UROGENITAL

MR-sequences for prostate cancer diagnostics: validation basedon the PI-RADS scoring system and targeted MR-guided in-borebiopsy

Lars Schimmöller & Michael Quentin & Christian Arsov & Andreas Hiester &

Christian Buchbender & Robert Rabenalt & Peter Albers & Gerald Antoch & Dirk Blondin

Received: 5 March 2014 /Revised: 24 May 2014 /Accepted: 5 June 2014# European Society of Radiology 2014

AbstractPurpose This study evaluated the accuracy of MR sequences[T2-, diffusion-weighted, and dynamic contrast-enhanced(T2WI, DWI, and DCE) imaging] at 3T, based on theEuropean Society of Urogenital Radiology (ESUR) scoringsystem [Prostate Imaging Reporting and Data System (PI-RADS)] using MR-guided in-bore prostate biopsies as refer-ence standard.Methods In 235 consecutive patients [aged 65.7±7.9 years;median prostate-specific antigen (PSA) 8 ng/ml] with

multiparametric prostate MRI (mp-MRI), 566 lesions werescored according to PI-RADS. Histology of all lesions wasobtained by targeted MR-guided in-bore biopsy.Results In 200 lesions, biopsy revealed prostate cancer (PCa).The area under the curve (AUC) for cancer detection was 0.70(T2WI), 0.80 (DWI), and 0.74 (DCE). A combination ofT2WI + DWI, T2WI + DCE, and DWI + DCE achieved anAUC of 0.81, 0.78, and 0.79. A summed PI-RADS score ofT2WI + DWI + DCE achieved an AUC of 0.81. For highergrade PCa (primary Gleason pattern≥4), the AUC was 0.85for T2WI + DWI, 0.84 for T2WI + DCE, 0.86 for DWI +DCE, and 0.87 for T2WI + DWI + DCE. The AUC for T2WI+ DWI + DCE for transitional-zone PCa was 0.73, and for theperipheral zone 0.88. Regarding higher-grade PCa, AUC fortransitional-zone PCa was 0.88, and for peripheral zone 0.96.Conclusion The combination of T2WI + DWI + DCEachieved the highest test accuracy, especially in patients withhigher-grade PCa. The use of ≤2 MR sequences led to lowerAUC in higher-grade and peripheral-zone cancers.Key Points• T2WI + DWI + DCE achieved the highest accuracy inpatients with higher grade PCa

• T2WI + DWI + DCE was more accurate for peripheral-than for transitional-zone PCa

• DCE increased PCa detection accuracy in the peripheralzone

• DWI was the leading sequence in the transitional zone

Keywords Prostate cancer .MRI . PI-RADS .MRsequences .MR-guided in-bore biopsy

AbbreviationsESUR European Society of Urogenital RadiologyPI-RADS Prostate Imaging Reporting and Data Systemmp-MRI Multiparametric magnetic resonance imaging

L. Schimmöller :M. Quentin (*) : C. Buchbender :G. Antoch :D. BlondinDepartment of Diagnostic and Interventional Radiology, UniversityDusseldorf, Medical Faculty, Moorenstr. 5, 40225 Dusseldorf,Germanye-mail: michael.quentin@med.uni-duesseldorf.de

L. Schimmöllere-mail: Lars.Schimmoeller@med.uni-duesseldorf.de

C. Buchbendere-mail: Christian.Buchbender@med.uni-duesseldorf.de

G. Antoche-mail: Antoch@med.uni-duesseldorf.de

D. Blondine-mail: Dirk.Blondin@sk-mg.de

C. Arsov :A. Hiester : R. Rabenalt : P. AlbersDepartment of Urology, University Dusseldorf, Medical Faculty,Moorenstr. 5, 40225 Dusseldorf, Germany

C. Arsove-mail: Cristian.arsov@med.uni-duesseldorf.de

A. Hiestere-mail: Andreas.Hiester@med.uni-duesseldorf.de

R. Rabenalte-mail: Robert.Rabenalt@med.uni-duesseldorf.de

P. Alberse-mail: urologie@uni-duesseldorf.de

Eur RadiolDOI 10.1007/s00330-014-3276-9

DCE Dynamic contrast-enhanced imagingMRSI Magnetic resonance spectroscopic imagingDWI Diffusion-weighted imagingROC Receiver operating characteristicAUC Area under the curvePCa Prostate cancerPSA Prostate-specific antigenTRUS Transrectal ultrasound

Introduction

Multiparametric magnetic resonance imaging (mp-MRI) ofthe prostate is gaining increasing relevance in the competitivefield of prostate cancer (PCa) diagnostics [1]. To overcome thelack of standardizedMRI protocols and diagnostic criteria, theEuropean Society of Urogenital Radiology (ESUR) publisheda unified scoring system [Prostate Imaging Reporting andData System (PI-RADS)] in 2012 [2]. Since then this scoringsystem has been evaluated and analysed by various studygroups [3, 4] and proved to have a good to moderateinterreader agreement [5]. The interreader reproducibility im-proved with increasing readers’ experience, but overall seemsto be lower for lesions in the transitional zone [6]. Althoughthe PI-RADS system offers major advances for interpretingmultiparametric prostate MRI, it should still be considered asa work in progress and requires further refinements [7].

Functional imaging techniques, such as diffusion-weighted(DWI), dynamic contrast-enhanced (DCE), and magnetic reso-nance spectroscopic MRS) imaging clearly improve diagnosticaccuracy compared with anatomical sequences and morpholo-gy alone has been repeatedly shown [8]. However, it is stillcontroversial as to whether every single functional MR (fMR)sequence is required for correct diagnosis [9, 10]. Additionally,the significance of the different MR sequences in relation tolocalisation of a suspect lesion is still under discussion [11, 12].

The aim of this study was to evaluate MR sequencesregarding their test accuracy for PCa detection based on thePI-RADS system using targeted MR-guided in-bore biopsy asthe standard of reference. Furthermore, differences relating tothe different MR sequences when assessing peripheral- andtransitional-zone PCa were analysed.

Materials and methods

Patients

The study was approved by the local ethics committee.Written informed consent was obtained from all patients.Between November 2011 and October 2013, 235 consecutivepatients [mean age 65.7±7.9 years, mean prostate volume

57.9±30.3 ml, mean prostate-specific antigen (PSA) value9.9±8.5 ng/ml; median PSA 8.0 ng/ml, lower/upper quartile5.7/11.0 ng/ml] with increased PSA levels (>4 ng/ml) and noknown PCa, who underwent diagnostic mp-MRI and at least4–6 weeks later a targeted MR-guided in-bore biopsy, wereretrospectively included in this study. Patients with (n=87)and without (n=148) previous negative transrectal ultrasound(TRUS) biopsy were included.

MRI protocol

Prostate mp-MRI was performed on a 3-Tesla MR scanner(Magnetom Trio™, Siemens AG, Healthcare Sector,Forchheim, Germany) with a six-channel phased-array bodycoil according to the recommendations of a EuropeanConsensus Meeting on the Standardization of Prostate MRI[13]. The MRI protocol, published previously [14], containedaxial, sagittal and coronal T2-weighted turbo spin-echo se-quences (axial: TR 10,630 ms, TE 117 ms, FOV 12.8 cm,voxel size 0.5×0.5×3.0 mm) and axial T1-weighted turbospin-echo images (TR 650 ms, TE 13 ms, FOV 30 cm, voxelsize 1.3×0.9×5.0 mm) for anatomic imaging. With DWIusing five b-values (0, 250, 500, 750, 1000 s/mm2) with fiveaverages and DCE imaging, two functional imaging tech-niques were performed: DWI was acquired with a single-shot spin-echo echo-planar sequence (TR 4,600 ms, TE90 ms, FOV 20.4 cm, voxel size 1.5×1.5×3.0 mm), andvolume-interpolated gradient echo (GRE) sequences (TR5.26 ms, TE 1.76 ms, FOV 19.2 cm, voxel size 1.5×1.5×3.0 mm, imaging time 5:05 min, temporal resolution 10 s)were applied for DCE. Apparent diffusion coefficient (ADC)parameter maps were calculated using the standardmonoexponential model. All patients received preliminarybutylscopolamine (20 mg Buscopan®, Boehringer IngelheimPharma, Ingelheim, Germany) to suppress bowel peristalsis.Gadoteric acid (Dotarem®, Guerbet, Aulnay-sous-Bois,France) was used for contrast-media injection in a weight-adapted standard dose (0.2 mmol/kg body weight) and aninjection rate of 3 ml/s.

MR image analysis

The most suspicious prostate lesions, up to three perpatient, were rated by an experienced uroradiologist(DB, MQ and LS with 6, 5 and 4 years of experiencein reading prostate MRI). For postprocessing of DCEimages the DynaCAD software (Invivo, Orlando, FL,USA) was used on an external workstation. All othersequences were evaluated on a standard Picture Archiveand Communication System (PACS) workstation.According to the ESUR guidelines, T2WI, DWI andDCE images were scored (1–5) using the PI-RADSsystem [2]. Additionally, a summed score of all three

Eur Radiol

MR sequences (T2WI + DWI + DCE) was calculatedfor all lesions [15]. For lesion localisation, a 27-regionscheme was applied [16]. For analysis purposes, alllesions were subsequently grouped as either peripheral-or transitional-zone lesions.

MR-guided in-bore biopsy

MR-guided biopsies were performed on a 3T MRI (MagnetomTrio™, Siemens AG, Healthcare Sector, Forchheim, Germany)by the same experienced uroradiologists. Patients were placedin a prone position. A needle guide marked with 2 % lidocainegel (Instillagel®, Farco-Pharma GmbH, Köln, Germany) andconnected to a portable biopsy device (DynaTRIM) was rectal-ly inserted (Invivo). T2-weighted half-Fourier-acquisition sin-gle shot turbo spin echo (HASTE) axial (TR 2,000 ms, TE76 ms, FOV 28 cm, voxel size 1.4×1.1×3.0 mm) and sagittal(TR 2,000 ms, TE 76 ms, FOV 28 cm, voxel size 1.4×1.1×3.0 mm) images were acquired with a six-channel phased-arraybody coil. Image data were transferred to a DynaCAD-workstation (Invivo) for biopsy planning. Two cores were takenof each lesion with an MR-compatible, 18-gauge, fully auto-matic biopsy gun [needle length 150 or 175 mm (Invivo)]. Intotal, no more than six cores were taken.

Histopathology

All biopsy cores were immediately stained with haematoxylinand eosin (H&E), fixed in formalin and underwent routinehistopathological evaluation. Higher-grade PCa was definedas primary Gleason pattern≥4 [17].

Statistics

Statistical analysis was performed using IBM® SPSS®Statistics 19 for Windows (SPSS Inc., Chicago, IL, USA).Statistical significance was defined as a p value<0.05. Alldata are expressed as mean±standard deviation (SD) or asmedian with interquartile range (IQR), if appropriate. All datawere tested for normal distribution with Kolmogorov–Smirnov test. Normally distributed parameters were comparedusing the independent sample t test, and nonparametric datawere tested with the Mann–Whitney U test (M-W-U).Receiver operating characteristics (ROC) analysis was per-formed corewise for each MR sequence separately and fordifferent combinations of MR sequences. ROC analysis wasadditionally performed separately for PCa in the peripheraland transitional zones.

Table 1 Baseline characteristics [age, prostate-specific antigen (PSA) value, prostate volume) of all patients with positive or negative biopsy results.Patients with verified malignant lesions were separated by all and higher-grade prostate cancer (PCa) (primary Gleason pattern≥4)

Baseline characteristics No. Age (years; mean±SD PSA value (ng/ml) Prostate volume (ml)

All 235 65.7±7.9 9.9±8.5 8.0 5.7 11.0 57.9±30.3

Negative biopsy 120 64.2±8.3 9.2±6.6 8.0 5.6 10.9 70.6±32.1

Positive biopsy 115 67.2±7.4 10.6±10.1 8.2 5.8 11.3 44.7±21.0

GS≥4+3=7 28 69.0±6.2 14.5±16.9 9.2 7.4 16.8 50.9±22.6

SD standard deviation, PSA prostate-specific antigen, GS Gleason score

Table 2 Prostate Imaging Reporting and Data System (PI-RADS) score (mean±standard deviation) of patients with positive and negative biopsyresults. Data are separated by lesion localisation (peripheral or transitional) and by higher-grade prostate cancer (PCa) (primary Gleason pattern≥4)

PI-RADS scoring Lesions (n) T2WI (mean±SD) DWI (mean±SD) DCE (mean±SD) Summed score (mean±SD)

All 566 3.8±0.8 4.0±0.8 2.7±1.4 10.6±2.2

Benign 366 3.6±0.8 3.8±0.8 2.3±1.1 9.6±1.6

Peripheral 102 3.7±0.7 3.7±0.7 2.0±1.1 9.4±1.8

Transitional 297 3.5±0.8 3.8±0.7 2.4±1.1 9.7±1.6

Malignant 200 4.2±0.6 4.6±0.6 3.5±1.4 12.2±2.2

GS≥4+3=7 69 4.3±0.7 4.8±0.4 4.1±1.1 13.2±1.6

Peripheral PCa 117 4.3±0.6 4.6±0.6 3.9±1.3 12.8±2.0

GS≥4+3=7 49 4.5±0.5 4.9±0.4 4.3±1.1 13.6±1.4

Transitional PCa 83 4.0±0.6 4.5±0.6 3.0±1.5 11.4±2.1

GS≥4+3=7 20 4.0±0.8 4.7±0.5 3.7±1.1 12.3±1.6

PI-RADS Prostate Imaging Reporting and Data System, GS Gleason Score, PCa prostate cancer, T2WI T2-weighted imaging, DWI diffusion-weightedimaging, DCE dynamic contrast-enhanced imaging, SD standard deviation

Eur Radiol

Results

Patients

In 115 patients (49 %) MR-guided in-bore biopsy confirmedPCa in 200 different lesions; 366 lesions were negative forPCa (benign lesions). Biopsy revealed higher-grade PCa in 69lesions (35 %) and lower grade PCa (≤3+4=7) in 131 lesions(65 %). Gleason Score distribution of all lesions was asfollows: 3+3=6 (47 lesions), 3+4=7 (84 lesions), 4+3=7(33 lesions), 4+4=8 (22 lesions), 4+5=9 (9 lesions) and 5+4=9 (5 lesions). Detailed baseline characteristics of all pa-tients separated for negative and positive biopsy and for aGleason Score≥3+4=7 are shown in Table 1.

MR image analysis

Mean PI-RADS score of all lesion (n=566) was 3.8±0.8 forT2WI, 4.0±0.8 for DWI and 2.7±1.4 for DCE. Tumour lesionshad a mean score of 4.2±0.6 (T2WI), 4.6±0.6 (DWI), and 3.5±1.4 (DCE). Benign lesions had amean score of 3.6±0.8 (T2WI),3.8±0.8 (DWI), and 2.3±1.1 (DCE) (for significance levels, seebelow). Mean overall PI-RADS score of benign and tumourlesions was 9.6±1.6, and 12.2±2.2, respectively. Detailed PI-RADS evaluation is shown in Table 2 separated for peripheral-(Fig. 1) and transitional-zone (Fig. 2) PCa.

Variance analysis

Variance analysis of patient data (age, PSA value,prostate volume) showed significant lower prostatevolume (p<0.01) and higher age (p<0.01) for allpatients with PCa compared with patients with benignlesions; there was no significant difference in PSAvalues (p=0.47). A significantly higher age (p<0.01)and PSA value (p=0.02) was found in patients withhigher-grade PCa compared with all patients withPCa. Analysis of the PI-RADS score expressed sig-nificant differences for each MR modality (T2WI,DWI, DCE) as well for all malignant, peripheral andtransitional PCa lesions compared with benign lesions(p<0.01). The summed PI-RADS score (T2WI + DWI+ DCE) likewise was significantly different for allmalignant, peripheral and transitional PCa lesionscompared with benign lesions and between all malig-nant lesions and higher-grade cancer lesions (p<0.01).Between peripheral and transitional PCa lesions, therewas a significant difference for all MR sequences andlikewise for the summed score (p<0.01). For aGleason Score≥4+3=7, peripheral and transitionalPCa lesions differed significantly only in T2WI(p<0.01), DCE (p=0.02) and the summed PI-RADSscore (p<0.01), whereas for DWI, there was no sig-nificant difference (p=0.13).

Fig. 1 Case 1: peripheral prostate cancer. a Axial T2-weighted image(T2WI); b, c apparent diffusion coefficient (ADC) map and diffusion-weighted imaging (DWI) on high b-value (1,000 s/mm2); d, e dynamic

contrast-enhanced (DCE) (washout) with a type 3 curve [Prostate Imag-ing Reporting and Data System (PI-RADS) score 5+5+5=15; GleasonScore 4+3=7]

Fig. 2 Case 2 transitional prostate cancer (PCs):. a Axial T2-weightedimage (T2WI); b, c apparent diffusion coefficient (ADC) map and diffu-sion-weighted imaging (DWI) on high b-value (1,000 s/mm2); d, e

dynamic contrast-enhanced (DCE) (washout) with type 2 curve [ProstateImaging Reporting and Data System (PI-RADS) score 4+4+4=12;Gleason Score 4+3=7)

Eur Radiol

Receiver operating curve analysis

Evaluation of the different MR sequences based on thePI-RADS score showed that the use of a single MRsequence led to a lower AUC for PCa detection. Amongthe tested sequences, DWI achieved the highest testaccuracy, followed by DCE and T2WI (Table 3). Thecombination of PI-RADS scores of two MR sequences(T2WI + DWI, T2WI + DCE or DWI + DCE) im-proved the accuracy. A summed PI-RADS score of allthree MR sequences showed the highest test accuracyfor general PCa detection. For higher-grade PCa, testaccuracy was considerably higher (Fig. 3). This wasalso the case for both peripheral- and transitional-zonePCa (Fig. 4). However, for PCa detection in the transi-tional zone, DCE showed the lowest AUC (0.63),whereas the combination of T2WI and DWI achievedthe highest test accuracy, with an AUC of 0.76. Thehighest AUC of all was achieved with T2WI + DWI +DCE for peripheral higher-grade PCa lesions (0.96).Detailed receiver operating characteristic (ROC) evalua-tions (AUC values) of MR sequences separately forperipheral or transitional PCa lesions and for all andhigher-grade PCa lesions are shown in Table 3.

Discussion

Selection and weighting of different MR sequences formp-MRI of the prostate is still competitively discussed[9, 18]. Our study investigated MR sequences recom-mended by ESUR regarding their individual and com-bined test accuracy for PCa detection. Our resultsshowed that a summed PI-RADS score of three MRsequences (T2WI + DWI + DCE) achieved the highesttest accuracy, apart from lower-grade PCa in the transi-tional zone. In patients with higher-grade (primaryGleason pattern≥4) and peripheral PCa in particular,the scoring system led to the highest AUC values.

DWI seems to be the most important MR sequencefor tumour detection [9, 19]. In view of single MRsequence analysis, we also verified the highest AUCvalues for overall PCa detection using DWI, as wellas for PCa localisation in the peripheral or transitionalzone. Nonetheless, additional T2WI and DCE sequencesimprove the detection rate of malignant lesions, espe-cially of higher-grade PCa. In the subgroup of patientslower-grade PCa, the summed PI-RADS score of T2WIand DWI only resulted in similar good test accuracy. Inagreement with mp-MRI after radiation therapy, therewere no additional benefits if DCE was added toT2WI and DWI [20]. However, for higher-grade PCadetection, DCE further improved data accuracy. AT

able3

Receiveroperatingcharacteristics(ROC)analysisof

theProstateIm

agingReportin

gandDataSystem(PI-RADS)

scoreforallm

alignant,transitionalandperipherallesions,andadditio

nally

for

higher-grade

prostatecancer

(PCa)

lesions(primaryGleason

pattern≥4

)separatedby

each

singlemagnetic

resonance(M

R)sequence

andtheadditio

nof

thescoreof

twoor

allthree

MRsequences

ROCanalysis(95%

CI)

Malignant

Transitional

Peripheral

All

GS≥4

+3=7

All

GS≥4

+3=7

All

GS≥4

+3=7

T2W

I0.698(0.653–0.743)

0.722(0.658–0.785)

0.642(0.577–0.707)

0.647(0.524–0.770)

0.738(0.686–0.790)

0.809(0.744–0.873)

DWI

0.797(0.757–0.837)

0.834(0.787–0.882)

0.756(0.696–0.816)

0.854(0.763–0.945)

0.828(0.780–0.876)

0.912(0.867–0.958)

DCE

0.735(0.689–0.781)

0.807(0.753–0.861)

0.629(0.556–0.703)

0.792(0.669–0.885)

0.809(0.758–0.860)

0.879(0.819–0.939)

T2W

I+DWI

0.811(0.772–0.850)

0.847(0.798–0.895)

0.762(0.701–0.823)

0.833(0.723–0.943)

0.852(0.807–0.897)

0.938(0.901–0.975)

T2W

I+DCE

0.777(0.734–0.820)

0.840(0.794–0.887)

0.680(0.611–0.750)

0.822(0.739–0.906)

0.852(0.807–0.898)

0.927(0.882–0.972)

DWI+DCE

0.787(0.744–0.830)

0.856(0.812–0.900)

0.701(0.632–0.770)

0.865(0.789–0.942)

0.855(0.808–0.902)

0.933(0.890–0.976)

T2W

I+DWI+DCE

0.812(0.771–0.853)

0.871(0.832–0.911)

0.731(0.665–0.797)

0.879(0.810–0.948)

0.877(0.833–0.921)

0.955(0.921–0.989)

ROCreceiver

operatingcharacteristics,T2W

IT2-weightedim

aging,DWIdiffusion-weightedim

aging,DCEdynamiccontrast-enhancedim

aging,GSGleason

Score

Eur Radiol

recent study demonstrated a benefit for determining PCalocalisation by combining T2WI, DWI and DCE [21].

ROC analysis of peripheral-zone PCa resulted inexcellent test accuracy both for all peripheral PCaand especially for higher-grade peripheral-zone PCa.Besides DWI, DCE was the most important MR se-quence for peripheral-zone PCa detection, with analmost equal AUC. Puech et al. called DCE imagingone of the cornerstones of mp-MRI, as it improvesaccuracy, detection and evaluation of PCa aggressive-ness [10]. Those authors pointed out that there arecertain lesions visible on T2WI or DWI alone, not onDCE; however, there are also lesions solely detectableby DCE. DCE showed far better accuracy for detect-ing peripheral-zone compared with transitional-zonePCa [18, 22].

Detecting transitional-zone PCa is much more chal-lenging than in the peripheral zone. For transitional-zone PCa, DWI or the combination of T2WI and DWI

achieved the highest test accuracy. Hoeks et al. reportedthat ADC values derived from DWI can differentiatecancerous from noncancerous transitional zone, but therewas only a moderate accuracy for differentiating higher-from lower-grade PCa [23]. A combination of sequence-specific PI-RADS scores (T2WI + DWI + DCE) seemsto slightly improve the detection rate of higher-gradetransitional-zone PCa. Conversely, varying evidence thatmp-MRI may not at all improve detection accuracy andlocalization of transitional-zone PCa at 3T has also beenpublished [12, 18]. Nevertheless, our data suggest theneed for further improvement of the PI-RADS scoringsystem; for instance, recommendations for weightingand selecting MR sequences for transitional-zone PCa diag-nostics. The PI-RADS scoring system distinguishesperipheral- and transitional-zone lesions on T2WI but doesnot offer differentiation in scoring DWI or DCE. Consideringour results, it seems that at least the DCE has a differentappearance regarding its localization. Neglecting DCE in

Fig. 3 Receiver operating characteristic (ROC) analysis. a, b All malig-nant lesions; and additionally, c, d of all higher-grade prostate cancer(PCa) lesions; primary Gleason pattern≥4 for Prostate Imaging Reporting

and Data System (PI-RADS) score of each magnetic resonance (MR)sequence and the addition of the score of two or all three MR sequences

Eur Radiol

transitional-zone lesions could possibly be a key to bettercancer detection and subsequent overtreatment. However,for clinical routine, mp-MRI excluding DCE cannot be rec-ommended according to our data due to the additional benefitin reading accuracy depending on tumour location and grade.

This study has some limitations. MR-guided in-bore pros-tate biopsies were used as the reference standard, and noradical prostatectomy or long-term follow-up was performed.Therefore, potentially missed lesions cannot be ruled outwhen using this technique. Nevertheless, MR-guided in-borebiopsy can optimally guarantee an exact histopathologicalcorrelation of all included lesions. In addition, we did notdifferentiate between biopsy-naive patients and patients withprior negative TRUS biopsy.Mean PSAvalues were relativelyhigh, which may have caused selection bias. Furthermore,different readers performed the PI-RADS score assessmentandMR-guided biopsy. Confidence intervals of ROC analysisoverlap in some cases. Larger and more independent patientcollectives can potentially clarify the challenge between clin-ical and statistical significance. Finally, the retrospective de-sign of this study a limitation.

In conclusion, the combination of three MR se-quences (T2WI, DWI and DCE) achieved the highesttest accuracy for peripheral and higher-grade PCa detec-tion in our patient population. For lower-gradetransitional-zone PCa, our results showed better accura-cy by neglecting DCE. Nonetheless, the use of two orsingle MR sequences led to lower AUC in patients withhigher-grade PCa. Our data suggest that the PI-RADSscoring system needs further improvement with respectto weighting and selection of MR sequences and regard-ing specific criteria for transitional-zone PCa.

Acknowledgments The scientific guarantor of this publication isDirk Blondin. The authors of this manuscript declare no relation-ships with any companies, whose products or services may berelated to the subject matter of the article. The authors state thatthis work has not received any funding. No complex statisticalmethods were necessary for this paper. Institutional Review Boardapproval was obtained. Written informed consent was obtainedfrom all patients in this study. No study participants or cohortshave been previously reported. Methodology: retrospective, diag-nostic or prognostic study, performed at one institution.

Fig. 4 Receiver operating characteristic (ROC) analysis of a, b all andadditionally all higher-grade prostate cancer (PCa) lesions (primaryGleason pattern≥4) of peripheral zone and c, d transitional zone for the

summed Prostate Imaging Reporting and Data System (PI-RADS) scoreof two or all three magnetic resonance (MR) sequences

Eur Radiol

References

1. Murphy G, Haider M, Ghai S et al (2013) The expanding role ofMRIin prostate cancer. AJR 201:1229–1238

2. Barentsz JO, Richenberg J, Clements R et al (2012) ESUR prostateMR guidelines 2012. Eur Radiol 22:746–757

3. Mozer P, Cornud F, Renard-Penna R, Misrai V, Thoulouzan M,Malavaud B (2012) Validation of the European Society ofUrogenital Radiology scoring system for prostate cancer diagnosison multiparametric magnetic resonance imaging in a cohort of repeatbiopsy patients. Eur Urol 62:986–996

4. Kuru TH, Roethke MC, Rieker P, Roth W, Fenchel M, HohenfellnerM, Schlemmer HP, Hadaschik BA (2013) Histology core-specificevaluation of the European Society of Urogenital Radiology (ESUR)standardised scoring system of multiparametric magnetic resonanceimaging (mpMRI) of the prostate. BJU Int 112:1080–1087

5. Schimmöller L, Quentin M, Arsov C, Lanzman RS, Hiester A,Rabenalt R, Antoch G, Albers P, Blondin D (2013) Inter-readeragreement of the ESUR score for prostate MRI using in-bore MRI-guided biopsies as the reference standard. Eur Radiol 23:3185–3190

6. Rosenkrantz AB, Lim RP, Haghighi M, Somberg MB, Babb JS,Taneja SS (2013) Comparison of interreader reproducibility of theprostate imaging reporting and data system and Likert scales forevaluation of multiparametric prostate MRI. AJR 201:601–618

7. Westphalen AC, Rosenkartz AB (2014) Prostate Imaging Reportingand Data System (PI-RADS): reflections on early experience with astandardized interpretation scheme for multiparametric prostate MRI.AJR 202:121–123

8. Mazaheri Y, Shukla-Dave A, Muellner A et al (2008) MR imaging ofthe prostate in clinical practice. MAGMA 21:379–392

9. Osugi K, Tanimoto A, Nakashima J, Shinoda K, Hashiguchi A, OyaM, Jinzaki M, Kuribayashi S (2013) What is the most effective toolfor detecting prostate cancer using a standard MR scanner? MagnReson Med Sci 12:271–280

10. Puech P, Sufana-Iancu A, Renard B, Lemaitrea L (2013) ProstateMRI: can we do without DCE sequences in 2013? Diagn IntervImaging 94:1299–1311

11. Sabach A, Babb JS, Matza BW, Taneja SS, Deng FM (2013) Prostatecancer: comparison of dynamic contrast-enhanced MRI techniquesfor localization of peripheral zone tumor. AJR 201:471–478

12. Hoeks CMA, Hambrock T, Yakar D, Hulsbergen-van de Kaa CA,Feuth T, Witjes JA, Fütterer JJ, Barentsz JO (2013) Transition zoneprostate cancer: detection and localization with 3-T multiparametricMR imaging. Radiology 266:207–217

13. Dickinson L, Ahmed HU, Allen C, Barentsz JO, Carey B et al (2011)Magnetic resonance imaging for the detection, localization and char-acterisation of prostate cancer: recommendations from a Europeanconsensus meeting. Eur Urol 59:477–494

14. Quentin M, Schimmöller L, Arsov C, Rabenalt R, Antoch G, AlbersP, Blondin D (2013) Increased signal intensity of prostate lesions onhigh b-value diffusion-weighted images as a predictive sign of ma-lignancy. Eur Radiol 24:209–213

15. Röthke M, Blondin D, Schlemmer HP, Franiel T (2013) PI-RADSclassification: structured reporting for MRI of the prostate. Röfo 185:253–261

16. Dickinson L, Ahmed HU, Allen C, Barentsz JO, Carey B, Futterer JJ,Heijmink SW, Hoskin P, Kirkham AP, Padhani AR, Persad R, PuechP, Punwani S, Sohaib A, Tombal B, Villers A, Emberton M (2013)Scoring systems used for the interpretation and reporting ofmultiparametric MRI for prostate cancer detection, localization, andcharacterization: could standardization lead to improved utilization ofimaging within the diagnostic pathway? J Magn Reson Imaging 37:48–58

17. SiddiquiMM,Rais-Bahrami S, TruongH, Stamatakis L, Vourganti S,Nix J, Hoang AN, Walton-Diaz A, Shuch B, Weintraub M, KrueckerJ, Amalou H, Turkbey B, Merino MJ, Choyke PL, Wood BJ, PintoPA (2013) Magnetic resonance imaging/ultrasound-fusion biopsysignificantly upgrades prostate cancer versus systematic 12-coretransrectal ultrasound biopsy. Eur Urol 64:713–719

18. Delongchamps NB, RouanneM, Flam Tet al (2011)Multiparametricmagnetic resonance imaging for the detection and localization ofprostate cancer: combination of T2-weighted, dynamic contrast-enhanced and diffusion-weighted imaging. BJU Int 107:1411–1418

19. Nagel KN, Schouten MG, Hambrock T, Litjens GJ, Hoeks CM, tenHaken B, Barentsz JO, Fütterer JJ (2013) Differentiation of prostatitisand prostate cancer by using diffusion-weighted MR imaging andMR-guided biopsy at 3 T. Radiology 267:164–172

20. Donati OF, Jung SI, Vargas HA, Gultekin DH, Zheng J, MoskowitzCS, Hricak H, Zelefsky MJ, Akin O (2013) Multiparametric prostateMR imaging with T2-weighted, diffusion-weighted, and dynamiccontrast-enhanced sequences: are all pulse sequences necessary todetect locally recurrent prostate cancer after radiation therapy?Radiology 268:440–450

21. Isebaert S, Van den Bergh L, Haustermans K et al (2013)Multiparametric MRI for prostate cancer localization in correlationto whole-mount histopathology. J Magn Reson Imaging 37:1392–1401

22. Chesnais AL, Niaf E, Bratan F, Mège-Lechevallier F, Roche S,Rabilloud M, Colombel M, Rouvière O (2013) Differentiation oftransitional zone prostate cancer from benign hyperplasia nodules:evaluation of discriminant criteria at multiparametric MRI. ClinRadiol 68:323–330

23. Hoeks CMA, Vos EK, Bomers JGR, Barentsz JO, Hulsbergen-van deKaa CA, Scheenen TW (2013) Diffusion-weighted magnetic reso-nance imaging in the prostate transition zone. Histopathologicalvalidation using magnetic resonance guided biopsy specimens.Investig Radiol 48:693–701

Eur Radiol