 infrastructure for digital signature,
• DICOM images,
• ebXML messages and transactions between IHE Actors,
• SMTP and HTTP message transport.
The support of the physicians and medical directors allowed
developing the appropriate organizational models. In these models,
also clinical evaluation protocols (specific for ischemic stroke as
SITS-MOST and NIH Stroke Scale) are employed, in addition to
telemedicine system and to CT scans.
In one of these models, a high quality audio and videoconference
system is even adopted. Indeed, since 2004, the scientific literature
has shown that videoconferencing can provide significant added value
to the telecounselling systems for Ischemic Stroke management [5].
In this way the real-time stream of images captured by three different
cameras is used by physicians at the central Hospital to evaluate from
remote the clinical conditions of the patient affected by stroke and
placed in the Emergency Room (ER) at the peripheral hospital,
respecting the national legislation on privacy.
In order to start up the telecounselling system to be tested between
two hospitals, it was necessary to provide appropriate training to all
clinicians and a coaching period in order to accustom the clinical to
use the technology thanks to the support of an expert technician.
Results
The use of clinical and technological standards described above
allowed structuring an organization that will treat in a new way the
patients with Stroke that access to peripheral structures.
Each patient that arrives at the ER with stroke symptoms is
urgently submitted to neurological checks. If the ischemic Stroke is
verified and the symptoms arose within 3 h, the physician consults a
neurologist at the Hub, thanks to the Telecounselling System. This
system allows the neurologist to evaluate if it is possible to apply
thrombolysis therapy to the patient and reduce Stroke’s damages. The
doctors use a structured form for the request and for the reply of
counselling where the evaluations of SITS-MOST Protocol and
NIHSS are indicated. The neurologist can be also supported by three
video-cameras to evaluate the patient as a face to face visit (Fig. 1).
Two of the video-cameras are fixed in order to show what happen in
the room and one can be moved directly by the neurologist at the Hub
structure. The third video-camera is mobile and is used by the
emergency doctor to show important details.
If in the Spoke structure it is possible to monitor the patient in a
dedicated area, the thrombolysis treatment will be executed in the
Fig. 1 Hub’s neurologist is evaluating health conditions of a patient
at peripheral hospital thanks to the assistance of video-camera
123
Int J CARS (2012) 7 (Suppl 1):S97–S102
peripheral structure without transfer the patient. Otherwise, patient
will be transferred in the Hub to be treated.
The neurologic telecounselling system between Treviso and
Conegliano hospitals is in operation. After a trial period, in 4 cases
out of 11, the treatment was authorised by the hub centre and
carried out in the peripheral clinic or hospital with a clinical out-
come that was comparable to treatment in the main hospital. The
procedure did not show any particular critical situations under a
clinical profile.
After the success of this experimentation, a similar system was
implemented in the province of Vicenza, and many LHAs are now
considering a deployment.
Conclusion
The use of telemedicine and clinical evaluation protocols allows to
collect all necessary clinical information that helps the referring
physicians at the Hub hospital to decide if the patient has to be treated
with thrombolysis or not, and if the patient can be treated also in
peripheral hospital. The time-saving is very decisive if we consider
the only 3 h on tap to do the thrombolysis.
Instead, the number of useless transfers can be reduced and a
better and equitable service can be offered to the citizen in the spoke
hospital.
Moreover, as we forecasted in the beginning of 2009, we have
optimized the traditional telecounselling services proposed by
HEALTH OPTIMUM Project adding the employment of video-
cameras and international recognized clinical evaluation protocols.
The traditional telecounselling service, applied on neurosurgical area,
provided positive results for a total of 2300 telecounsellings per year
[2]. The neurological telecounselling service applied to the same
platform as the neurosurgical one service is giving again positive
results at this experimentation stage. Many other LHAs are now
considering the possibility of a deployment.
References
[1] Alberts MJ, Hademenos G, Latchaw RE, Jagoda A, Marler JR,
Mayberg MR, Starke RD, Todd HW, Viste KM, Girgus M,
Shephard T, Emr M, Shwayder P, Walker MD, for the Brain
Attack Coalition. Recommendations for the establishment of
primary stroke centers. JAMA. 2000;283: 3102–3109
[2] Dario C., Scannapieco G., Carraro M.G., Giometto B., Gion M.,
Scienza R., Pengo V., Battistella B., Favaro M., Vio E., (2011),
Quaderno 2/2011 Progetto HEALTH OPTIMUM, Arsenàl.IT
[3] www.consorzioarsenal.it
[4] www.ihe.net/Technical_Framework/
[5] Heinrich J. Audebert, et al, Telemedicine for Safe and Extended
Use of Thrombolysis in Stroke. The Telemedic Pilot Project for
Integrative Stroke Care (TEMPiS) in Bavaria, American Heart
Association, Inc. 2004, DOI: 10.1161/01. STR.0000153015.
57892.66
The vascular medicine training application
E. Tripoliti1, A. Sakellarios1, M. Peroulis2, E.G.M. Petrakis1
1
Technical University of Crete (TUC), Department of Electronic
and Computer Engineering, Chania, Greece
2
University Hospital of Ioannina, Department of Surgery, Vascular
Surgery Unit, Ioannina, Greece
Keywords e-Learning
Training scenario
Vascular medicine
Purpose
We present a surgical training application enabling experts in vascular
medicine prepare educational, training content for their trainees
(medical students, endovascular surgeons under training). The train-
ing application offers tutors (experienced surgeons) a distance
learning solution for training their students at any time and any place
Int J CARS (2012) 7 (Suppl 1):S97–S102
with a minimum cost in budget and time. The trainees (future sur-
geons) can use real patient data, process this data (mainly the
available imaging examinations), go through the whole condition
assessment and, finally, take all the important decisions (whether to
surgically treat the patient or not, choose surgical intervention method
etc.) suggested by a training scenario. The training application is
developed as part of the RT3S project [1] of the EU.
Method
The RT3S training application is a novel tool for designing, managing
and delivering online collaborative learning activities. It provides
tutors with a highly intuitive visual authoring environment for cre-
ating sequences of learning activities (i.e., training scenarios).
Learning activities range from individual tasks and small group work,
to class activities. The RT3S training application implements a
modular architecture where tools and learning activities can be added
on-the-fly. The implementation platform is LAMS [2]. The archi-
tecture consists of two main components, namely the Core and the
Activity tools. The Core has modules for Authoring, Administration
and Learner/Trainee roles. The Core manages the elements that form
the structure of the Learning Design [3], the order of the activities,
allocating users to groups, managing when users can progress to
particular activities, takes care of logins, system administration, etc.
The Activity Tools component includes self-contained modules,
which form most of the visual functionality of the content the Learner
interacts with in the RT3S Training Application. Activity tools are
used in order to allow tutors to present information in the form of text,
images, links, flash files and any other HTML features, to create a
series of questions, to add content into a sequence, to run a syn-
chronous chat or forum over the course of a lesson etc.
Results
A realistic scenario (i.e., based on a real patient case) has been
developed demonstrating basic pre-operation steps (before an angio-
plasty operation takes place) that are routinely followed once a patient
has been accepted into the hospital. A limited number of users (rep-
resenting different ages, sexes, and experience with relevant
applications) were selected for the evaluation of the application using
the above scenario. They were provided with a short description of the
application, and then they were allowed to use it for almost an hour.
The users completed an evaluation questionnaire about functionality,
aesthetics, ergonomics, and completeness. Figure 1 illustrates an
overview of the evaluation results. The results from this evaluation
activity were encouraging since most aspects of the application
yielded an evaluation above average that is between 3 and 3.5, with 5
being the highest score, which means that users were satisfied, but
improvements must be made. The evaluation of the proposed training
Fig. 1 An overview of the evaluation results
S101
application also reveals that the application moves in the correct
direction since the scenario-based approach, which was a core deci-
sion taken in the early steps of the development, was indeed a very
good one and is very well-received by the users.
Conclusion
We presented a Web-based open source training application which
provides learners with unlimited access to training opportunities.
More specifically it promotes individual training exercises, facilitates
self-paced practice sessions, and captures case details for objective
assessment. The training application supports endovascular surgeons
at their first steps in the specific competence and their further on-the-
job training. By completing the scenarios they become more familiar
with various cases and prepare themselves for making decisions for
their patients. Overall, it supports trainees in improving their skills
required for competence in vascular medicine.
Acknowledgments
This work was supported by project ‘‘RT3S: Real Time Simulation
for Safer vascular Stenting’’ (project number 248801) of the EU.
References
[1] www. http://www.rt3s.eu/
[2] http://www.lamsinternational.com/
[3] http://www.imsglobal.org/learningdesign/
RadMiner: improved access to radiological free-text data
and images
E. Kotter1, A. Gerstmair1, K. Simon2, C. Simon2, P. Daumke2
1
Universitätsklinikum Freiburg, Abteilung Röntgendiagnostik,
Freiburg, Germany
2
Averbis GmbH, Freiburg, Germany
Keywords Data mining
Radiology report
Free text
Image
content
Purpose
To build and evaluate a retrieval system for radiology reports
enabling improved access to radiological image and free-text data.
Methods and materials
A scalable text analysis platform is used to extract relevant infor-
mation from German radiology reports and to create semantic links
between images and textual descriptions. The platform consists of
domain-specific analysis components dealing with different aspects of
medical language such as negations, abbreviations, dimensions,
proper names and the morphological analysis of word compounds.
Additionally it contains an image reference and context recognition
component allowing the creation of semantic links between text and
image data. Given an explicit picture reference mentioned in a report,
e.g. ‘IMA 12-15’, and its surrounding description this component tries
to deduce the corresponding DICOM series ID based on explicit or
implicit hints and loads the referenced image(s) from the local PACS
server. Finally, key images and relevant text passages are subse-
quently made available to radiologists in a search engine for ad-hoc
querying.
Results
On a corpus of 113.325 radiology reports the system was able to
identify 32.533 image references and successfully retrieve 32.157
images from the local PACS server. Randomized validations of
identified image references showed that the extraction component is
able to resolve the correct reference with mean-average precision of
P:0.97.
Access to the system is made through a dedicated web-interface.
Results are delivered almost immediately and consist of a collection
of reports retrieved by the search algorithm. Significant findings in the
report are highlighted and a mouse-over displays the corresponding
images containing the finding. If access to the complete examination
123
S102
is needed, a direct link to the PACS database and viewer might be
used.
Conclusion
Our analysis shows that comprehensive syntactic and semantic text
analysis is required to achieve high quality information extraction
from radiology reports. It’s the basis for a number of value-added
applications such as image search, data mining and decision support.
The system is used in our hospital since September 2011 and has
proved to be very useful for different applications. It allows easy
collection and review of cases for retrospective studies. It is also very
123
Int J CARS (2012) 7 (Suppl 1):S97–S102
suitable to collect teaching cases for training of students and resident
training. In this aspect, it acts as an addendum to the department’s
teaching file collection. The system also is very useful when a radi-
ologists wants to compare images with other similar cases for
differential diagnosis, i.e. if you suspect a rather seldom diagnosis,
you can look-up the whole RIS database for cases with this diagnosis.
In the future, this technology might also be used for the follow-up
of patients, e.g. in order to give a feedback to the radiologist when-
ever a corresponding pathological report becomes available to a
previous radiological examination.