Prostate Magnetic RChallenges of Im

Ronald Loch, MD, Kathryn FowJoseph Ippolito, MD, PhD, Cary Sie

technology into an efcient multidisciplinary model ofpatient care. This article reviews several of the challengesthat radiologists should be aware of when integrating

identify patients who are appropriate for AS.9-1126There are many arguments in favor of adding MPP-MRI to the AS algorithm. Initial studies show thatMPP-MRI can more accurately classify patients to ASwhen combined with clinical classication schemes(eg, the D'Amico, Epstein, and Cancer of the ProstateRisk Assessment systems).10 MPP-MRI may identifytumors and allow for targeted biopsies, which is

Curr Probl Diagn Radiol 2015;44:2637.& 2014 Mosby, Inc. All rights reserved.0363-0188/$36.00 + 0http://dx.doi.org/10.1067/j.cpradiol.2014.05.009

From the Mallinckrodt Institute of Radiology, Washington UniversitySchool of Medicine, St. Louis, MO.Reprint requests: Ronald Loch, MD, Mallinckrodt Institute of Radiology,510 South Kingshighway Blvd, St. Louis, MO 63110. E-mail: ronloch@gmail.com.

Curr Probl Diagn Radiol, January/February 2015screening have shown increased diagnosis of prostatecancer in screened populations, leading to increased

increasingly being used to document tumor burden inpatients and offers a potentially powerful tool to helpprostate MRI into their clinical practice.

IntroductionProstate cancer is a common cause of cancer andcancer deaths among men in the United States, with anestimated 233,000 new cases and 29,480 deaths eachyear.1 In recent decades, screening and diagnosis ofprostate cancer relied on serum prostate-specic anti-gen (PSA) and digital rectal examination, followed bytargeted or saturation biopsy for positive screens.However, PSA screening has become controversial.Recent randomized trials evaluating the utility of PSAProstate cancer is among the most common causes ofcancer and cancer deaths in men. Screening methods andoptimal treatments have become controversial in recentyears. Prostate magnetic resonance imaging (MRI) is gain-ing popularity as a tool to assist diagnosis, risk assessment,and staging. However, implementation into clinical practicecan be difcult, with many challenges associated withimage acquisition, postprocessing, interpretation, repor-ting, and radiologic-pathologic correlation. Although state-of-the-art technology is available at select sites for target-ing tissue biopsy and interpreting multiparametric prostateMRI, many institutions struggle with adapting this newesonance Imaging:plementation

ler, MD, Ryan Schmidt, MD,gel, MD, and Vamsi Narra, MD

interventions but no difference in prostate cancermortality or all-cause mortality.2,3 The lack of survivalbenet and added morbidity associated with interven-tion ultimately led to a grade D recommendation4

against routine PSA screening by the U.S. PreventiveServices Task Force (USPSTF).Beyond screening, the optimal treatment of local-

ized prostate cancer is controversial with studiesdemonstrating little or no mortality benet whencomparing radical prostatectomy with observation.5-7

The inability to prospectively identify and differentiatehigh-risk tumors from indolent tumors, many times,leads to overtreatment and the psychological stressassociated with a cancer diagnosis for patients.8

In the wake of the USPSTF recommendation againstroutine PSA screening and the mounting evidenceagainst radical prostatectomy for men with low-risktumors, there has been renewed interest in observa-tional management. In particular, there is increasingenthusiasm for active surveillance (AS) of prostatecancer. AS denes monitoring and treatment triggersaimed at minimizing interventions for indolent can-cers.9 Optimal monitoring and treatment algorithms forAS are not yet well dened within the urologycommunity. Gleason score on repeat biopsies, serialPSA monitoring, and patient preferences play a keyrole in determining treatment strategy. Multiparametricprostate magnetic resonance imaging (MPP-MRI) is

important to accurately assess tumor grade given thatsaturation biopsies randomly sample less than 0.5% ofthe gland. Early retrospective studies have shown thatMPP-MRI may result in removal of up to 29% ofpatients from AS after conrmatory repeat biopsy.Likewise, MPP-MRI may provide a high negativepredictive value for large or high-grade tumors,allowing patients to stay on AS with more con-dence.11 Table 1 summarizes the potential roles of

pathologic correlation. This article provides an

imaging (DWI) in addition to acquiring T2-weightedimages (T2WIs), and more than half of centers useddynamic contrast enhancement (DCE) in addition toT2WI and DWI. Only 21% of surveyed centers usedMR spectroscopy, all of which were academic centers.Guidelines and consensus statements for prostate MRIacquisition and reporting have been proposed inEurope13-15 but not yet by a North American society.These guidelines and a thorough review of theliterature by Hoeks et al16 suggest that MPP-MRI beperformed with a minimum of DWI and DCE inaddition to T2WI. We therefore focus our discussionon these 3 techniques. The challenges of prostate MRIprotocol design are discussed later with reference tosupportive literature where available. We also discussour approach and solutions to challenges at our own

both phased-array coil and an integrated endorectal-pelvic phased-array coil system during the same

TABLE 1. Potential benets of MPP-MRI in ASImprove negative predictive value of tumor burden assessmentImprove accuracy of tumor grade determinationReduce unnecessary biopsiesContribute to accurate assignment of patients to ASProvide a tool to follow up patients noninvasively once diagnosis isestablished

D

EcS

189Noverview of the role of MPP-MRI in AS as well asthe challenges associated with its implementation.

Challenges in AcquisitionThere is no uniformly agreed protocol in the com-munity. In a survey of predominantly academiccenters, there was a near-even split between acquiringimages using an endorectal coil at 1.5 T, a pelvicphased-array coil at 3 T, and an endorectal coil at3 T.12 Nearly all centers performed diffusion-weighted

TABLE 2. Nominal MR parametersProtocols T2WI

Sequence type Turbo spin echoOptions Phase encoding in left-right direction

which helps reduce motion artifactfrom rectal peristalsis

TR 4000-6000 msTE 90-120 msFlip angle 180 (4150)Echo train length 8-16MPP-MRI in AS.In light of the possible clinical utility of MPP-MRI,

the volume of imaging requests have gone up in manycenters, and radiologists must be familiar with theperformance and interpretation of these studies.MPP-MRI can be challenging for many reasons,including issues related to technical acquisition, post-processing, interpretation, reporting, and radiologic-Matrix size 512 512 1Field of view (cm2) 16 16 1Number of averages 4 5Slice thickness (mm) 3 3

NA, not applicable.

Curr Probl Diagn Radiol, January/February 2015examination. The endorectal coil images showed sig-nicantly improved prospective diagnostic accuracy(from 59%-83%) and more importantly specicity(from 62%-98%) of extracapsular extension detected

WI DCE

ho planar imaging Spoiled 3D gradient echopectral fat saturation, partialFourier acquisition, b 0, 500,and 1000 s/mm2

Acquire T1 map before contrastadministration if a quantitativeanalysis is desired

1,200 ms Minimal TR9 ms Minimal TE0 6-12A 128 128 256 2566 16 16 16

1-23

27institution. Table 2 shows typical sequence parame-ters. Table 3 shows some of the more commonchallenges encountered with the various sequencesas well as potential solutions to these challenges.

Field Strength and Coil SelectionMost academic centers use an endorectal coil when

imaging at 1.5 T, and most of the literature supportsreasonable accuracy for staging with this coil selec-tion.12,16 The performances of endorectal coils andphased-array coils in the same patient population at1.5 T was prospectively compared by 2 studies.17,18

In 1 study, 81 patients underwent prostate MRI using

endorectal coil. Placement requires radiologist pres-

TABLE 3. Challenges and solutionsSequence Challenge Solutions

itionincrtal mectiotum

ducmotat whole-mount pathologic examination.18 The limi-tation of this and other studies evaluating endorectalcoil at 1.5 T is that the focus was on staging and notprospective diagnosis, hence DWI and DCE sequen-ces were not performed (ie, the accuracy of MPP-MRIwas not truly tested). At our institution, MPP-MRI at1.5 T with an endorectal coil is reserved only forpatients who have (1) a hip arthroplasty or othermaterials that may lead to excessive artifact at 3 T or(2) MR-conditional surgical devices that may not besafe at 3 T.Another parameter under debate is whether an

endorectal coil should be used when imaging at 3 T.Initially as 3 T scanners became more widely avail-able, several studies compared performance of 1.5 Twith endorectal coil and 3 T with phased-array coil. Intheory, the added signal-to-noise ratio (SNR) fromimaging at higher eld strength may obviate the needfor an endorectal coil. Initial studies demonstratedlittle difference in diagnostic performance between1.5 T with endorectal and 3 T with phased-arraycoil.19-21 As endorectal coils became available for3 T systems, the question of coil selection wasrevisited. In a study of 46 men undergoing prostatec-tomy, performance of preoperative MRI performedwithout and with an endorectal coil was evaluated,with whole-mount pathologic correlation. The use ofan endorectal coil improved image quality, tumordetection, and staging performance in a subgroupanalysis of experienced radiologists.22 Again, this

T2WI Wrap artifact Increase FOV, phase oversample, and posT2WI Motion artifact Reduce imaging time by parallel imaging,

Consider glucagon injection to reduce recAssign phase-encoding in the right-left dir

DWI Susceptibilityartifact

Reduce rectal gas with enemas, empty recpatients with hip arthroplasties

DCE Motion artifact Phase encoding in the left-right direction resoftware for correction of gross patient

FOV, eld of view.study was focused primarily on staging with T2WIand did not include DWI or DCE. A more recent studyaimed at assessing the diagnostic value of MPP-MRI(including DWI and T2WI) was performed in20 patients with 51 foci of cancer detected onwhole-mount pathologic analysis.23 The authorsreported signicantly improved sensitivity and pos-itive predictive value when using an endorectal coil,with a trend of detecting more small cancers with the

28ence at the imaging center; the retention balloon mustbe lled with special media (peruorocarbon or a lessexpensive alternative such as barium) at 3 T to preventdistortion from gas, and there is added cost and time tothe overall examination. Additionally, the coil maydistort the prostate gland, having implications forradiologic-pathologic correlation and targeting biop-sies, which are discussed in detail in the subsequentsections. Further investigation is needed to draw moredenitive conclusions in this area.

T2-Weighted ImagingT2WI is an essential part of MPP-MRI, providing

anatomical detail of the entire prostatic gland. Sometypical sequence parameters are reviewed in Table 2.Fast spin-echo acquisitions are performed in at least2 planes including the transverse plane. The transverseimages should be oriented orthogonal to the long axisof the urethra (although the urethra curves anteriorly atendorectal coil. Despite early evidence of equivalence,of the 5 patients with extracapsular extension, 4 weredetected with the endorectal coil and only 1 wasdetected without the endorectal coil.Although there is some evidence that suggests the

use of an endorectal coil may improve imagingaccuracy, there are other considerations to take intoaccount when designing a protocol. Patient comfortand compliance may be reduced with the use of an

hands out of imaging planeeased echo train lengthotionn to reduce rectal motion transmitted through the prostatebefore imaging, displace gas with endorectal coil, and image at 1.5 T in

es rectal motion artifact, antiperistaltic medications, and postprocessingionthe level of the verumontanum prohibiting a trueorthogonal plane throughout). The high-resolutionand small eld-of-view requirements for this sequencepresent some unique challenges. The phase-encodingdirection should ideally be oriented right to left toprevent any motion artifact from rectal peristalsis to betransmitted through the prostate. If inadequate eld ofview or phase oversampling is selected, this can resultin wrap artifact or aliasing such as from the arms

Curr Probl Diagn Radiol, January/February 2015

also aid in detection of transitional zone tumors.29

Some typical sequence parameters are reviewed in

values (41000 s/mm2). A major limitation of ultrahighalong the patient's sides (Fig 1). Because the right-to-left direction is often the longest dimension in thepelvis, this adds to the length of the acquisition. Thelength of acquisition is important both for workowefciency and to help prevent bulk motion, whichtends to increase along with the time of the acquis-ition. Measures such as parallel imaging, increasedecho train length, and partial k-space lling acquis-itions can be implemented to shorten acquisitionlength, but may come at the expense of decreasedSNR and increased image blurring. Specic absorp-tion rate (SAR) at 3 T is another issue, and some

FIG 1. Small eld of view and left-right phase-encoding direction usedon T2WI increases chances for wrap artifact (indicated by the arrow) ifprotocol is not optimized. In this case, the patient's hands werepositioned at the side. The patient's hands were repositioned. Increas-ing FOV and phase oversampling are other solutions to eliminatingwrap artifact. FOV, eld of view.vendors provide solutions to SAR without compro-mising T2 contrast or adding pauses, such ashyperecho-turbo spin-echo sequences and low SARradiofrequency pulses.In addition to phase encoding in the left-right

direction, agents that inhibit peristalsis (eg, glucagon)can be given at the time of examination to reduce theeffect of rectal peristalsis. The use of antiperistalticagents in the absence of an endorectal coil isincompletely evaluated in the literature and is oftendetermined by radiologist preference and workowconsiderations.

Diffusion-Weighted ImagingThe addition of DWI to MPP-MRI can improve

diagnostic accuracy,24,25 indicate information aboutGleason grade26 and cellular density,27,28 and may

Curr Probl Diagn Radiol, January/February 2015b values (41000 s/mm2) is the relatively poor SNR,especially in the absence of an endorectal coil. One ofthe initial studies using ultrahigh b values reported thata b value of 1000 s/mm2 outperformed a b value of2000 s/mm2, but this study was limited by the use of alonger TE in the b 2000 s/mm2 acquisition.30 Subse-quent studies reported improved performance of DWIacquired with high b values, but they did not includeADC maps in their comparisons.31 When including thecorresponding ADC maps, there have been mixedreports.32,33 Although there is no consensus in theliterature, at our institution, we use 3 b values with amaximum of 1000 s/mm2.Optimization of DWI is perhaps the most impor-

tant component (and hence worth spending the mostacquisition time on) of the MPP-MRI as it has shownthe highest diagnostic accuracy.25 DWI in the pelvisis acquired as a free-breathing sequence as respira-tory motion is not a major factor. Increasing aver-ages, reducing TE, and optimizing echo spacing andnoise level are important factors for improving theSNR. Most MRI vendors provide the suggestedparameters.The greatest limitation of DWI imaging relates to

echo planar pulse sequence design, making it sensitiveto artifacts. Susceptibility artifacts related to rectal gasand metal implants are the most commonly encoun-tered artifacts (Fig 2). Strategies to reduce suscepti-bility artifact from gas in the rectum include use ofenemas, instructions to empty the rectum immediatelybefore imaging, or imaging with an endorectal coil. Inpatients with hip prostheses, imaging at 1.5 T shouldbe considered to reduce the associated susceptibilityartifact.

DCE ImagingDetection of prostate cancer with DCE relies on the

increased vascular permeability of tumor neovascula-ture typically seen in tumors, and has been shown toadd value to MPP-MRI. Some typical sequence

29Table 2. DWI requires at least 2 acquisitions withdifferent b values to calculate an apparent diffusioncoefcient (ADC). The European guidelines suggestapplying diffusion gradients in 3 orthogonal directionswith b values14 of 0, 800, and 1000 s/mm2. There havebeen conicting reports regarding the utility of high b

10 seconds has been suggested.13-15 Gadolinium-based contrast is typically injected via an antecubitalparameters are reviewed in Table 2. DCE imagingrequires serial fast 3-dimensional (3D) spoiledgradient-echo acquisitions over a period of approx-imately 4-6 minutes following administration of intra-venous gadolinium contrast. Differentiation betweenthe vascularity of the tumor and normal prostatenecessitates a high temporal resolution and completecoverage of the gland, which comes at the cost ofspatial resolution and SNR.Assigning optimal parameters can be challenging

and often requires a compromise between T1 weight-ing, SNR, and temporal resolution. Imaging for atleast 5 minutes with a temporal resolution less than

FIG 2. Warped diffusion-weighted image (A) owing to air in therectum (despite use of KY jelly). Gas in the rectum is a major hurdle toimaging, as it results in phase distortion. In this case, a Gleason9 adenocarcinoma (indicated by the arrow) was found in the middleright peripheral zone on T2WI (B) but obscured on the DWI. Use of anendorectal coil can displace gas in the rectum. Some promote the useof KY jelly to displace gas from the rectum.

30vein at 2-4 mL/s, followed by a 20-mL saline ush.Choosing a minimal TE improves T1 weighting butdecreases the SNR. Increasing the ip angle increasesT1 weighting but decreases the SNR. Utilization ofk-space undersampling techniques can help to achievegood spatial resolution and SNR without loss oftemporal resolution.34 In a detailed review and scien-tic article by Li et al,35 the authors suggested using aip angle of twice the optimal Ernst angle expectedbased on the precontrast images, as the Ernst anglechanges with the arrival of contrast. In addition to theDCE images, a T1 map, generated from multiple ipangle T1-weighted sequences before contrast admin-istration, is required to derive the relationship betweenvoxel signal intensity on postcontrast images andgadolinium concentration. This relationship is neces-sary for quantitative postprocessing.The serial acquisition over a period of approxi-

mately 4-6 minutes makes DCE images particularlyvulnerable to motion, including from the rectum,bladder, and gross patient motion. Most postprocess-ing systems have rigid motion correction that canrealign images.Interpretation of DCE images may be intimidating

to some radiologists, especially those without experi-ence in postprocessing this type of data set. DCEimages can be interpreted using qualitative, semi-quantitative, or quantitative methods, and an excep-tional review of these techniques is discussed byVerma et al.36 There is no consensus regarding theuse of quantitative methods in routine practice, butthere is support for performing at least semiquantita-tive analysis.37 The most widely used model forquantitative analysis is the Tofts model, which calcu-lates Ktrans, Kep, and Ve. Ktrans represents the trans-fer constant from the blood plasma space to theextracellular extravascular space. Kep represents thetransfer constant of contrast in the opposite direction.Ve represents the volume of the extracellular extrava-scular space and equals Ktrans divided by Kep. Thismodel requires an arterial input function (AIF), eitherfrom a region of interest selection over the arteries inthe patient or a pooled standardized AIF adjusted forindividual patient factors. Pooled standardized AIFsare offered with some software packages and can beadjusted based on the patient's weight and volume ofcontrast injected. There are several commerciallyavailable and independent institutional postprocessing

Curr Probl Diagn Radiol, January/February 2015

systems for prostate DCE. However, no single methodhas been tested and shown to be superior. The

eligible for AS, 48.7% of patients with tumorso1 cm in diameter measured on ADC maps hadcommercially available systems often offer manymodeling options and differ mostly in their reportingfunctions, usability, access to technical support, andcost. The decision to purchase a postprocessing sys-tem should be supported by on-site testing of thesystem whenever possible to ensure user preference.

Challenges in Interpretation and ReportingProviding a clinically meaningful interpretation of aprostate MRI examination begins with knowledge ofthe patient's clinical history and the indication for theexamination. Referring urologists may prefer theradiologist tailoring the interpretation to focus onspecicity and staging information, such as evaluationfor extracapsular extension in patients being consid-ered for curative therapies. Alternatively, in a patientwith rising PSA but negative random biopsies, thereport should perhaps focus on sensitivity and detect-ing foci of potential cancer. A multidisciplinaryapproach including the radiologist, urologist, andpathologist is essential to developing a reportingsystem that provides this information. The perform-ance of prostate MRI under different clinical situationshas been extensively studied and was thoroughlyreviewed by Hoeks et al.16 In this section, we discussthe typical appearance of prostate cancer, pitfalls, andcurrent approaches to reporting.

InterpretationThe classic features of prostate cancer include dis-

crete hypointense lesions on T2WI, diffusion restric-tion on ADC maps, and brisk enhancement followedby washout on the DCE images. Signs of extracap-sular extension include asymmetric capsular bulge,obliteration of the rectoprostatic angle, asymmetry ofthe neurovascular bundle, and seminal vesicle inva-sion. Figure 3 shows a tumor demonstrating classicfeatures on T2WI, the ADC map, and processed DCEimages.There is a growing body of literature to suggest that

in addition to diagnosis, prognostic information maybe acquired by MPP-MRI. The T2 intensity of thecancer relative to muscle38 and its ADC value26

inversely correlate with Gleason grade. The ADCvalue of focal lesions may also correlate with like-lihood of repeat biopsy showing adverse histology inlow-risk patients on AS.39 In a study of patients

Curr Probl Diagn Radiol, January/February 2015insignicant disease at prostatectomy vs only 24.7%of patients with tumors 41 cm.40 The ability forMPP-MRI to provide this additional information ispromising and requires further validation.Although the classic features of prostate cancer are

well documented in the literature, in practice there aremany sources of error leading to both false-positive andfalse-negative results.41 T2 hypointensity and possiblefalse-positive reading can result in patients with pros-tatitis, hemorrhage from prior biopsy, radiation, orhormonal-deprivation treatment. Wedge-shaped lesionsand lesions with diffuse extension but no mass effect inthe peripheral zone have been reported as signs ofbenignity.42 If MPP-MRI is performed after a recentbiopsy, it should perhaps be delayed 4-6 weeks to helpmitigate the effects of hemorrhage. Although biopsycan complicate the interpretation, postbiopsy hemor-rhage can also unmask a focus of cancer. This isreferred to as the hemorrhage exclusion sign. Speci-cally, hemorrhage manifest as T1 hyperintense bloodproducts (ie, methemoglobin) will appear as hyper-intense on T1WIs and will outline a tumor appearinghypointense on T2WI. This has been termed thehemorrhage exclusion sign and has been reported ashighly specic when present with a positive predictivevalue of 95%-96%.43 An example is shown in Figure 4.False-negative results are also not uncommonly

encountered. Sparse or diffuse tumors have beenshown to have signal intensity on T2WI and ADCssimilar to normal peripheral zone parenchyma.27

Additionally, sensitivity for detection of tumorssmaller than 1 cm3 decreases in tumors with lowerGleason grades (such as Gleason 6).44 It may beargued that some of these are the tumors that do notmerit intervention.40

The high incidence of benign prostatic hyperplasia(BPH) complicates assessment of the transitional zoneand deserves special attention. On T2WI, the presenceof homogeneously low signal, ill-dened margins,lack of capsule, lenticular shape, and invasion of theanterior bromuscular stroma have been studied assigns to discriminate tumors from BPH nodules.45

There have been conicting reports regarding the useof MPP-MRI as opposed to only T2WI, with somestudies reporting no signicant improvement in diag-nostic accuracy when DWI and DCE are added,25,46,47

despite a single study showing the promise of DCE inimproving accuracy.48 There is, however, a growing

31

body of literature to suggest that the use of b values of1000 s/mm2 or greater may improve tumor detectionin the transitional zone.29,31,49 Further research isneeded to verify the usefulness of DWI in detectingtransitional zone tumors. In our practice, the T2WI ismost important when assessing tumors within a regionof BPH.

ReportingAs with any radiologic study, reports should be well

structured and should highlight clinically relevantinformation in a concise manner. The Prostate Imag-ing Reporting and Data System was proposed by theEuropean Society of Urogenital Radiology14 with agoal of reducing interreader variability, improving

FIG 3. T2WI (A), ADC map (B), kTrans map (C), and area under theadenocarcinoma demonstrating a triple parametric match involving the anbromuscular stroma (indicated by the arrows). Owing to their anterior locbiopsies or at prostatectomy with some partial embedding protocols. (Colo

32diagnostic value, and improving communicationbetween radiologists and urologists. In this system,the location, maximal diameter, and a summed overallscore are assigned to each lesion detected. The loca-tion is based on dividing the prostate into either 16 or27 regions. The overall Prostate Imaging Reportingand Data System score for a lesion is the sum ofscores given to lesions for each technique. A lesion'sscore ranges from 1-5, with a score of 5 given tolesions that are highly likely to contain signicantcancer. Extracapsular extension, seminal vesicle inva-sion, distal sphincter invasion, bladder invasion, andother pertinent ndings such as osseous or nodaldisease are also reported. Studies evaluating theaccuracy of this scale when correlated with targetedbiopsies using MRI-3D ultrasound (US) fusion

curve (AUC) map (D). These images depict pathologically proventerior horn of the peripheral zone on the left as well as the anterioration, these tumors are more likely to be missed on random transrectalr version of gure is available online.)

Curr Probl Diagn Radiol, January/February 2015

intermediate-risk groups, the performance of T2WI,DCE, and MR spectroscopy were evaluated with

PZ, peripheral zone.guidance have demonstrated promising results.50,51

In a study using MR-guided biopsies, there wasmoderate interreader agreement, high sensitivity, andhigh negative predictive values.52 This system maynot perform as well regarding accuracy and interreaderagreement for transitional zone tumors, and renementfor tumors in this location may be needed.53,54 Inaddition to continued renement, the development ofan image-rich atlas similar to the BI-RADS atlaswould likely be useful to radiologists interpretingFIG 4. The hemorrhage exclusion sign. A large tumor involving muchof the right PZ, anterior gland, and central gland is outlined byhemorrhage in the uninvolved left peripheral zone on this precontrastT1WI. The left border of the tumor is indicated by the arrowhead andis visible owing to the hemorrhage in the normal left peripheral zone,indicated by the arrow. T2WI, DCE, and DWI showed typical ndings.prostate MRI.

Challenges in Radiology-PathologyCorrelationRadiology-pathology correlation deserves specialmention as the accurate correlation of radiologyinterpretations with pathologic results is necessaryfor continued validation of new techniques as wellas for improving our own diagnostic accuracy. Severalmethods of radiologic-pathologic correlation appear inthe literature regarding the type of pathology (whole-mount specimens vs biopsy), method of correlation(image directed biopsy vs mapping based uponlocation coordinates), and stringency with which thecorrelation method is interpreted. An example ofcorrelation with a whole-mount specimen from ourown institution is shown in Figure 5. In a prospectivestudy55 of 70 patients predominantly in low- to

Curr Probl Diagn Radiol, January/February 2015whole-mount specimens as the reference standard. Morespecically, the prostate was divided into 30 regionsbased on anatomical landmarks during both imaginginterpretation and histopathologic examination. In araw stringent approach, a true positive was onlycounted when a lesion detected on imaging corre-sponded to a cancer found at histopathologic examina-tion within the same region. In their neighboringapproach, cancers found at histopathologic examinationin an immediately adjacent region were also counted astrue positives. This was argued to reduce the negativeeffect of prostate shrinkage and deformation as well asdifferences in sectioning plane during grossing of theprostatectomy specimens. The neighboring approachyielded better performance measures. In an attempt toreduce the error introduced when sectioning the prosta-tectomy specimen at angles that may be different thanthe imaging planes, the same group subsequentlyreported using a customized mold to guide the section-ing.56 The mold was generated by a 3D printer and wasbased on an individual's prostate contours as drawn onT2WI. These 2 sophisticated methods of correlatinghistopathologic ndings on prostatectomy specimens areobviously ideal in research settings when compared withother methods used in routine clinical practice, such asthose involving partial embedding of the specimen,57

which may miss more anterior tumors. However, manyacademic and private institutions are not equipped forthis type of tissue processing.In practice, most histopathologic correlation will be

from a combination of the systematic extended sextanttransrectal USguided biopsies, MRI-US fusionguided biopsies, and in-bore MRI-guided biopsies.Promising results have been reported with bothMRI-US fusion-guided biopsies58,59 and in-boreMRI-guided biopsies.60,61 An additional importantconsideration with MRI in the setting of AS is that anegative MPP-MRI examination nding has beenshown to have a high negative predictive value ofdetecting clinically signicant disease at biopsy inlow-risk patients, therefore allowing the urologist toavoid the unnecessary risks of repeat biopsies.62

Regardless of the method of pathologic correlationused, whether in research or clinical practice, it isimperative that multidisciplinary teams criticallythink about their method of validation to advanceemerging techniques and prove diagnostic accuracyand repeatability.

33

Further Challenges in Implementation IntoRoutine PracticeCare should be taken when interpreting the literatureand trying to generalize results to one's own practice.For example, when comparing reported performanceof techniques from different publications, there areoften several variables that differ, including patientpopulation characteristics, eld strength, coil selec-tion, b values in DWI, postprocessing techniques inDCE, denition of positive ndings, and method ofpathologic verication. Additionally, most publishedexperience with prostate MRI represents single insti-tution, often high-end academic centers with smallsample sizes. Hence, the reproducibility of theseresults in general practice remains to be shown.Similarly, although once promising, standard ADC

FIG 5. Prostate adenocarcinoma suspected in the left peripheral zone (indishown). This was conrmed when correlated with whole-mount histopatholored. (Color version of gure is available online.)

34cutoff values reported in the literature cannot bereliably applied to one's practice as ADC has beenshown to be dependent on b values in addition to otherimaging parameters.63,64 Open discussion with theurologists and pathologists is essential to developrealistic expectations and determine the role thatprostate MRI will play in our own institution.

ConclusionIn the post-USPSTF era of prostate cancer screeningand diagnosis, urologists are seeking new tools to helpmanage patients and avoid unnecessary treatments.MPP-MRI has a yet underdeveloped but promisingrole in AS. There are several challenges associatedwith implementing MPP-MRI in clinical practice. Amajor barrier to general acceptance will be precise

cated by the arrows) on T2WI (A), ADC map (B), and DCE images (notgic specimens (C). Note that the areas of conrmed tumor are colored

Curr Probl Diagn Radiol, January/February 2015

radiologic-pathologic correlation, which has a majoreffect on the perceived accuracy of MPP-MRI. Multi-

13. Dickinson L, Ahmed HU, Allen C, et al. Magnetic resonanceimaging for the detection, localisation, and characterisation ofprostate cancer: Recommendations from a European consen-

15. Kirkham APS, Haslam P, Keanie JY, et al. Prostate MRI:Who, when, and how? Report from a UK consensus meetingClin Radiol 2013;68(10):1016-23.sus meeting. Eur Urol 2011;59(4):477-94.14. Barentsz JO, Richenberg J, Clements R, et al. ESUR prostate

MR guidelines 2012. Eur Radiol 2012;22(4):746-57.

Curr Probl Diagn Radiol, January/February 2015disciplinary collaboration between urologists, radiol-ogists, and pathologists is integral to the successfulimplementation of MPP-MRI.

REFERENCES1. Siegel R, Ma J, Zou Z, et al. Cancer statistics, 2014. CA

Cancer J Clin 2014;64(1):9-29.2. Andriole GL, Crawford ED, Grubb RL 3rd, et al. Mortality

results from a randomized prostate-cancer screening trial.N Engl J Med 2009;360(13):1310-9.

3. Rosario DJ, Lane JA, Metcalfe C, et al. Short term outcomesof prostate biopsy in men tested for cancer by prostate specicantigen: Prospective evaluation within ProtecT study. Br MedJ 2012;344:d7894.

4. Moyer VA. U.S. Preventive Services Task Force. Screeningfor prostate cancer: U.S. Preventive Services Task Forcerecommendation statement. Ann Intern Med 2012;157(2):120-34.

5. Iversen P, Madsen PO, Corle DK. Radical prostatectomyversus expectant treatment for early carcinoma of the prostate.Twenty-three year follow-up of a prospective randomizedstudy. Scand J Urol Nephrol Suppl 1995;172:65-72.

6. Wilt TJ, Brawer MK, Jones KM, et al. Radical prostatectomyversus observation for localized prostate cancer. N EnglJ Med 2012;367(3):203-13.

7. Bill-Axelson A, Holmberg L, Ruutu M, et al. Radical pros-tatectomy versus watchful waiting in early prostate cancer.N Engl J Med 2011;364(18):1708-17.

8. Welch HG, Albertsen PC. Prostate cancer diagnosis andtreatment after the introduction of prostate-specic antigenscreening: 1986-2005. J Natl Cancer Inst 2009;101(19):1325-9.

9. Ganz PA, Barry JM, Burke W, et al. National Institutes ofHealth State-of-the-Science Conference: Role of activesurveillance in the management of men with localizedprostate cancer. Ann Intern Med 2012;156(8):591-5.

10. Turkbey B, Mani H, Aras O, et al. Prostate cancer: Canmultiparametric MR imaging help identify patients who arecandidates for active surveillance? Radiology 2013;268(1):144-52.

11. Stamatakis L, Siddiqui MM, Nix JW, et al. Accuracy ofmultiparametric magnetic resonance imaging in conrmingeligibility for active surveillance for men with prostate cancer.Cancer 2013;119(18):3359-66.

12. Leake JL, Hardman R, Ojili V, et al. Prostate MRI: Access toand current practice of prostate MRI in the United States.J Am Coll Radiol 2014;11(2):156-60.16. Hoeks CMA, Barentsz JO, Hambrock T, et al. Prostate cancer:Multiparametric MR imaging for detection, localization, andstaging. Radiology 2011;261(1):46-66.

17. Hricak H, White S, Vigneron D, et al. Carcinoma of theprostate gland: MR imaging with pelvic phased-array coilsversus integrated endorectalpelvic phased-array coils. Radio-logy 1994;193(3):703-9.

18. Ftterer JJ, Engelbrecht MR, Jager GJ, et al. Prostate cancer:Comparison of local staging accuracy of pelvic phased-arraycoil alone versus integrated endorectal-pelvic phased-arraycoils. Local staging accuracy of prostate cancer using endor-ectal coil MR imaging. Eur Radiol 2007;17(4):1055-65.

19. Beyersdorff D, Taymoorian K, Knsel T, et al. MRI of prostatecancer at 1.5 and 3.0 T: Comparison of image quality in tumordetection and staging. Am J Roentgenol 2005;185(5):1214-20.

20. Torricelli P, Cinquantini F, Ligabue G, et al. Comparativeevaluation between external phased array coil at 3 T andendorectal coil at 1.5 T: Preliminary results. J Comput AssistTomogr 2006;30(3):355-61.

21. Park BK, Kim B, Kim CK, et al. Comparison of phased-array3.0-T and endorectal 1.5-T magnetic resonance imaging in theevaluation of local staging accuracy for prostate cancer.J Comput Assist Tomogr 2007;31(4):534-8.

22. Heijmink SWTPJ, Ftterer JJ, Hambrock T, et al. Prostatecancer: Body-array versus endorectal coil MR imaging at3 Tcomparison of image quality, localization, and stagingperformance. Radiology 2007;244(1):184-95.

23. Turkbey B, Merino MJ, Gallardo EC, et al. Comparison ofendorectal coil and nonendorectal coil T2W and diffusion-weighted MRI at 3 Tesla for localizing prostate cancer:Correlation with whole-mount histopathology. J Magn ResonImaging 2013;39(6):1443-8.

24. Langer DL, van der Kwast TH, Evans AJ, et al. Prostatecancer detection with multi-parametric MRI: Logistic regres-sion analysis of quantitative T2, diffusion-weighted imaging,and dynamic contrast-enhanced MRI. J Magn Reson Imaging2009;30(2):327-34.

25. Delongchamps NB, Rouanne M, Flam T, et al. Multipara-metric magnetic resonance imaging for the detection andlocalization of prostate cancer: Combination of T2-weighted,dynamic contrast-enhanced and diffusion-weighted imaging.BJU Int 2011;107(9):1411-8.

26. Hambrock T, Somford DM, Huisman HJ, et al. Relationshipbetween apparent diffusion coefcients at 3.0-T MR imagingand Gleason grade in peripheral zone prostate cancer.Radiology 2011;259(2):453-61.

27. Langer DL, van der Kwast TH, Evans AJ, et al. Intermixednormal tissue within prostate cancer: Effect on MR imagingmeasurements of apparent diffusion coefcient and T2-sparseversus dense cancers. Radiology 2008;249(3):900-8.

28. Zelhof B, Pickles M, Liney G, et al. Correlation of diffusion-weighted magnetic resonance data with cellularity in prostatecancer. BJU Int 2009;103(7):883-8.

29. Yoshizako T, Wada A, Hayashi T, et al. Usefulness ofdiffusion-weighted imaging and dynamic contrast-enhanced

35

magnetic resonance imaging in the diagnosis of prostatetransition-zone cancer. Acta Radiologica 2008;49(10):1207-13.

30. Kim CK, Park BK, Kim B. High-b-value diffusion-weighted

44. Vargas HA, Akin O, Shukla-Dave A, et al. Performancecharacteristics of MR imaging in the evaluation of clinicallylow-risk prostate cancer: A prospective study. Radiologyimaging at 3 T to detect prostate cancer: Comparisonsbetween b values of 1,000 and 2,000 s/mm2. Am J Roent-genol 2010;194(1):W33-7.

31. Katahira K, Takahara T, Kwee TC, et al. Ultra-high-b-valuediffusion-weighted MR imaging for the detection of prostatecancer: Evaluation in 201 cases with histopathological corre-lation. Eur Radiol 2011;21(1):188-96.

32. Kitajima K, Takahashi S, Ueno Y, et al. Clinical utility ofapparent diffusion coefcient values obtained using highb-value when diagnosing prostate cancer using 3 tesla MRI:Comparison between ultra-high b-value (2000 s/mm2) andstandard high b-value (1000 s/mm2). J Magn Reson Imaging2012;36(1):198-205.

33. Rosenkrantz AB, Hindman N, Lim RP, et al. Diffusion-weighted imaging of the prostate: Comparison of b1000 andb2000 image sets for index lesion detection. J Magn ResonImaging 2013;38(3):694-700.

34. Tudorica LA, Oh KY, Roy N, et al. A feasible highspatiotemporal resolution breast DCE-MRI protocol for clin-ical settings. Magn Reson Imaging 2012;30(9):1257-67.

35. Li X, Huang W, Rooney WD. Signal-to-noise ratio, contrast-to-noise ratio and pharmacokinetic modeling considerations indynamic contrast-enhanced magnetic resonance imaging.Magn Reson Imaging 2012;30(9):1313-22.

36. Verma S, Turkbey B, Muradyan N, et al. Overview ofdynamic contrast-enhanced MRI in prostate cancer diagnosisand management. Am J Roentgenol 2012;198(6):1277-88.

37. Rosenkrantz AB, Sabach A, Babb JS, et al. Prostate cancer:Comparison of dynamic contrast-enhanced MRI techniquesfor localization of peripheral zone tumor. Am J Roentgenol2013;201(3):W471-8.

38. Wang L, Mazaheri Y, Zhang J, et al. Assessment of biologicaggressiveness of prostate cancer: Correlation of MR signalintensity with Gleason grade after radical prostatectomy.Radiology 2008;246(1):168-76.

39. Van As NJ, de Souza NM, Riches SF, et al. A study ofdiffusion-weighted magnetic resonance imaging in men withuntreated localised prostate cancer on active surveillance. EurUrol 2009;56(6):981-7.

40. Lee DH, Koo KC, Lee SH, et al. Tumor lesion diameter ondiffusion weighted magnetic resonance imaging could helppredict insignicant prostate cancer in patients eligible foractive surveillance: Preliminary analysis. J Urol 2013;190(4):1213-7.

41. Rosenkrantz AB, Taneja SS. Radiologist, be aware: Tenpitfalls that confound the interpretation of multiparametricprostate MRI. Am J Roentgenol 2014;202(1):109-20.

42. Cruz M, Tsuda K, Narumi Y, et al. Characterization of low-intensity lesions in the peripheral zone of prostate on pre-biopsy endorectal coil MR imaging. Eur Radiol 2002;12(2):357-65.

43. Barrett T, Vargas HA, Akin O, et al. Value of the hemor-rhage exclusion sign on T1-weighted prostate MR images forthe detection of prostate cancer. Radiology 2012;263(3):751-7.

362012;265(2):478-87.45. Akin O, Sala E, Moskowitz CS, et al. Transition zone prostate

cancers: Features, detection, localization, and staging atendorectal MR imaging. Radiology 2006;239(3):784-92.

46. Haider MA, van der Kwast TH, Tanguay J, et al. CombinedT2-weighted and diffusion-weighted MRI for localization ofprostate cancer. Am J Roentgenol 2007;189(2):323-8.

47. Hoeks CMA, Hambrock T, Yakar D, et al. Transition zoneprostate cancer: Detection and localization with 3-T multi-parametric MR imaging. Radiology 2013;266(1):207-17.

48. Li H, Sugimura K, Kaji Y, et al. Conventional MRIcapabilities in the diagnosis of prostate cancer in the transitionzone. Am J Roentgenol 2006;186(3):729-42.

49. Oto A, Kayhan A, Jiang Y, et al. Prostate cancer: Differentiationof central gland cancer from benign prostatic hyperplasia byusing diffusion-weighted and dynamic contrast-enhanced MRimaging. Radiology 2010;257(3):715-23.

50. Portalez D, Mozer P, Cornud F, et al. Validation of theEuropean Society of Urogenital Radiology scoring system forprostate cancer diagnosis on multiparametric magnetic reso-nance imaging in a cohort of repeat biopsy patients. Eur Urol2012;62(6):986-96.

51. Fiard G, Hohn N, Descotes J-L, et al. Targeted MRI-guidedprostate biopsies for the detection of prostate cancer: Initialclinical experience with real-time 3-dimensional transrectalultrasound guidance and magnetic resonance/transrectal ultra-sound image fusion. Urology 2013;81(6):1372-8.

52. Schimmller L, Quentin M, Arsov C, et al. Inter-readeragreement of the ESUR score for prostate MRI usingin-bore MRI-guided biopsies as the reference standard. EurRadiol 2013;23(11):3185-90.

53. Rosenkrantz AB, Kim S, Lim RP, et al. Prostate cancerlocalization using multiparametric MR imaging: Comparisonof Prostate Imaging Reporting and Data System (PI-RADS)and Likert scales. Radiology 2013;269(2):482-92.

54. Rosenkrantz AB, Lim RP, Haghighi M, et al. Comparison ofinterreader reproducibility of the prostate imaging reportingand data system and Likert scales for evaluation of multi-parametric prostate MRI. Am J Roentgenol 2013;201(4):W612-8.

55. Turkbey B, Pinto PA, Mani H, et al. Prostate cancer: Value ofmultiparametric MR imaging at 3 T for detection-histopathologic correlation. Radiology 2010;255(1):89-99.

56. Turkbey B, Mani H, Shah V, et al. Multiparametric 3 Tprostate magnetic resonance imaging to detect cancer: Histo-pathological correlation using prostatectomy specimens proc-essed in customized magnetic resonance imaging basedmolds. J Urol 2011;186(5):1818-24.

57. Sehdev AE, Pan CC, Epstein JI. Comparative analysis ofsampling methods for grossing radical prostatectomy speci-mens performed for nonpalpable (stage T1c) prostatic adeno-carcinoma. Hum Pathol 2001;32(5):494-9.

58. Rastinehad AR, Turkbey B, Salami SS, et al. Improvingdetection of clinically signicant prostate cancer: MRI/TRUSfusion-guided prostate biopsy. J Urol 2013;191(6):1749-54.

Curr Probl Diagn Radiol, January/February 2015

59. Walton Diaz A, Hoang AN, Turkbey B, et al. Can magneticresonance-ultrasound fusion biopsy improve cancer detectionin enlarged prostates? J Urol 2013;190(6):2020-5.

60. Baur ADJ, Maxeiner A, Franiel T, et al. Evaluation of theprostate imaging reporting and data system for the detectionof prostate cancer by the results of targeted biopsy of theprostate. Invest Radiol 2014;49(6):411-20.

61. Hoeks CMA, Somford DM, van Oort IM, et al. Value of 3-Tmultiparametric magnetic resonance imaging and magneticresonance-guided biopsy for early risk restratication in activesurveillance of low-risk prostate cancer: A prospective multi-center cohort study. Invest Radiol 2014;49(3):165-72.

62. Numao N, Yoshida S, Komai Y, et al. Usefulness of pre-biopsymultiparametric magnetic resonance imaging and clinicalvariables to reduce initial prostate biopsy in men with suspectedclinically localized prostate cancer. J Urol 2013;190(2):502-8.

63. Mazaheri Y, Vargas HA, Nyman G, et al. Diffusion-weightedMRI of the prostate at 3.0 T: Comparison of endorectal coil(ERC) MRI and phased-array coil (PAC) MRIThe impactof SNR on ADC measurement. Eur J Radiol 2013;82(10):e515-20.

64. Peng Y, Jiang Y, Antic T, et al. Apparent diffusion coefcientfor prostate cancer imaging: Impact of B values. Am JRoentgenol 2014;202(3):W247-53.Curr Probl Diagn Radiol, January/February 2015 37

Prostate Magnetic Resonance Imaging: Challenges of ImplementationIntroductionChallenges in AcquisitionField Strength and Coil SelectionT2-Weighted ImagingDiffusion-Weighted ImagingDCE Imaging

Challenges in Interpretation and ReportingInterpretationReporting

Challenges in Radiology-Pathology CorrelationFurther Challenges in Implementation Into Routine PracticeConclusionReferences