Content uploaded by Jonas Ivarsson
Author content
All content in this area was uploaded by Jonas Ivarsson on Apr 24, 2017
Content may be subject to copyright.
THE APPLICATION OF IMPROVED, STRUCTURED AND
INTERACTIVE GROUP LEARNING METHODS IN DIAGNOSTIC
RADIOLOGY
Jonas Ivarsson 1,*, Hans Rystedt 1,2, Sara Asplund 3,4, A
˚se Allansdotter Johnsson5,6 and Magnus Ba
˚th3,4
1
Department of Education, Communication and Learning, University of Gothenburg, Gothenburg, Sweden
2
Department of Teacher Education, Universityof Turku, Turku, Finland
3
Department of Radiation Physics, Institute of Clinical Sciences, Sahlgrenska Academy,
Universityof Gothenburg, Gothenburg, Sweden
4
Department of Medical Physics and Biomedical Engineering, SahlgrenskaUniversity Hospital,
Gothenburg, Sweden
5
Department of Radiology, Institute of Clinical Sciences, Sahlgrenska Academy,University of Gothenburg,
Gothenburg, Sweden
6
Department of Radiology, Sahlgrenska University Hospital, Gothenburg, Sweden
*Corresponding author: jonas.ivarsson@gu.se
This study provides an example on how it is possible to design environments in a diagnostic radiology department that could
meet learning demands implied by the introduction of new imaging technologies. The innovative aspect of the design does not
result from the implementation of any specific tool for learning. Instead, advancement is achieved by a novel set-up of existing
technologies and an interactive format that allows for focussed discussions between learners with dierent levels of expertise.
Consequently, the study points to what is seen as the underexplored possibilities of tailoring basic and specialist training that
meet the new demands given by leading-edge technologies.
INTRODUCTION
One persistent dilemma for education is the struggle
to provide up-to-date training tailored for high-end
technologies. The ever-increasing pace in which such
technologies are implemented means that compe-
tences needed for their successful use have to be devel-
oped before any formal education programmes exist.
Consequently, there is a need for designing work-
place-learning environments in which both experi-
enced professionals and novices can develop and
improve essential skills
(1)
. The present study investi-
gates an approach adopted in a radiology depart-
ment, which was aimed at improving the detection of
pulmonary nodules arising from the application of a
novel medical imaging technology (tomosynthesis)
(2)
.
The main purpose of this study was to investigate the
suggested approach in terms of the desirable qualities
involved in developing more professional modes of
image analysis and reasoning among novices in the
field.
For many years, modern medicine has been fused
with—and is thereby partly dependent on—a vast
number omaging technologies. The skills needed for
interpreting various forms of medical images are thus
at the core of diagnostic reasoning, and there is
demand for developing systematic and eective
methods for training such skills
(3)
. An associated
challenge concerns the ongoing and ever-expanding
developments in imaging technology that can rapidly
change the very basis of diagnostic reasoning
(4)
.
When the foundation of diagnostic methods is
aected by the introduction of new technology, even
experienced radiologists need to re-calibrate their
analytical methods and interpretative skills
(5,6)
.
A common approach in advanced training for the
professions is the application of some kind of appren-
ticeship model in which students work with and ob-
serve experienced clinicians as they act in real settings.
This approach to learning possesses many advan-
tages, but is not free from shortcomings. The general
apprenticeship model builds upon an epistemic asym-
metry between two nodes where the knowledge is
both fixed in scope and owned by the master. Even
if this configuration does enable novice learning, it
provides little room for the entire organisation to
develop its accumulated knowledge base in the face of
changed conditions
(7)
.
In search for more advanced pedagogical methods
for the training of professionals, the research group
has aimed to design eective training models that
can address the learning needs of students, as well as
faculty, with varying degrees of experience in radi-
ology. This undertaking has addressed the needs for
developing forms oearning that are able to keep up
with the rapid pace of technological development
and enable successful use of such innovations before
any formal education has been put in place. The
work draws upon existing research into methods for
Radiation Protection Dosimetry (2016),Vol. 169, No. 1-4, pp.416-421 doi:10.1093/rpd/ncv497
teaching professional modes of reasoning by promot-
ing active involvement by all participants in problem-
solving processes
(8)
.
Previous studies have comprised a multitude of
approaches, such as ‘Clinical Reasoning Theater’
(9)
,
educational rounds
(10)
and computer-supported case-
based learning
(11)
. These studies clearly indicate that
the demonstration of expert reasoning in examples of
positive diagnosis and the opportunities for learners
to discuss and reflect upon the process of reasoning
are key factors for successful learning. Similarly, the
use of real cases has been shown to be advantageous
in representing the variation and ambiguity of clinical
symptoms. There is also support for the importance
of immediate feedback and exposure to diagnostic
reasoning early in the learning curve
(12)
. One further
finding of particular interest concerns the use of
shared visual fields that are open to manipulation.
When striving to clarify professional reasoning, this
type of presentation can encourage group participa-
tion and associated collective knowledge that may be
modified through group interactions aimed towards a
common goal
(13)
.
For novices, as well as for experts in the field of
radiology, it is often demanding to interpret radio-
logical section images. Since these images consist of
greyscale images in different planes, exclusively based
on the differences in attenuation of tissues, the inter-
pretation of what they represent is far from straight-
forward. Among other things, this involves how to
identify objects and how these correspond to various
anatomical structures. Furthermore, if a suspected
object is identified in the two-dimensional image, it
can be hard to judge where it is located in the three-
dimensional body. Another difficulty is to assess if
such objects are to be regarded as pathological or as a
variation of the normal anatomy.
Based on the reviewed pedagogical principles, what
became called a Technology-enhanced Learning Session
(TLS) was developed. The design involved a set-up of
imaging technologies and an interactive format that
was intended to facilitate discussions between experi-
enced radiologists and novices in the field. The primary
aim of the work carried out during the TLS was to
improve diagnostic accuracy following the introduction
of tomosynthesis at a thoracic radiology department
and to identify potential pitfalls regarding nodule de-
tection in this new modality. The secondary aim was to
investigate to what extent and in what ways the TLS
could simultaneously support novice learning.
The present article is focussed on the secondary
aim of the TLS and starts from observations made in
a prior study that investigated the effects of the TLS
on the detection of pulmonary nodules. In this initial
study, a number of observers with varying experience
of chest tomosynthesis analysed tomosynthesis cases
for the presence of nodules. The same tomosynthesis
cases were analysed before and after the TLS, and the
difference in performance between the two readings
was calculated. The results showed significant improve-
ments in performance after the TLS for observers inex-
perienced in tomosynthesis
(14)
. Whilst this analysis
provided evidence that the intervention indeed had
been successful, it gave few insights into why this was
the case. The present study, therefore, re-visited the pre-
viously collected data, but instead of addressing the
outcomes, as measured byobserver performance, a dif-
ferent set of questions were posed. By grounding the
new analysis in video recordings of the actual process,
the study first addresses ways in which the TLS could
display and instruct inexperienced observers in profes-
sional modes of reasoning. Secondly, it opens up a dis-
cussion on the ways the findings can be generalised to
inform the design of basic and specialist training in-
volving new imaging technologies.
DESIGN OF THE TECHNOLOGY-ENHANCED
LEARNING SESSION
The TLS was designed by an interdisciplinary research
group from the departments of radiation physics and
radiology in cooperation with researchers in the learn-
ing sciences. The design of the session was prefaced on
the assumption that more deliberate and systematic
methods for exploring the criteria for making judge-
ments would improve diagnostic accuracy.
To support this, there was a preparatory phase of
the TLS where the observers first performed individ-
ual assessments of nodules in tomosynthesis examina-
tions and thereafter received individual feedback on
their results. As a second phase, there was a collective
review session. Here, the individual results of the
observers as well as the reference answers and the
corresponding tomosynthesis and computed tomog-
raphy (CT) images were displayed on large screens (see
Figure 1). One radiologist responsible for the reference
Figure 1. The CT and tomosynthesis images are projected
on a screen in front of the room.
J. IVARSSON ET AL.
Page 2 of 6
by guest on January 4, 2016http://rpd.oxfordjournals.org/Downloaded from
answers navigated the CT images, and another re-
searcher navigated the results of the tomosynthesis
assessments. For each assessment, the observers were
asked to give their reasons for making false positives,
false negatives or giving low ratings to true nodules.
In order to assess which qualities of the TLS facili-
tated improved professional modes of reasoning, an
analysis that could account for how the observers pro-
duced this in practice was needed. It should be made
clear that the analysis did not seek to explain factors
affecting learning, but rather targets those enabling
conditions or qualities of the learning process that
create the opportunities for improved professional
judgements. By investigating the consistencies in how
these judgements were disseminated to those involved,
including the less experienced observers, a number of
critical conditions were highlighted.
ANALYSIS OF THE TECHNOLOGY-
ENHANCED LEARNING SESSION
The TLS study on how best to judge radiological section
images produced by a new technology highlighted three
interrelated conditions of the TLS approach to develop-
ing professional diagnostic judgements. These condi-
tions are presented and discussed.
Juxtaposing past and present actions
In order to understand how the TLS could contribute
to an understanding of the ways in which experienced
radiologists make their decisions ( for the benefits of
novices), it is important to outline how the session,
and the technologies involved, were arranged. One
significant premise was that the results of each indi-
vidual rating and the results of the reference method
were displayed in a field next to the projected tomo-
synthesis image (see Figure 2). Another premise was
that these appeared simultaneously as the annotated
tomosynthesis section image was shown. Through
this specific arrangement, each individual answer
could be instantly compared with (1) the ‘true’ answer
(is this a nodule or not), (2) the results of the other
observers’ answers and (3) the object in the image to
which the classification referred. Technically, this was
made possible using the visualisation software
ViewDEX
(15 –17)
.
Epistemic asymmetry and accounting practices
There were several implications of the particular
juxtaposition of outcomes. First, both the correctness
and incongruences of earlier answers were effectively
displayed. Every single mistake was thus established
as an inescapable fact open for general review. As a
result, when the past actions were shown to be incor-
rect, this called for some kind of ‘corrective action’.
Frequently when one of the more experienced radiol-
ogists was responsible for an incorrect outcome, they
would initiate an exploration of the grounds for their
own assessment. Typically, such accounts started with
statements such as ‘is this on me?’ or ‘I’m responsible’
and would then develop into possible ways to analyse
the specific case. For the benefit of the less experienced
observers, these extended accounts would act as models
of advanced diagnostic reasoning, typically pertaining
to borderline cases or particularly difficult areas.
These accounts also reveal one of the more sensi-
tive aspects of the TLS. As a part of the design, there
was an epistemic asymmetry in the group of obser-
vers, with four experienced specialists in thoracic radi-
ology and three non-specialists. For professionals, the
opening up of their actions to scrutiny by peers, and
even less experienced participants, could be uncom-
fortable and requires a large degree of mutual trust.
Even if this practice may be contrary to expected
hierarchy and normal processes, it was perceived to be
an important element when striving for improving
Figure 2. The visual representations used during the TLS. CT (left), TS (middle) and compiled answers (right).
INTERACTIVE GROUP LEARNING
Page 3 of 6
by guest on January 4, 2016http://rpd.oxfordjournals.org/Downloaded from
professional diagnostic capabilities in general. In this
way, professional judgements were not only being
challenged, but it was also being clearly demonstrated
and analysed by discussion to what standard and by
which methods the group would deem any judgement
valid for the clinical application of a new technology.
The inclusion of inexperienced observers also resulted
in additional mistakes in the batch of cases to be
discussed. Rather than regarding this as a flaw in the
design, it should be considered an important resource.
As a consequence, a large number of cases became
objects for clarification that required further explanation.
Such explanations were implemented, i.e. communica-
tively designed, in ways that were also comprehensible
for non-specialists in the field. Consequently, knowl-
edge/awareness, that under normal conditions would be
taken for granted in communication between specialists,
now had to be articulated.
Moreover, as the TLS arrangement provides an op-
portunity for elaborating the reasons for all mistakes,
the discussion was not limited to a selection of cases
regarded as typical or noteworthy from some pre-
given point of view. In this way, abroad range of prob-
lematic interpretations associated with the variability
of human anatomy and pathologies were discussed.
Throughout the session, a number of interpretative
pitfalls were thus found and formulated.
Coordination work displayed
A central constituent of the TLS for making judge-
ments public was the simultaneous use of two separ-
ate projector screens. All participating observers
(equipped with laser pointers) were facing these
screens, which served as a shared point of reference.
The fact that two projections were arranged side by
side allowed for comparisons of the CT and tomo-
synthesis images and enabled constant assessments of
each highlighted structure and the corresponding
structure as it appeared in the reference method.
However, since the systems did not have comparable
coordinate systems that could define the exact corre-
sponding localisation, comparisons were not immedi-
ately available. Furthermore, due to the two different
types of imaging technologies involved, the same
structure would look dissimilar in the two images.
Adding to this, the tomosynthesis images were only
available in the frontal (coronal) plane, whereas the
preferred view of the CT scans was the transversal
section images (see Figure 1).
Although these differences might be interpreted as
putting extra demands on the observers, it was found
that the work needed to organise the two imaging
systems in alignment constituted important grounds
for uncovering implicit radiological reasoning. This
work necessitated concerted effort by the two partici-
pants controlling the computers as well as the group
of observers. New cases were first opened up in the
tomosynthesis material. Although the radiologist re-
sponsible for the reference method had noted in
advance where the corresponding structure should be
located in the CT images, additional effort was
required in order to determine the exact section image
that best matched the highlighted structure. The
group of observers regularly guided this coordination
effort. Moreover, it was commonly requested to also
see the adjacent sections in the tomosynthesis stack,
which did not always coincide with the highlighted
region. In this way, it became possible to discern in
which section the structure was most clearly visible,
and thereby, to establish a more precise perception of
its localisation. Through this manual calibration, a
solid relation between the two technologies was
reached, making sure that the two representations
corresponded to the same object.
A central component in the diagnostic work of
finding suspected nodules is to grasp their precise loca-
tion. Thereby, the work of aligning the two representa-
tions fulfilled several purposes. Most importantly, it
clarified the means through which decisions about
localisation are made. In this work, a number of
resources central to the practitioners’ diagnostic rea-
soning came into view, such as knowledge about
anatomy for navigating the regions of interest, search
routines for reading radiological section images by
scrutinising sequences of images and methods for pre-
cisely pinpointing significant features of the lung.
DISCUSSION
The design of the TLS built on earlier work pinpoint-
ing a number of essential features of successful learn-
ing environments: the demonstration of expert
reasoning, the use of authentic cases, the sharing of a
reference space for learning, and opportunities for
discussion and reflection
(8–10)
. Even if these findings
acted as an introduction for the current study, the new
results expand on the understanding of their applica-
tion. In addition, the present analysis clarifies the out-
comes of the previous study by the research group
(14)
.
This demonstrated that inexperienced observers
improved their detection skills after taking part in the
TLS. In highlighting some interesting features of the
TLS, the present analysis adds to the understanding
of how such improvements were reached.
One major outcome of the analysis is the central
role played by broad-based visibility. That is, the visi-
bility both of diagnostic reasoning in action and the
material grounds on which this reasoning was based.
In the TLS, competent detection was made transpar-
ent and turned into instructions for the less experi-
enced. The problem of coordinating the two image
representations is a case in point. A technological ap-
proach to the matter would probably suggest that this
task should be automated and handed over to the
technology. On the contrary, the findings show that
J. IVARSSON ET AL.
Page 4 of 6
by guest on January 4, 2016http://rpd.oxfordjournals.org/Downloaded from
such a move would come at a pedagogical cost. The
work needed to perform the calibration was exactly
what brought radiological reasoning into view.
Through this work, the reasoning and skills of the
experienced observers, particular and domain specific
as they are, became evident features in the TLS, i.e.
things to be seen and learned.
One part of the educational design was the inclu-
sion of preparatory work by each observer. This
meant that much time and effort had been invested in
working through the cases. It was also known that the
assessments would be subjected to scrutiny under a
collective regime at a later stage. Arguably these expe-
riences would serve as grounds for taking part in the
ensuing discussions. Through this relation between
past and present, the TLS targeted diagnostic report-
ing practices. The use of openly available records of
past actions triggered a number of reports—extensive
explanations and close examinations of mistakes and
their corrections. Given the difference between obser-
vers, the comments provided had to be constructed in
ways that displayed sensitivity to the variation in par-
ticipants’ experience and enabled everyone to take
part. This detailed level of description was additional-
ly motivated by the presence of the new tomosynth-
esis technology for medical imaging. In the process of
reaching consensus, tacit understandings or things
that had been taken for granted were revisited and
collectively reviewed. Again in relation to the less
experienced observers, this discussion served as a
form of elaborate and detailed instruction in active
professional reasoning.
Implications for design
In the present work, the success was created by the in-
tegration of novel imaging technologies in the work-
place with specific forms of interaction. Based on the
analysis of outcomes, three principles are proposed—
ideas to be considered when designing learning envir-
onments for teaching professional modes of reasoning
in radiology:
†Ensure accountability of past and present actions:
The ways in which participants with different
levels of experience interact and communicate
have a large impact on the outcome of the activity.
By displaying records of past actions, everyone
can become involved and mistakes become dis-
sected rather than hidden.
†Exhibit work in action: Experts working on authen-
tic cases give prominence to case-specific details, dis-
ambiguation practices, and several dimensions of
variation (in representations, anatomy, pathology
etc.). Professional modes of reasoning, when being
made publically visible, then operate as instructions.
†Provide participants with shared access to visual
materials: Given different set-ups, participants
will have different possibilities of establishing
shared references and partake in reasoning that
can build on visual details. As was noted, the
observers’ ability to notice, discuss and investigate
particular features of the radiological images
became a necessary requirement for the accom-
plishment of their collaborative group working.
FUNDING
This work was supported by the University of
Gothenburg Learning and Media Technology Studio,
the LearnMedImage project (the Academy of Finland,
128766), the Swedish Research Council (2010-5105,
2011-488, 2013-3477), the Swedish Radiation Safety
Authority (2012-2021, 2013-2982), the King Gustav V
Jubilee Clinic Cancer Research Foundation (2008:50),
the Health & Medical Care Committee of the Region
Va
¨stra Go
¨taland (VGFOUREG-81341, VGFOUREG-
483951) and the Swedish Federal Government
under the LUA/ALF agreement (ALFGBG-136281,
ALFGBG-428961).
REFERENCES
1. Billett, S. and Choy, S. Learning through work.
Emerging perspectives and new challenges. JWL 24,
264–276 (2013).
2. Vikgren, J., Zachrisson, S., Svalkvist, A., Johnsson, A
˚.
A., Boijsen, M., Flinck, A., Kheddache, S. and Ba
˚th, M.
Comparison of chest tomosynthesis and chest radiography
for detection of pulmonary nodules: human observer study of
clinical cases. Radiology 249, 1034– 1041 (2008).
3. O’Brien, K. E., Cannarozzi, M. L., Torre, D. M.,
Mechaber, A. J. and Durning, S. J. Training and assess-
ment of CRX/basic radiology interpretation skills:
results from the 2005 CDIM survey. Teach. Learn. Med.
20, 157– 162 (2007).
4. Mouratev, G., Howe, D., Hoppmann, R., Poston, M. B.,
Reid, R., Varnadoe, J., Smith, S., McCallum, B., Rao, V.
and DeMarco, P. Teaching medical students ultrasound
to measure liver size: Comparison with experienced clini-
cians using physical examination alone. Teach. Learn.
Med. 25, 84–88 (2013).
5. Lymer, G., Ivarsson, J., Rystedt, H., Johnsson, A
˚. A.,
Asplund, S. and Ba
˚th, M. Situated abstraction. From the
particular to the general in second order diagnostic work.
Discourse Stud. 16, 182–212 (2014).
6. Rystedt, H., Ivarsson, J., Asplund, S., Johnsson, A
˚.A.
and Ba
˚th, M. Rediscovering radiology. New technologies
and remedial action at the worksite. Soc. Stud. Sci. 41,
867–891 (2011).
7. Greenfield, P. M. Historical change and cognitive
change: a two-decade follow-up study in Zinacatan, a
Maya community in Chiapas, Mexico.MCA6, 92–108
(1999).
8. Jacobson, K., Fisher, D. L., Hoffman, K. and Tsoulas,
K. D. Integrated case section: a course designed to
promote clinical reasoning in year 2 medical students.
Teach. Learn. Med. 22, 312 –316 (2010).
INTERACTIVE GROUP LEARNING
Page 5 of 6
by guest on January 4, 2016http://rpd.oxfordjournals.org/Downloaded from
9. Borleffs, J. C., Custers, E. J., van Gijn, J. and ten Cate,
O. T. ‘‘Clinical reasoning theater’’: a new approach to
clinical reasoning education. Acad. Med. 78, 322– 325
(2003).
10. Nendaz, M. R., Junod, A. E., Vu, N. V. and Bordage, G.
Eliciting and displaying diagnostic reasoning during educa-
tional rounds in internal medicine: who learns from whom?
Acad. Med. 73, 54– 56 (1998).
11. Kourdiokova, E. V., Verstraete, K. L. and Valcke, M.
The quality and impact of computer supported collabora-
tive learning (CSCL) in radiology case-based learning.
Eur. J. Radiol. 78, 353– 362 (2011).
12. Kassirer, J. P. Teaching clinical reasoning: case-based and
coached. Acad. Med. 85, 1118– 1124 (2010).
13. Henderson, K. Flexible sketches and inflexible data
bases: visual communication, conscription devices, and
boundary objects in design engineering. STHV 16,
448–473 (1991).
14. Asplund, S. et al.Learning aspects and guidelines regard-
ing detection of pulmonary nodules and developing
quality criteria for chest tomosynthesis. Acta Radiol. 52,
503–512 (2011).
15. Bo
¨rjesson, S. et al.A software tool for increased effi-
ciency in observer performance studies in radiology.
Radiat. Prot. Dosim. 114, 45 –52 (2005).
16. Ha
˚kansson, M., Svensson, S., Zachrisson, S., Svalkvist,
A., Ba
˚th, M. and Ma
˚nsson, L. G. ViewDEX: an efficient
and easy-to-use software for observer performance
studies. Radiat. Prot. Dosim. 139, 42– 51 (2010).
17. Svalkvist, A., Svensson, S., Ha
˚kansson, M., Ba
˚th, M.
and Ma
˚nsson, L. G. ViewDEX: a status report. Radiat.
Prot. Dosim. doi:10.1093/rpd/ncv543.
J. IVARSSON ET AL.
Page 6 of 6
by guest on January 4, 2016http://rpd.oxfordjournals.org/Downloaded from