Magnetic resonance imaging for the detection, localisation, and characterisation of prostate cancer: recommendations from a European consensus meeting.

Abstract

Multiparametric magnetic resonance imaging (mpMRI) may have a role in detecting clinically significant prostate cancer in men with raised serum prostate-specific antigen levels. Variations in technique and the interpretation of images have contributed to inconsistency in its reported performance characteristics.
Our aim was to make recommendations on a standardised method for the conduct, interpretation, and reporting of prostate mpMRI for prostate cancer detection and localisation.
A consensus meeting of 16 European prostate cancer experts was held that followed the UCLA-RAND Appropriateness Method and facilitated by an independent chair.
Before the meeting, 520 items were scored for "appropriateness" by panel members, discussed face to face, and rescored.
Agreement was reached in 67% of 260 items related to imaging sequence parameters. T2-weighted, dynamic contrast-enhanced, and diffusion-weighted MRI were the key sequences incorporated into the minimum requirements. Consensus was also reached on 54% of 260 items related to image interpretation and reporting, including features of malignancy on individual sequences. A 5-point scale was agreed on for communicating the probability of malignancy, with a minimum of 16 prostatic regions of interest, to include a pictorial representation of suspicious foci. Limitations relate to consensus methodology. Dominant personalities are known to affect the opinions of the group and were countered by a neutral chairperson.
Consensus was reached on a number of areas related to the conduct, interpretation, and reporting of mpMRI for the detection, localisation, and characterisation of prostate cancer. Before optimal dissemination of this technology, these outcomes will require formal validation in prospective trials.

Full text

Page 1

Platinum Priority – Prostate Cancer
Editorial by Axel Heidenreich on pp. 495–497 of this issue
Magnetic Resonance Imaging for the Detection, Localisation, and
Characterisation of Prostate Cancer: Recommendations from a
European Consensus Meeting
Louise Dickinsona,b,c,*, Hashim U. Ahmeda,b, Clare Allend, Jelle O. Barentsze, Brendan Careyf,
Jurgen J. Futterere, Stijn W. Heijminke, Peter J. Hosking, Alex Kirkhamd, Anwar R. Padhanih,
Raj Persadi, Philippe Puechj, Shonit Punwanid, Aslam S. Sohaibk, Bertrand Tomball,
Arnauld Villersm, Jan van der Meulenc,n, Mark Embertona,b,c
aDivision of Surgery and Interventional Science, University College London, London, UK
bDepartment of Urology, University College London Hospitals NHS Foundation Trust, London, UK
cClinical Effectiveness Unit, Royal College of Surgeons of England, London, UK
dDepartment of Radiology, University College London Hospitals NHS Foundation Trust, London, UK
eDepartment of Radiology, Radboud University Medical Centre, Nijmegen, The Netherlands
fDepartment of Radiology, Leeds Teaching Hospitals NHS Trust, Leeds, UK
gDepartment of Oncology, Mount Vernon Cancer Centre, Middlesex, UK
hDepartment of Radiology, Mount Vernon Cancer Centre, Middlesex, UK
iDepartment of Urology, University Hospitals Bristol NHS Foundation Trust, Bristol, UK
jDepartment of Radiology, Hospital Claude Huriez, Universitaire Lille Nord de France, F-59000, Lille, France
kDepartment of Radiology, Royal Marsden NHS Foundation Trust, London, UK
lDepartment of Urology, Cliniques Universitaires Saint-Luc, Brussels, Belgium
mDepartment of Urology, Hospital Claude Huriez, Universitaire Lille Nord de France, F-59000, Lille, France
nHealth Services Research Unit, London School of Hygiene and Tropical Medicine, London, UK
EURO PEAN UROLOGY 59 ( 2011) 477–494
available at www.sciencedirect.com
journal homepage: www.europeanurology.com
Article info
Article history:
Accepted December 10, 2010
Published online ahead of
print on December 21, 2010
Keywords:
Consensus methods
Multiparametric MRI
Prostate cancer
Abstract
Background: Multiparametric magnetic resonance imaging (mpMRI) may have a
role in detecting clinically significant prostate cancer in men with raised serum
prostate-specific antigen levels. Variations in technique and the interpretation of
images have contributed to inconsistency in its reported performance character-
istics.
Objective: Our aim was to make recommendations on a standardised method for
the conduct, interpretation, and reporting of prostate mpMRI for prostate cancer
detection and localisation.
Design, setting, and participants: A consensus meeting of 16 European prostate
cancer experts was held that followed the UCLA-RAND Appropriateness Method
and facilitated by an independent chair.
Measurement: Before the meeting, 520 items were scored for ‘‘appropriateness’’
by panel members, discussed face to face, and rescored.
* Corresponding author. Department of Urology, 250 Euston Road, London, NW1 2BU, UK.
Tel. +44 207 380 9194; Fax: +44 207 3809303.
E-mail address: ldickinson@doctors.org.uk (L. Dickinson).
0302-2838/$ – see back matter # 2010 European Association of Urology. Published by Elsevier B.V. All rights reserved.doi:10.1016/j.eururo.2010.12.009

Page 2

1.Introduction
Magnetic resonance imaging (MRI) has considerable poten-
tialtoimprovethe prostatecancerdiagnostic pathway. Until
fairly recently, the accuracy of morphologic MRI to detect,
localise,andcharacteriseprostatecancerswaslimited,andas
aresult,MRIhasnotbeenroutinelyincorporatedintoclinical
care. However, evidence is accumulating that suggests an
improved performance of MRI, provided that modern
sequences are used and their outputs combined in so-called
multiparametric MRI (mpMRI). Currently these include T1-
and T2-weighted images, dynamic contrast, diffusion
weighting, and proton spectroscopy [1,2].
Although experts in the field generally regard the
performance characteristics of mpMRI of the prostate as
promising [3], there exists professional disagreement on its
accuracy and usefulness in clinical practice [4], limiting
wider adoption. These concerns relate in part tothe variable
quality and methodology of studies that have resulted in
marked variation in indication, conduct, interpretation, and
reporting [5–11]. These issues have made it difficult to
summarise the literature in any meaningful way [1].
This problem is not a new one. Over the last 2 decades,
breast cancer experts have had to manage similar issues in
relation to x-ray mammography and, more recently, breast
MRI [12,13]. The solution to this problem was formal
attempts to establish agreement among experts on areas
of uncertainty. As a result, a series of recommendations
emerged on the minimum standards acceptable for mam-
mography [14–17] and breast MRI [18,19]. The principal
innovationprovedtobetheincorporationofscoringsystems
tocommunicatethelikelihoodofmalignancyinwomenwith
suspected breast cancer. These recommendations conferred
at least two benefits: reduced interobserver variability
[20,21] and improved positive predictive value for obtaining
pathology from breast biopsy [21].
Based on the breast cancer experience, it seemed timely
and necessary to see if experts in the field of prostate cancer
and prostate MRI could achieve similar consensus. This
paper reports the recommendations of a panel of urora-
diology experts who participated in a formal consensus
process aimed at defining when prostate MRI should be
applied and how it should be conducted and reported.
2.Materials and methods
2.1. The consensus method
A number of formal consensus methods have been used in health care
settings [22,23]. The RAND-UCLA Appropriateness Method (RAM) was
chosen as the most appropriate for our objectives [24]. RAM includes a
combination of postal and face-to-face consensus rounds. It is most
suited to topic areas where there is little or poor quality evidence to
enable a gold standard recommendation. Appropriateness levels are
used to communicate the perceived balance between risks and benefits
of each item under discussion. The RAM User’s Manual was followed
throughout the process [24].
2.2.Panel selection
Leading clinicians from the United Kingdom, France, Belgium, and the
Netherlands with known subspecialty expertise in prostate cancer
diagnostic imaging were approached, all prominent European members
published in the field. The chair, Jan van der Meulen, is a clinical
epidemiologist with experience using formal consensus methods to
developclinicalguidelines.Commercialbiaswasexcludedthroughtheuse
of institutional funds.
2.3. Construct of the questionnaire
A questionnaire containing 537 items was constructed between August
and November 2009. The first draft was produced by panel members
Louise Dickinson, Mark Emberton, Hashim U. Ahmed, Clare Allen, Shonit
Punwani, Alex Kirkham, and Anwar R. Padhani. Refinements were made
through consultation with the other panel members.
The questionnaire was designed to address the minimum and
optimum imaging requirements for performance, interpretation, and
deliveryofresultsforprostateMRI,includingT1-weighted,T2-weighted,
diffusion-weighted, dynamic contrast-enhanced, and proton spectros-
copy. These aspects related to the localisation and detection of prostate
cancer in men with suspected or known cancer in the absence of biopsy
artefact (ie, before biopsy or at an appropriate time frame after
diagnostic prostate biopsies). It focused on early T staging only (without
breach of the capsule) and did not address lymph node or metastatic
assessments.
Results and limitations: Agreement was reached in 67% of 260 items related to
imaging sequence parameters. T2-weighted, dynamic contrast-enhanced, and
diffusion-weighted MRI were the key sequences incorporated into the minimum
requirements. Consensus was also reached on 54% of 260 items related to image
interpretation and reporting, including features of malignancy on individual
sequences. A 5-point scale was agreed on for communicating the probability of
malignancy, with a minimum of 16 prostatic regions of interest, to include a
pictorial representation of suspicious foci. Limitations relate to consensus meth-
odology. Dominant personalities are known to affect the opinions of the group and
were countered by a neutral chairperson.
Conclusions: Consensus was reached on a number of areas related to the conduct,
interpretation, and reporting of mpMRI for the detection, localisation, and char-
acterisation of prostate cancer. Before optimal dissemination of this technology,
these outcomes will require formal validation in prospective trials.
# 2010 European Association of Urology. Published by Elsevier B.V. All rights reserved.
EUROPEAN UROLOGY 59 (2011) 477–494
478

Page 3

2.4. First-round questionnaire completion before the meeting
Each item was scored on a scale between 1 (inappropriate/strongly
disagree) and 9 (appropriate/strongly agree). A midpoint score of 5
indicated uncertainty. Panel members were asked to provide a score on
all items they considered themselves sufficiently knowledgeable to
answer.
2.5. Meeting format
ThemeetingwasconvenedattheRoyalCollegeofSurgeonsofEnglandfor
a singledayin December 2009. The panellists comprised 16 uroradiology,
urology, and oncology experts. Two neutral observers were in attendance
(Paul Cathcart and Michael Baum) together with the meeting organiser
(Louise Dickinson) to document key points of discussion.
All panel members were sent questionnaires to complete before the
meeting with relevant literature on previous scoring systems used for
breast imaging (a list of articles is available on request). For each
individual item, the scores were presented and discussed during the
consensus meeting, after which the panellists rescored that item.
At given time points during the consensus meeting, panel members
presented on the following topics: current scoring systems in use,
diffusion-weighted MRI, dynamic contrast-enhanced MRI, spectroscopy,
definition of clinically significant cancers, and validation of the scoring
system. Speakers were asked to summarise the evidence in the given
area and to highlight areas of controversy.
Some items were found to be either inadequate or ambiguous. If that
was the case, the item was reworded to improve clarity. Some items
were added after full agreement by panel members and scored during
the meeting.
2.6.Interpretation of the results
The results were interpreted according to the RAM User’s Manual. If <16
panellists scored an item, consensus was defined using the manual’s
recommendations based on that particular panel size. Only those items
scored by at least eight panel members were included in the results.
3. Results
The premeeting questionnaire included 537 items, reduced
to 520 items during the meeting. Nineteen items were
added and 36 items omitted. Wording changes were made
to 20 items.
3.1.Areas of consensus
Consensus was reached in 27% of items in the premeeting
questionnaire, which increased to 61% during the meeting
(Table 1). Postmeeting consensus was reached on 67% of
260 items related to imaging parameters for tumour
detection and localisation, and on 54% of 260 items related
to imaging interpretation and reporting, including features
on individual MRI sequences indicating malignancy. Tables
2–4 summarise the key items of consensus. Table 7 lists the
full postmeeting results for the individual items. Further
discussion took place electronically among all panel
members following the consensus meeting, once the
outcomes of the questionnaires were known.
Table 1 – Proportion of items scored inappropriate, uncertain, or appropriate
Inappropriate
with consensus
Uncertain
with consensus
Uncertain
no consensus
Appropriate
with consensus
Premeeting
Postmeeting
10% (54)
31% (161)
2% (10)
3% (15)
73% (392)
39% (205)
15% (81)
27% (139)
Table 2 – Areas of consensus for general magnetic resonance imaging components
Minimal requirements Optimal requirements
The data set should include T1-weighted, T2-weighted, diffusion-weighted,
and contrast-enhanced MRI but not MR spectroscopy
Imaging could be adequately performed at 1.5 T
A pelvic phased-array coil is required
The data set should include T1-weighted, T2-weighted,
diffusion-weighted, contrast-enhanced MRI
Imaging should be performed at 3 T
A pelvic phased-array coil, endorectal coil, power injector,
and bowel relaxant are required
MR = magnetic resonance; MRI = magnetic resonance imaging.
Table 3 – Areas of positive consensus for disease detection and characterisation for individual magnetic resonance sequences
It is possible to gain the following information from each MR sequence in isolation: MR sequence
T1WT2WDW CEMRSI
Detection of any cancer in the peripheral zone
The Gleason grade of lesions in the peripheral zone
Exclusion of clinically significant disease as defined by a lesion size ?0.2 cm3
(approximately 7 mm) in the peripheral zone
Exclusion of clinically significant disease as defined by a lesion size ?0.5 cm3
(approximately 10 mm) in the peripheral zone
Exclusion of clinically significant cancer according to the definition of a
lesion ?0.5 cm3and/or Gleason ?4 + 3 in the peripheral zone
CE = contrast enhanced; DW = diffusion weighted; MR = magnetic resonance; MRSI = magnetic resonance spectroscopy; T1W = T1 weighted; T2W = T2
weighted; H = areas considered appropriate (positive consensus).
HH
H
H
H
HH
H
E UR O P E A N U R OL O GY 59 ( 2 01 1) 4 7 7 – 4 9 4
479

Page 4

Members of the panel were able to agree on the
following 5-point scale for the scoring of all MRI sequences:
score 1, clinically significant disease is highly unlikely to be
present; score 2, clinically significant cancer is unlikely to
be present; score 3, the presence of clinically significant
cancer is equivocal; score 4, clinically significant cancer is
likely to be present; score 5, clinically significant disease is
highly likely to be present. Clinically significant disease was
defined as Gleason ?4 + 3 and/or lesions ?0.5 cm3in
volume.
Fig. 1 shows the minimal and optimal number of regions
ofinterestandtheirdivisionsasagreedbyconsensus.Asper
the meeting results, it is intended that a 1–5 score be
assigned for each region of interest and for each individual
lesion identified.
The panel agreed that diffusion-weighted imaging is
capable of providing information on Gleason grade and of
excluding lesions of both 0.2 cm3and 0.5 cm3in the
peripheral zone.
3.2. Areas lacking consensus
Several important areas of continued uncertainty or lack of
consensus were identified, some of which are detailed in
Table 5. The value of including MR spectroscopy for optimal
imaging remains uncertain. The use of certain hardware
such as endorectal coils, power injectors, and bowel
relaxants as a minimum standard failed to reach consensus,
although their use was recommended as part of optimum
imaging.
There was no consensus on whether different MR
sequences are able to detect lesions fulfilling other
definitions of clinical significance that are more conserva-
tivethanavolumeof0.5 cm3anddominantGleasonpattern
4 (Table 6).
4. Discussion
4.1. Summary of results
The use of the RAM method produced consensus in 61% of
the items (315 of 520). Consensus was achieved in several
key areas where inconsistency between studies had previ-
ously been a problem. In particular, the panel recommended
that all sequences (T2-weighted, diffusion-weighted, and
dynamic contrast enhanced sequences) except proton
spectroscopy should comprise the minimum standard.
Recent evidence from a large prospective multicentre study
showing no benefit of spectroscopy for prostate cancer
localisation compared with T2-weighted imaging alone
Table 4 – Areas of consensus on imaging interpretation, scoring, and reporting
Areas of positive consensus
When scoring the prostate for the presence or absence of cancer for T2-weighted, diffusion-weighted, contrast-enhanced, and MR spectroscopy sequences,
the range of scores should be 1–5, for each imaging type
Both individual lesions and areas of the prostate should be separately scored for probability of malignancy
The maximum diameter of the largest abnormal lesion should be recorded
The following should be scored for involvement with an individual scoring range of 1–5:
- Extracapsular extension
- Seminal vesicles (extra- and intraprostatic)
- Distal sphincter
- Rectal wall
- Neurovascular bundles
- Bladder neck
As a minimum requirement, the prostate should be divided into 16 regions of interest (apical, mid, and base quadrants) and, as an optimum
requirement, into 27 regions of interest
The ADC value should be stated for any suspicious lesion detected
Dynamic contrast-enhanced MRI should be scored according to the morphological enhancement pattern
The following clinical information is important for reporting the imaging and should be included:
- PSA level
- Digital rectal examination
- Time scale since prostate biopsies and results of previous biopsies
- Results of previous MRI scans
- History of previous prostate treatment or intervention (eg, TURP, prostate radiotherapy)
- History of medical treatment (eg, 5a-reductase inhibitors, hormones)
As a minimum requirement, each MRI should be assessed and scored by one radiologist and, as an optimal requirement, scored by two radiologists
independently and discrepancies referred for consensus
If one of the modalities within the minimum data set is noninterpretable due to artefact, the denominator of the scoring system should be changed
to allow for a lack of score for the affected sequence
Dedicated software for imaging interpretation should be developed for this purpose with the ability to display, co-register, segment, fuse, and analyse
every tool in an integrated single work space
The final report should be presented electronically, in both number and picture form, and should include relevant images
ADC = apparent diffusion coefficient; MR = magnetic resonance; MRI = magnetic resonance imaging; PSA = prostate-specific antigen; TURP = transrectal
resection of the prostate.
EUROPEAN UROLOGY 59 (2011) 477–494
480

Page 5

supports this recommendation [25]. Consensus was also
reached in the area of reporting. A 5-point scoring scale
should be used in allreports to communicate the probability
of prostate cancer. However, several areas were identified
where no consensus could be derived, either due to lack of
evidencetoinformopinionorthepaneldecidingthatcurrent
evidencewassufficientlycontroversialtopermitprofession-
al disagreement. For example, endorectal coil use, an area of
currentclinicalinconsistency,wasrecommendedforoptimal
practice,butnoconsensuswasreachedonitsuseinstandard
(minimal) practice.
4.2. Methodologic limitations
Despite achieving high levels of concordance in key areas,
some caution is called for. Face-to-face consensus methods
or expert group discussions are prone to biases. Dominant
personalities are known to be capable of influencing scoring
to a significant extent [26]. During our meeting a neutral
chairperson, who had experience running formal consensus
meetings, moderated the discussions.
One of the benefits of the group approach that does not
arise in postal processes is that the group can agree on poor
[()TD$FIG]
Fig. 1 – (A) Sixteen regions/sectors standardised magnetic resonance imaging (MRI) prostate reporting scheme. Posteriorly (p), average axial sections at
prostate base and midgland are subdivided into four regions (midlobar and lateral) and at the prostate apex into two regions. Anteriorly (a), prostate
base, midgland, and apex are divided into two regions. The anterior region starts 17 mm from the prostatic posterior surface (biopsy core length). A 10-
core extended biopsy scheme would be expected to sample the 10 posterior sectors. (B) Twenty-seven regions/sectors standardised MRI prostate reporting
scheme. Posteriorly (p), average axial sections at prostate base, midgland, and apex are subdivided into four regions (midlobar and lateral). Anteriorly, the
prostate is divided into four anterior regions (a) (midlobar and lateral) and three anterior stroma regions (as). The anterior region starts 17 mm from the
prostatic posterior surface (biopsy core length). A 12-core extended biopsy scheme would be expected to sample the 12 posterior sectors.
Table 5 – Areas lacking consensus on imaging interpretation, scoring, and reporting
Areas of the prostate should be scored separately rather than by individual lesions
The overall score for probability of tumour given by the radiologist should be influenced by other clinical results (eg, PSA level)
The overall score should be based purely on imaging appearances
A separate radiologist’s ‘‘hunch’’ score should be given that represents the radiologist’s personal overall hunch view on the likelihood of malignancy
regardless of the objective radiologic score
The final score should be given as individual scores, a sum of the individual scores, or as a radiologist’s overall opinion score
T staging should be a formal part of the final report
PSA = prostate-specific antigen.
E UR O P E A N U R OL O GY 59 ( 2 01 1) 4 7 7 – 4 9 4
481

Page 6

or imprecise questions and agree on amendments. The
group structure also meant that all panellists were exposed
to the latest evidence, and as a result the consensus
approach had high face validity.
The other problem worth identifying relates to the
European versus US perspective. There are clear differences
in the conduct of MRI between the two. The main one is that
endorectal coils and spectroscopy are used more frequently
in the United States. Our consensus is largely European,
although we did make great efforts to include the
worldwide literature. Of relevance to spectroscopy, only a
fewpanellists chose toscoreseveral questions duetolackof
clinical experience and interpretation software. As a result,
these could not be analysed.
Finally, our recommendations were based on published
studies of limited clinical evidence (four or fewer on the
Oxford Centre for Evidence-Based Medicine levels of
evidence), a problem previously shared by the breast
imaging community and demanding the use of formal
methodology to derive consensus opinion.
4.3. Clinical implications
The ability to detect, locate, and characterise prostate
cancers radiologically has significant implications for the
patient diagnostic pathway. It is hoped that widespread
incorporation of the recommendations of this meeting will
allow a more consistent and standardised approach to MRI.
The consequence of this will be that reports from different
centres might be amenable to pooled analyses.
If the accuracy of mpMRI demonstrates levels that
warrant widespread adoption, we may have the beginnings
of a triage test that, by ruling out the presence of clinically
significant disease, could result in fewer men needing to
undergo biopsy. The incorporation of mpMRI before biopsy
in men with suspected prostate cancer is currently being
performed in a few centres. The benefits of this approach
have been discussed elsewhere [27]. In addition, the
detection and localisation components of such a test could
assist in targeting biopsies, a strategy that has well served
other cancer detection pathways.
Imaging in the form of mpMRI is likely to have important
roles in the tissue-preserving strategies of active surveil-
lance [28] and focal therapy [29]. These two therapies differ
slightly in their requirements from the pure diagnostic
challenge because mpMRI would need to be repeated over
time to detect interval change. This process will demand
high levels of reliability, which can only be achieved once
standardised conduct and reporting have been agreed on
and implemented.
4.4. Research implications
We do not regard our consensus statements as an end
product. We hope and expect that they will stimulate and
provoke other groups to adapt and improve on our outputs.
It is key that research be prioritised to areas with little or no
consensus. To both integrate and validate the recommen-
dations from this meeting in future imaging protocols and
research, they must be embedded in prospective trials. The
ideal population would be men with a raised serum
prostate-specific antigen (PSA) who undergo prostate
mpMRI before histologic verification with biopsies. Thus
verification or work-up bias would be limited and the at-
risk population evaluated. This will require an accurate
reference standard that can be applied to this population.
The random and systematic errors inherent in transrectal
ultrasound (TRUS) guided biopsy rule this out as a
reference standard. Most studies currently use whole-
mount prostatectomy specimens as the reference standard.
This introduces work-up bias. Furthermore, correlation of
MRI regions of interest with equivalent regions on
prostatectomy specimens is inconsistent. Transperineal
template mapping biopsies are probably the optimal
reference standard because they overcome random and
systematic errors by sampling the gland every 5 mm,
provide three-dimensional coordinates for correlation to
imaging,andcanbeappliedtoallmen.Amulticentretrialin
Table 6 – Areas lacking consensus or with negative consensus on disease detection and characterisation by individual sequences
It is possible to be gain the following information from each MR sequence in isolation:MR sequence
T1W T2WDW CEMRSI
Detection of any cancer in the peripheral zones
Detection of any cancer or exclusion of clinically significant disease as defined by both lesion
sizes (?0.2 cm3and ?0.5 cm3) in the transition zone
The Gleason grade of lesions or the exclusion of clinically significant disease as defined by a lesion
size ?0.2 cm3(approximately 7 mm) in the peripheral zone
The Gleason grade of lesions in the transition zone
Exclusion of clinically significant disease as defined by a lesion size ?0.5 cm3(approximately 10 mm)
in the peripheral zone
Differentiation between low-grade and intermediate/high-grade tumours (defined as a tumour
with a Gleason 4 grading component) in the peripheral or transition zones
Exclusion of clinically significant cancer according to the definition of a lesion ?0.2 cm3and/or
Gleason ?3 + 4 in the peripheral and transition zones
Exclusion of clinically significant cancer according to the definition of a lesion ?0.5 cm3and/or
Gleason ?4 + 3 in the peripheral zone
Exclusion of clinically significant cancer according to the definition of a lesion ?0.5 cm3and/or
Gleason ?4 + 3 in the transition zone
CE = contrast enhanced; DW = diffusion weighted; MR = magnetic resonance; MRSI = magnetic resonance spectroscopy; T1W = T1 weighted; T2W = T2
weighted;? = areas lacking consensus; X = areas considered inappropriate (negative consensus); H = areas with consensus.
X
X
?
?
H
?
H
?
H
?
X?
H
??
X
X
X
?
?
H
?
H
?
?
XX???
X????
X?
H
??
X????
EUROPEAN UROLOGY 59 (2011) 477–494
482

Page 7

Table 7 – Full table of results
InappropriateUncertain Appropriate
1. Overview
Minimum and optimum MR data sets for the detection and localisation of prostate cancer should be establishedX
A scoring system is necessary for the interpretation of diagnostic MRI of the prostateX
A standard method dividing the prostate into sectors needs to be agreed onX
A standard method for reporting the results needs to be agreed onX
2. General MRI components
Minimum imaging requirements
The data set should include T1-weighted, T2-weighted, diffusion-weighted, and contrast-enhanced MRIX
The dataset should include MR spectroscopyX
Imaging could be adequately performed at 0.5 TX
Imaging could be adequately performed at 1.5 TX
Imaging should be performed at 3 TX*
A pelvic phased-array coil is requiredX
An endorectal coil and bowel preparation are requiredX
A power injector is requiredX*
Bowel relaxant is requiredX*
Optimum imaging requirements
The data set should include T1-weighted, T2-weighted, diffusion-weighted, and contrast-enhanced MR imagesX
The data set should include MR spectroscopy X*
Imaging should be performed at 0.5 TX
Imaging should be performed at 1.5 TX*
Imaging should be performed at 3 TX
A pelvic phased-array coil, power injector, and bowel relaxant are requiredX
An endorectal coil is requiredX*
Bowel preparation is requiredX
All imaging
If bowel relaxant is used, both Buscopan and glucagon are suitable agentsX
An endorectal coil must be used for 1.5-T and 3-T imagingX
E UR O P E A N U R OL O GY 59 ( 2 01 1) 4 7 7 – 4 9 4
483

Page 8

Table 7 (Continued)
InappropriateUncertain Appropriate
3. MRI sequence parameters
a. T2-weighted MRI
Minimum requirements
Maximum slice thickness for diagnostic imaging at 1.5 T should be 4 mmX
Maximum slice thickness for diagnostic imaging at 3 T should be 3 mmX
The planes of imaging should include the axial plane and one other planeX
The in-plane image resolution should be 0.5 ? 0.5 mm to 0.7 ? 0.7 mm at both 1.5 T and 3 T
Quantification of T2 relaxation time should be performed as standard (minimal) practice
X
X
Optimum requirements
Maximum slice thickness for diagnostic imaging at 1.5 T should be 3 mmX
Maximum slice thickness for diagnostic imaging at 3 T should be 3 mmX
The planes of imaging should include the axial, sagittal, and coronal planesX
The in-plane image resolution should be 0.5 ? 0.5 mm to 0.7 ? 0.7 mm at 1.5 T
The in-plane image resolution should be 0.3 ? 0.3 mm to 0.5 ? 0.5 mm at 3 T
Quantification of T2 relaxation time should be performed as optimal practice
X
X
X*
b. Diffusion-weighted imaging
Minimum requirements
Maximum slice thickness should be 5 mm for diagnostic imaging at both 1.5 T and 3 TX
The planes of imaging should include the axial plane onlyX
The in-plane image resolution should be 1.5 ? 1.5 mm to 2 ? 2 mm at both 1.5 T and 3 T
ADC quantification should be performed as standard (minimal) practice
X
X*
An ADC map should be used for diagnostic interpretation as standard (minimal) practiceX
Minimum requirements for diffusion-weighted imaging:
- A single ‘‘high b-value’’ diffusion-weighted imageX*
- A b0 image and a single diffusion-weighted image allowing quantification of ADC via monoexponential fittingX*
- A b0 image and multiple (?2) b values allowing quantification of ADC via monoexponential fitting
- Multiple b values allowing biexponential ADC derivation
X*
X
For calculation of ADC, the highest b-value that should be used is 800 (with adequate signal-to-noise ratio)X
EUROPEAN UROLOGY 59 (2011) 477–494
484

Page 9

If a high single b-value image is used for lesion detection, the minimal required is any of 800, 1000, or 1400X*
If a high single b-value image is used for lesion detection, the minimal required is either 2000 or 3000X
Diffusion tensor imaging should be part of the minimum requirement for prostate diffusion imagingX
Optimum requirements
Maximum slice thickness for diagnostic imaging should be 5 mm for 1.5-T imaging and 4 mm for 3-T imagingX
The planes of imaging should include the axial plane onlyX
The in-plane image resolution should be 0.3 ? 0.3 mm to 0.5 ? 0.5 mm or 0.5 ? 0.5 mm to 0.7 ? 0.7 mm or 0.7 mm ? 0.7 mm to 1 ? 1 mm at 1.5 T
The in-plane image resolution should be 1 ? 1 mm to 1.5 ? 1.5 mm or 1.5 ? 1.5 mm to 2 ? 2 mm at 1.5 T
The in-plane image resolution should be 1 ? 1 mm to 1.5 ? 1.5 mm at 3 T
ADC quantification should be performed as part of optimal practice
X
X*
X
X
An ADC map should be used for diagnostic interpretation as part of optimal practiceX
Optimum requirements for diffusion-weighted imaging:
- A single ‘‘high b-value’’ diffusion-weighted imageX
- A b0 image and a single diffusion-weighted image allowing quantification of ADC via monoexponential fitting X*
- A b0 image and multiple (?2) b values allowing quantification of ADC via monoexponential fitting
Multiple b-values allowing biexponential ADC derivation
X
X*
For calculation of ADC, the highest b-value that should be used is 1000 (with adequate signal-to-noise ratio)X
If a high single b-value is used for lesion detection, this should be at a value of 1400 for optimal imagingX
Diffusion tensor imaging should be part of the optimum requirement for prostate diffusion imagingX
All diffusion-weighted imaging
A standard set of b-values should be used at all institutions for prostate imagingX
ADC values should be quoted with the b-values used for calculationX
Perfusion-insensitive ADC values should be quoted separately from the total ADC valueX
Diffusion tensor imaging
- Aids in the characterisation of nodules in the transition zone when combined with trace diffusion imagesX
- Aids in the characterisation of nodules in the peripheral zone when combined with trace diffusion imagesX*
c. Contrast-enhanced imaging
Minimum requirements
Maximum slice thickness should be 4 mm for both 1.5-T and 3-T imagingX
Imaging should be performed in the axial planeX
E UR O P E A N U R OL O GY 59 ( 2 01 1) 4 7 7 – 4 9 4
485

Page 10

Table 7 (Continued)
InappropriateUncertain Appropriate
Quantitative or semiquantitative dynamic contrast-enhanced MRI should be performed as standard (minimal) practiceX
The minimum contrast-enhanced MRI data set should be sufficient for:
- Detecting early arterial enhancement and early washout only (ie, baseline, early arterial image, and venous phase image)X*
- Plotting a signal intensity-time curve allowing extraction of curve parameters (slope of enhancement, maximum enhancement,
time to enhancement, and curve shape)
X*
- Physiologic modelling via contrast agent quantificationX
For signal intensity time curves, the in-plane image resolution should be 0.7 ? 0.7 mm to 1 ? 1 mm at 1.5 T and 0.5 ? 0.5 mm to 0.7 ? 0.7 mm at 3 T
For analysis of signal intensity time curve parameters, the maximum temporal resolution should be 10–15 s
X
X
For dynamic contrast-enhanced imaging, a power injector should be used as part of minimal practiceX
The rate of injection should be 3 ml/sX
Optimum requirements
Maximum slice thickness for diagnostic imaging should be 3 mm for both 1.5- and 3-T imaging.X
Contrast enhanced imaging should be performed in the axial plane X*
Contrast enhanced imaging should be performed in the coronal or sagittal planesX
Contrast enhanced imaging should be performed as an isometric acquisitionX
Quantitative or semiquantitative dynamic contrast enhanced MRI should be performed as optimal practiceX
The optimum contrast-enhanced MRI data set should be sufficient for:
- Detecting early arterial enhancement and early washout only (ie, baseline, early arterial image, and venous phase image)X*
- Plotting a signal intensity time curve allowing extraction of curve parameters (slope of enhancement, maximum enhancement,
time to enhancement, and curve shape)
X*
- Physiologic modelling via contrast agent quantificationX*
For dynamic contrast-enhanced imaging, a power injector should be used as part of optimal practiceX
The rate of injection should be 3 ml/sX
All contrast-enhanced imaging
A single dose of contrast agent should be usedX
For the three time-point technique, apart from the baseline and 90-s time points, a third time point should be specifiedX*
For DCE imaging. acquisition should be continued for 5 min to detect washoutX
d. MR spectroscopy
Single-voxel spectroscopy should be performed in areas of concern following review of the imaging by a radiologist as part of standard (minimal) practiceX
EUROPEAN UROLOGY 59 (2011) 477–494
486

Page 11

Chemical shift imaging through the whole prostate should be performed as part of standard (minimal) practiceX
The maximum voxel size should be 0.5 cm3
X
For voxels that are considered usable on signal-to-noise ratio grounds, as a minimum requirement for benign versus malignant assignment, two
adjacent voxels only can be considered
X
MR spectroscopy analysis should be qualitative using visual classification of the patterns observedX
Optimum requirements
Single-voxel spectroscopy should be performed in areas of concern following review of the imaging by a radiologist as part of optimal practiceX
Chemical shift imaging through the whole prostate should be performed as part of optimal practiceX*
The maximum voxel size should be 0.5 cm3
X
For voxels that are considered usable on signal-to-noise ratio grounds, as a minimum requirement for benign versus malignant assignment, three
or more adjacent voxels only can be considered
X
MR spectroscopy analysis should be qualitative using visual classification of the patterns observed or quantitative using ratios of heights
of spectroscopy curves
X*
MR spectroscopy analysis should involve quantitative and qualitative ratios togetherX*
MR spectroscopy analysis should involve full quantification with metabolite concentrationsX
NB: Insufficient numbers of panellists were able to answer further questions about MR spectroscopy and conduct; therefore a report of these statements is not included
4. Image interpretation
a. Detection of any cancer
It is possible to be highly suspicious for the presence of any cancer in the peripheral zone only on the following sequences in isolation:
- T1 weightedX
- T2 weightedX
- Diffusion weightedX
- Contrast enhancedX
- MR spectroscopyX*
It is possible to be highly suspicious for the presence of any cancer in the transition zone only on the following sequences in isolation:
- T1 weightedX
- T2 weightedX*
- Diffusion weighted X*
- Contrast enhanced X*
- MR spectroscopy X*
E UR O P E A N U R OL O GY 59 ( 2 01 1) 4 7 7 – 4 9 4
487

Page 12

Table 7 (Continued)
InappropriateUncertain Appropriate
b. Information about Gleason grade alone
The following sequences in isolation provide information about the Gleason grade of lesions in the peripheral zone:
- T1 weightedX
- T2 weighted X*
- Diffusion weightedX
- Contrast enhancedX*
- MR spectroscopyX*
The following sequences in isolation provide information about the Gleason grade of lesions in the transition zone:
- T1 weightedX
- T2 weightedX
- Diffusion weighted X*
- Contrast enhancedX*
- MR spectroscopyX*
c. Information about lesion size alone
The following sequences in isolation are useful to exclude ‘‘clinically significant’’ disease as defined by a lesion size of ?0.2 cm3(approximately 7 mm) in the peripheral zone:
- T1 weightedX
- T2 weighted X*
- Diffusion weightedX
- Contrast enhancedX*
- MR spectroscopyX*
The following sequences are useful in isolation to exclude ‘‘clinically significant’’ disease as defined by a lesion size of ?0.2 cm3(approximately 7 mm) in the transition zone:
- T1 weightedX
- T2 weightedX*
- Diffusion weightedX*
- Contrast enhancedX*
- MR spectroscopy X*
The following sequences are useful to exclude ‘‘clinically significant’’ disease as defined by a lesion size of ?0.5 cm3(approximately 10 mm) in the peripheral zone in isolation:
EUROPEAN UROLOGY 59 (2011) 477–494
488

Page 13

- T1 weightedX
- T2 weighted X*
- Diffusion weightedX
- Contrast enhancedX
- MR spectroscopy X*
The following sequences are useful to exclude ‘‘clinically significant’’ disease as defined by a lesion size of ?0.5 cm3(approximately 10 mm) in the transition zone in isolation:
- T1 weightedX
- T2 weightedX*
- Diffusion weighted X*
- Contrast enhanced X*
- MR spectroscopyX*
d. Differentiation between low-and intermediate/high-grade tumours
The following sequences can be used to differentiate between low-grade and intermediate/high-grade tumours (defined as a tumour with a Gleason 4 grading component) in the peripheral zone in isolation:
- T1 weightedX
- T2 weightedX
- Diffusion weighted X*
- Contrast enhancedX*
- MR spectroscopy X*
The following sequences can be used to differentiate between low-grade and intermediate/high-grade tumours (defined as a tumour with a Gleason 4 grading component) in the transition zone in isolation:
- T1 weightedX
- T2 weightedX
- Diffusion weightedX*
- Contrast enhanced X*
- MR spectroscopy X*
e. Exclusion of ‘‘clinically significant’’ disease
It is possible to exclude ‘‘clinically significant’’ cancer on the following sequences in isolation according to definition 1: lesion ?0.2 cm3and/or Gleason ?3 + 4 in the peripheral zone:
- T1 weightedX
- T2 weighted X*
- Diffusion weightedX*
- Contrast enhanced X*
- MR spectroscopyX*
E UR O P E A N U R OL O GY 59 ( 2 01 1) 4 7 7 – 4 9 4
489

Page 14

Table 7 (Continued)
InappropriateUncertainAppropriate
It is possible to exclude ‘‘clinically significant’’ cancer on the following sequences in isolation according to definition 1: lesion ?0.2 cm3and/or Gleason ? 3 + 4 in the transition zone:
- T1 weightedX
- T2 weightedX*
- Diffusion weighted X*
- Contrast enhancedX*
- MR spectroscopyX*
It is possible to exclude ‘‘clinically significant’ cancer on the following sequences in isolation according to definition 1: lesion ?0.5 cm3and/or Gleason ? 4 + 3 in the peripheral zone:
- T1 weightedX
- T2 weighted X*
- Diffusion weightedX
- Contrast enhanced X*
- MR spectroscopyX*
It is possible to exclude ‘‘clinically significant’’ cancer on the following sequences in isolation according to definition 1: lesion ?0.5 cm3and/or Gleason ? 4 + 3 in the transition zone:
- T1 weightedX
- T2 weightedX*
- Diffusion weighted X*
- Contrast enhancedX*
- MR spectroscopyX*
f. Scoring system
When scoring the prostate for the presence or absence of cancer for T1-weighted imaging, the range of scores should be 1–4, 1–6, 1–10, or 1–20X
When scoring the prostate for the presence or absence of cancer for T1-weighted imaging, the range of scores should be 1–3 or 1–5X*
When scoring the prostate for the presence or absence of cancer for T2-weighted, diffusion-weighted, contrast enhanced, and MR spectroscopy sequences,
the range of scores should be 1–5
X
Each individual lesion should be separately scored for probability of malignancyX
Only areas of the prostate should be scored rather than by individual lesionsX*
Both individual lesions and areas of the prostate should be given a scoreX
The maximum diameter of the largest abnormal lesion should be recordedX
The following should be scored for involvement:
- Extracapsular extensionX
- Seminal vesicles (extra- and intraprostatic)X
EUROPEAN UROLOGY 59 (2011) 477–494
490

Page 15

- Distal sphincterX
- Rectal wallX
- Neurovascular bundlesX
- Bladder neckX
All extracapsular involvement features should have an individual scoring range of 1–5X
g. Regions of interest
As a minimum requirement, the prostate should be divided into 16 regions of interest (apical, mid, and base quadrants)X
As an optimum requirement, the prostate should be divided into 27 regions of interestX
h. Sequence reporting methodology
The ADC value should be stated for any suspicious region detectedX
Dynamic contrast-enhanced MRI should be scored according to the morphological enhancement patternX
Physiologic modelling is a method that can be appropriately used in the present timeX
Full quantification is a method of scoring that should be reconsidered in the next 2–5 yrX*
i. Other recommendations
The following clinical information is important for reporting the imaging and should be included:
- PSA levelX
- Digital rectal examinationX
- Other markers (eg, PCA3 level)X*
- Times scale since prostate biopsies and results of previous biopsiesX
- Results of previous MRI scansX
- History of previous prostate treatment or intervention (eg, TURP, prostate radiotherapy)X
- History of medical treatment (eg, 5a-reductase inhibitors, hormones)X
The overall score should be based purely on imaging appearances X*
The overall score for probability of tumour given by the radiologist should be influenced by other clinical results (eg, PSA level)X*
The radiologist should be blinded to the patient’s clinical detailsX
As a minimum requirement, each MRI should be assessed and scored by one radiologistX
As an optimal requirement, each MRI should be assessed and scored by two radiologists independently and discrepancies referred for consensusX
When used for the detection or localisation of disease, 10 or 20 prostate MRI studies should be the minimum annual reporting requirement for
reporting radiologists
X
When used for the detection or localisation of disease, 50, 80, or 100 prostate MRI studies should be the minimum annual reporting requirement for
reporting radiologists
X*
A separate radiologist’s ‘‘hunch’’ score should be given that represents the radiologist’s personal overall hunch view on the
likelihood of malignancy regardless of the objective radiologic score
X*
E UR O P E A N U R OL O GY 59 ( 2 01 1) 4 7 7 – 4 9 4
491