The British Journal of Radiology, 82 (2009), 148вЂ“156 Dynamic contrast-enhanced MRI for prostate cancer localization 1 A S N JACKSON, FRCR, 2S A REINSBERG, PhD, 3S A SOHAIB, FRCR, 2E M CHARLES-EDWARDS, MSc, S JHAVAR, MD, DMRT, 5T J CHRISTMAS, MD, FRCS, 5A C THOMPSON, FRCS, 6M J BAILEY, MS, FRCS, 7 C M CORBISHLEY, FRCPath, 8C FISHER, MD, FRCPath, 2M O LEACH, PhD and 1D P DEARNALEY, MD, FRCR 4 1 Academic Department of Radiotherapy and Oncology, 2CRUK Magnetic Resonance Imaging Group and, 3Department of Radiology, Royal Marsden NHS Foundation Trust and Institute of Cancer Research, Downs Rd, Sutton SM2 5PT, 4Cancer Genetics and Molecular Carcinogenesis, The Institute of Cancer Research and the Royal Marsden Hospital NHS Foundation Trust, Cotswold Road, Sutton SM2 5NG, 5Department of Urology, Royal Marsden NHS Foundation Trust, Fulham Road, London SW3 6JJ, Departments of 6Urology and 7Cellular Pathology, St Georges Hospital, Blackshaw Road, London SW17 0QT, and 8Department of Histopathology, Royal Marsden NHS Foundation Trust, Fulham Road, London SW3 6JJ, UK ABSTRACT. Radiotherapy dose escalation improves tumour control in prostate cancer but with increased toxicity. Boosting focal tumour only may allow dose escalation with acceptable toxicity. Intensity-modulated radiotherapy can deliver this, but visualization of the tumour remains limiting. CT or conventional MRI techniques are poor at localizing tumour, but dynamic contrast-enhanced MRI (DCE-MRI) may be superior. 18 patients with prostate cancer had T2 weighted (T2W) and DCE-MRI prior to prostatectomy. The prostate was sectioned meticulously so as to achieve accurate correlation between imaging and pathology. The accuracy of DCE-MRI for cancer detection was calculated by a pixel-by-pixel correlation of quantitative DCE-MRI parameter maps and pathology. In addition, a radiologist interpreted the DCE-MRI and T2W images. The location of tumour on imaging was compared with histology, and the accuracy of DCE-MRI and T2W images was then compared. Pixel-by-pixel comparison of quantitative parameter maps showed a significant difference between the benign peripheral zone and tumour for the parameters Ktrans, ve and kep. Calculation of areas under the receiver operating characteristic curve showed that the pharmacokinetic parameters were only вЂ�вЂ�fairвЂ™вЂ™ discriminators between cancer and benign gland. Interpretation of DCE-MRI and T2W images by a radiologist showed DCE-MRI to be more sensitive than T2W images for tumour localization (50% vs 21%; p50.006) and similarly specific (85% vs 81%; p50.593). The superior sensitivity of DCE-MRI compared with T2W images, together with its high specificity, is arguably sufficient for its use in guiding radiotherapy boosts in prostate cancer. Radiotherapy dose escalation improves tumour control rates in prostate cancer, particularly in less favourable tumours [1вЂ“5], but this is at the expense of greater rectal side effects. One strategy to improve the therapeutic ratio of increased radiotherapy would be to dose escalate only discrete tumour foci within the prostate, and contemporary radiotherapy planning techniques are able to create the necessary dose distributions to do this [6, 7]. The major limitation here is difficulty in delineating tumour using available imaging techniques. CT, the standard method for radiotherapy target volume localization, provides poor soft-tissue contrast, and focal tumour in the prostate is often not seen. MRI has superior soft-tissue contrast resolution compared with CT, but conventional techniques still lack the necessary accuracy to delineate tumour satisfactorily. Dynamic contrast-enhanced MRI (DCE-MRI) uses differences in the time course of enhancement following Address correspondence to: A S N Jackson, Academic Department of Radiation Oncology, Christie Hospital NHS Trust, Wilmslow Rd, Manchester M20 4BX, UK. E-mail: Andrew.Jackson@manchester. ac.uk 148 Received 12 December 2007 Revised 12 February 2008 Accepted 7 March 2008 DOI: 10.1259/bjr/89518905 вЂ™ 2009 The British Institute of Radiology intravenous administration of contrast in order to distinguish benign from malignant tissues. Enhancement can be expressed semi-quantitatively using parameters derived from signal intensity changes associated with the passage of contrast agent. Enhancement can also be described qualitatively, based on calibrated contrast agent concentrationвЂ“time curves [8] and pharmacokinetic models used to express the passage of contrast agent between tissue compartments in terms of physiologically relevant parameters. DCE-MRI has been investigated in the prostate; prostate cancer is often found to enhance more quickly and to a greater degree, and to show more washout than the benign peripheral zone [9вЂ“13]. Studies that derived concentrationвЂ“time curves described similar differences between cancerous and benign prostate according to either descriptive parameters relating to the concentrationвЂ“time curve [12] or pharmacokinetic models [14, 15]. In terms of clinical application, some studies have suggested that DCE-MRI may be of use in characterizing hypointense lesions on conventional T2 weighted (T2W) MRI sequences [10, 11]. Some studies selected a single The British Journal of Radiology, February 2009 Dynamic contrast-enhanced MRI in prostate cancer slice for DCE-MRI analysis [10], one thought likely to contain tumour on the basis of T2W sequences [9, 11]. Later studies produced more clinically relevant results by comparing the ability of radiologists to correctly localize tumour to a region of the prostate using DCEMRI and T2W images, and suggesting superiority for DCE-MRI [16, 17] This work describes two approaches to evaluate the ability of DCE-MRI to correctly localize tumour within the prostate, as assessed against histological analysis of whole-mount prostatectomy specimens. Firstly, quantitative pharmacokinetic parameter maps derived from DCE-MRI images were compared on a pixel-by-pixel basis with the distribution of tumour within the prostate, in effect assessing the accuracy of automated tumour delineation. Secondly, a radiologist viewed standard T2W images and raw DCE-MRI data to determine the presence or absence of tumour within specific regions of the prostate. Table 1. Summary of patient characteristics Parameter Mean Range Age (years) PSA (ng lвЂ“1) No. of core biopsies Cores involved (%) Interval biopsy to MRI (weeks) 59 10.76 8 35.7 31 42вЂ“73 4.3вЂ“18.9 6вЂ“12 12.5вЂ“58 8вЂ“151 Parameter Clinical T stage: T1c T2a T2b T3a Unavailable Gleason score: 3+3 3+4 4+4 Number 10 2 1 1 4 12 5 1 PSA, prostate-specific antigen. Methods and materials The research protocol was approved by the Royal Marsden NHS Trust and Institute of Cancer Research Committee for Clinical Research and Regional Ethics Committee, and Wandsworth Regional Ethics Committee. Patients The 19 patients included in this study were undergoing radical prostatectomy as primary treatment for histologically confirmed adenocarcinoma of the prostate. The patients were required to have no contra-indication to MRI scanning and were ineligible if they had received prior hormonal treatment or if prostatectomy was a salvage procedure following radiotherapy. Histological diagnosis of prostate cancer had been reached by transrectal ultrasound-guided biopsy in all cases. Remaining patient characteristics can be found in Table 1. In one patient, the size of the prostate precluded its removal as a whole gland; therefore, correlation between histology and imaging was not feasible for this patient. In two patients, satisfactory calibration data were not acquired at the time of MRI scanning and, consequently, quantitative pharmacokinetic analysis could not take place. Pixel-by-pixel analysis is therefore restricted to 16 patients and radiologist interpretation to 18 patients. examination. The DCE-MRI volume was placed such that its slices interrogated the same position and angulations as those of the T2W scan, although the DCE-MRI volume was limited to eight slices to ensure an acquisition time of ,3 s per image volume. Owing to restrictions on gradient slew-rate, the minimum repetition time achievable with the gradient echo sequence varied with patient-dependent imaging volume angulation. The DCE-MRI scan therefore comprised 150 consecutive imaging time-points of duration 2.9вЂ“3.2 s, during which contrast agent (0.5 mmol mlвЂ“1 gadodiamide (Omniscan; Amersham Health AS, Oslo, Norway)) was administered intravenously after the 10th time-point вЂ“1 to a total dose of 0.4 ml kg at a rate of 3 ml sвЂ“1. Parameters associated with the imaging sequences are detailed in Table 2. Surgery Three consultant urologists collaborated on the project, two of whom perform radical prostatectomy via a retropubic approach, and the other via a transperineal approach. Three patients therefore underwent transperineal prostatectomy and the remainder retropubic prostatectomy. The aim was to remove the gland whole with the seminal vesicles attached. 14 patients underwent surgery within 24 h of MRI, with the remaining 5 patients having surgery 2 days, 4 days, 9 days and 21 days (2 patients) after MRI. Imaging Imaging was performed using a 1.5 T Philips Intera Gyroscan scanner (Philips Medical Systems, Reigate, UK) with a phased-array flexible pelvic coil. A T2W sagittal scan was performed to identify the posterior surface of the prostate. Axial T2W scans were aligned such that 18 6 4 mm contiguous slices were acquired perpendicular to the posterior surface of the gland. A three-dimensional T1 weighted gradient echo sequence comprising eight 4 mm contiguous slices was used in the DCE-MRI The British Journal of Radiology, February 2009 Histopathological analysis The left and right sides (ВЎ the anterior surface) of the prostate were inked to allow subsequent orientation. The seminal vesicles were detached and thin sections from each urethral extremity of the gland were taken with a scalpel to assess margin status. The ex vivo gland was sectioned perpendicularly to its posterior surface using a multi-bladed knife and cradle developed within the institution (UK patent number 2404607) (Figure 1). This 149 A S N Jackson, S A Reinsberg, S A Sohaib et al Table 2. Details of MRI sequences used Sequence TE (ms) TR (ms) Flip angle FOV (mm) Acquired matrix Reconstructed matrix Slice thickness/ gap (mm) T2W coronal scout T2W sagittal T2W axial small FOV DCE-MRI 80 132 100 Shortest Shortest 5080 Shortest 3.7вЂ“4.1 90 90 90 10 3756262.5 3706259 1706170 3006210 2566179 3206224 2566256 1766123 2566179 5126358 2566256 2566179 5/0 3/0.4 4/0 4/0 TR, repetition time; TE, echo time; FOV, field of view; T2W, T2 weghted; DCE, dynamic contrast-enhanced. apparatus ensured that the prostate was held in a known and stable orientation for slicing, and resulted in tissue slices of 4 mm thickness, i.e. the same slice thickness and orientation as the images derived from the DCE-MRI and axial T2W scans. The resulting slices were put into cassettes in a known orientation, i.e. cranial or caudal face of the slice downward in the cassette. One slice was selected by another investigator for subsequent RNA extraction [18], placed in a buffer solution and refrigerated. All of the slices of the main specimen were photographed using a Nikon D100 digital camera (Nikon UK Limited, Kingston upon Thames, UK). All slices except that used for RNA extraction were fixed in 10% neutral buffered formaldehyde solution for a minimum of 24 h before being processed using an automated processor (Tissue-TekH VIP 5 series, Sakura, Torrance CA) and embedded in paraffin wax. The wax blocks were sectioned with a rotary microtome to a thickness of 3 mm, mounted as whole sections on slides and stained with haemotoxylin and eosin. Image co-registration Stained slices were photographed, and each image (with the tumour indicated by a consultant histopathologist (C.F or C.C)) was matched with the image of its fresh unstained counterpart and, in turn, its corresponding MRI image. Assignment of MRI with the corresponding pathology was achieved by an overall вЂ�вЂ�best fitвЂ™вЂ™ of T2W images to pathology, based on morphology of the fresh unstained slices. T2W images with their superior anatomical detail were used for this initial co-registration. The stained slice was orientated using the inked margins for left and right orientation, and anatomical Figure 1. Prostate slicing using a multi-bladed knife and holder developed in-house. 150 features ВЎ the inked anterior surface for anteroposterior orientation. To compensate for shrinkage, which in other series has been estimated as between 10% and 33% [19вЂ“22], and distortion during processing of fresh tissue, image morphing software was used to apply transformations to the final stained slice in order to reshape it to its original corresponding MRI image (gtkMorph; GNU, general public licence). One control mesh was created between the image of the fresh slice and its T2W MRI counterpart, and another between the fresh slice and the stained section. Landmarks such as the capsule, urethra and ejaculatory ducts, but not features thought to represent tumour, were used as reference points for the control mesh. Using the approach illustrated in Figure 2, each image of a stained slice with the tumour outlined was morphed to the shape of the corresponding T2W MRI slice via the intermediate step of the image of the unfixed (fresh) corresponding prostate slice. The area of each histologically demonstrated individual tumour focus was also recorded, and the total tumour volume per patient was estimated by multiplying the sum of tumour areas per patient by the slice thickness (4 mm). From the slice coordinates of the T2W slice, the coincident DCE-MRI image was identified. An additional visual comparison between corresponding T2W and dynamic images was made to control for any patient and/or prostate motion taking place between the imaging sequences. Generation of pharmacokinetic parameter maps The previously calibrated DCE-MRI data were corrected for anteroposterior prostate motion occurring during the dynamic series acquisition using software developed in-house. The correction was then manually checked by sequentially scrolling through the sequence of images for each imaging slice. The motion-corrected dynamic sequences were then transferred to a separate workstation running in-house software (Magnetic Resonance Imaging Workbench (MRIW)) [23]. The software converts intensity changes observed in the images to changes in T1 values, and hence contrast agent concentration, using the method described by Parker et al [8] and subsequently generates pharmacokinetic parameter maps using a model described by Tofts et al [24]. In this model, the movement of contrast agent between tissue compartments is related to three physiologically based parameters: Ktrans (the volume transfer constant between blood plasma and extracellular extravascular space (EES)), ve (the EES fractional volume) and kep (the flux rate constant between the EES and plasma). A standardized arterial The British Journal of Radiology, February 2009 Dynamic contrast-enhanced MRI in prostate cancer Figure 2. The process of correcting for shrinkage and distortion of the specimen using gtkMorph software. input was used in the pharmacokinetic model (derived in a manner similar to that described by Walker-Samuel et al [25]) using data from arterial sampling reported by Fritz-Hansen et al [26]. Typical output parameter maps from the MRIW software are shown in Figure 3. Pixel-by-pixel analysis A pixel-by-pixel comparison between the pharmacokinetic maps and the corresponding histopathology sections was performed using software developed inhouse. The prostate and its peripheral and central zones were defined on the corresponding T2W images. These regions along with those defined as being tumour (morphed as described earlier) were then transferred automatically to the DCE-MRI image data. Mean pharmacokinetic enhancement parameters for the tumour, benign central gland and benign peripheral zone were calculated for each patient. The sensitivity and specificity on a pixel-by-pixel basis for varying thresholds of each parameter were calculated for each slice in order to generate receiver operating characteristic (ROC) curves for each patient and for each parameter, and also a вЂ�вЂ�summaryвЂ™вЂ™ ROC for all pixels for all patients. Analysis by qualitative radiological assessment Our second approach involved an experienced radiologist (S.A.S) identifying tumour based on an overall impression of the raw dynamic images viewed using the manufacturerвЂ™s analysis software (Easyvision; Philips). Specifically, focal areas of rapid enhancement and/or Table 3. Mean values of pharmacokinetic parameters for 16 patients analysed Benign peripheral zone Benign central gland Tumour kep (min21) Ktrans (min21) ve 0.502 (0.002) 0.513 (0.026) 0.685 0.219 (0.001) 0.264 (0.109) 0.328 0.318 (0.003) 0.382 (0.179) 0.412 Parenthesized values indicate p-values derived from Wilcoxon signed ranks tests between parameters for the benign gland and tumour. The British Journal of Radiology, February 2009 151 A S N Jackson, S A Reinsberg, S A Sohaib et al Figure 3. Dynamic contrast-enhanced MRI images of the pelvis with overlayed quantitative parameter maps for (a) Ktrans, (b) ve and (c) kep. (d) The contrast time vs concentration curve for an individual voxel. washout were interpreted as representing tumour. The prostate was first considered in the lower, middle and upper thirds based on the number of pathology slices obtained from the gland. Each axial section was then divided into 4 quadrants, resulting in 12 zones in total. The assignment between histopathology slices and imaging was the same as that used in the pixel-by-pixel analysis. Blinded to the results of pathology, the radiologist examined each slice and identified probable tumour from the qualitative parameter maps and from the raw dynamic images. Each region of interest was designated as being within one of the twelve zones. e.g. вЂ�вЂ�inferior, posterior, leftвЂ™вЂ™. The same way of describing the 152 location of tumour was used for the histopathology sections. On a different occasion, the radiologist used only the T2W images to identify the presence or absence of disease in a particular prostate zone, with focal areas of low signal intensity taken to represent tumour. To determine the effect of lesion size on the radiologistвЂ™s ability to identify malignant lesions, regions of interest were divided into quartiles according to area, and separately analysed. As there are data to suggest that lesions ,0.5 cm3 in size may not be prognostically significant [27, 28], further analysis was carried out after subdividing patients according to whether their total tumour volume was .0.5 cm3 or .1.0 cm3. The British Journal of Radiology, February 2009 Dynamic contrast-enhanced MRI in prostate cancer Statistical analysis Discussion Statistical analysis was carried out using SPSS (SPSS Inc, Chicago, IL). The Wilcoxon signed rank test was used to compare means of paired datasets, and the MannвЂ“Whitney U test was used for non-paired data. A pvalue of ,0.05 was taken to represent statistical significance. Results from the automated pixel-by-pixel analysis indicate a statistically significant difference in pharmacokinetic parameters between cancer and the benign peripheral zone, but the discriminatory value of the test, as judged by the area under the ROC curve, is only fair. There was no significant difference in the enhancement parameters between cancer and the benign central gland. The qualitative radiological assessment method of analysis has shown that DCE-MRI is superior to T2W scans in terms of the ability of a radiologist to correctly localize tumour to geographical regions of the prostate. We have not clearly demonstrated a relationship between tumour size and imaging accuracy, but sample sizes were small when patients and/or lesions were divided according to size. Other studies have examined the use of DCE-MRI in prostate cancer. Initial work tended to map retrospectively histological tumour to corresponding DCE-MRI imaging or parameter maps in order to explore which parameters appeared to be most discriminatory between tumour and benign gland. Some have used simple descriptive parameters relating to the time enhancement or timeвЂ“concentration curve, whereas others have applied pharmacokinetic models. Turnbull et al [14] correlated pharmacokinetic parameters with histopathology findings in 12 subjects and demonstrated that neoplastic tissue had a greater amplitude and higher contrast exchange rate than fibromuscular benign prostatic hypertrophy (BPH). Using a two-compartment pharmacokinetic model, Kiessling et al [13] found that prostate cancer has a higher amplitude of enhancement, kep and area under the timeвЂ“signal intensity curve than the benign gland. In the publication by Jager et al [9], a dynamic subtraction turbo-FLASH (fast low-angle shot) sequence with an endorectal coil was used in addition to standard MRI images of the prostate. The patient group comprised 57 men who were about to undergo radical prostatectomy for prostate cancer. A single slice thought to contain both tumour and benign tissue was chosen for the dynamic sequence from the T2W images. For correct tumour localization, the sensitivity was 57.5% vs 73.5% and specificity 80.5% vs 81% for T2W and turbo-FLASH sequences, respectively. For the detection of extracapsular spread on a quadrant by quadrant basis, sensitivity was 18.4% vs 50% and specificity 98.3% vs 99.2% for T2W and turbo-FLASH images, respectively. Engelbrecht et al [12] analysed 36 men with prostate cancer prior to radical retropubic prostatectomy and performed DCE-MRI with an endorectal coil, in addition to standard MRI sequences. ConcentrationвЂ“time curves Results Pixel-by-pixel analysis Table 3 shows the mean and the range of values for each of the three pharmacokinetic parameters for the 16 patients analysed in this way. Data are presented for the benign central gland, benign peripheral zone and tumour. A significant difference was seen between the benign peripheral zone and tumour for the values of Ktrans, ve, and kep. When comparing benign central gland to tumour, a significant difference was seen only for kep. Table 4 summarizes the area under the ROC curve for each parameter, i.e. the ability of a specific parameter to discriminate benign from malignant tissue. Data are presented for the whole gland and are restricted to the peripheral zone. Analysis by qualitative radiological assessment Out of a possible 12 zones, the average number of zones per patient thought to be suspicious for containing tumour was 2.9 (range, 1вЂ“6) on the basis of DCE-MRI alone, and 1.7 (range, 0вЂ“4) on the basis of T2W images alone. The average number of zones per patient confirmed by histology to contain tumour was 4.6 (range, 1вЂ“11). The median estimated tumour volume per patient was 0.676 cm3 for all 18 patients and 0.83 cm3 for the 16 patients included in the pixel-by-pixel analysis; the range for both groups was 0.01вЂ“4.76 cm3. Table 5 shows the accuracy of the technique as determined by the qualitative radiological assessment method. The sensitivity of DCE-MRI was significantly greater than that for T2W scans for all sizes of lesion, with a similar specificity. No significant sensitivity or specificity difference was seen between patients with a tumour volume greater or smaller than 0.5 cm3, or greater or smaller than 1 cm3 for DCE-MRI or T2W images. There was no significant difference in the sensitivity of DCE-MRI when all lesions were considered compared with the largest 25% of lesions only (50.5% vs 69.6%; p50.25) (data not shown). Table 4. Mean area under the receiver operating characteristic (ROC) curve for each parameter Ktrans kep Whole gland Area under summary ROC 0.61 curve (all patients) Range of areas under curve 0.33вЂ“0.74 for individual patients ve Peripheral zone Whole gland Peripheral zone Whole gland Peripheral zone 0.6 0.6 0.6 0.55 0.56 0.46вЂ“0.77 0.33вЂ“0.73 0.46вЂ“0.76 0.33вЂ“0.74 0.32вЂ“0.75 The British Journal of Radiology, February 2009 153 A S N Jackson, S A Reinsberg, S A Sohaib et al Table 5. Accuracy as assessed by qualitative radiological assessment for DCE-MRI and T2W images DCE MRI Sensitivity All patients (n518) 0.5 (0.006) 0.49 Total tumour volume .0.5 cm3 (n511) Total tumour volume .1.0 cm3 (n56) 0.63 T2W MRI Specificity Sensitivity Specificity 0.85 (0.593) 0.85 0.21 0.25 0.89 0.90 0.93 0.34 1.0 Parenthesized values indicate p-values derived from Wilcoxon signed ranks tests comparing DCE-MRI and T2W images. DCE, dynamic contrast-enhanced; T2W, T2 weighted. were calculated and a four-parameter model applied, with onset time, time to peak, peak enhancement and washout being chosen as the parameters of interest. The optimal parameter for the discrimination of tumour within the peripheral zone or central gland was found to be relative peak enhancement. In this study, the exclusion of small lesions and the characterization of each lesion by selected pixels, demonstrating the maximal value of the parameter of interest, could lead to bias and an overestimation of the association between that parameter and the presence of tumour. Rouviere et al [11] imaged 39 patients with a DCE-MRI T1 weighted FLASH sequence prior to radical prostatectomy. A single slice thought to contain tumour was chosen for the dynamic images. Area under the ROC curve varied from 0.602вЂ“0.698 for the identification of tumour within adenoma, and 0.614вЂ“0.827 for the depiction of tumour within the peripheral zone. Preziosi et al [10] studied 11 patients who were scanned with conventional sequences and a dynamic contrast sequence. A single slice containing a focal area of enhancement was selected for analysis. All 13 of the enhancing regions of interest corresponded to tumour, as seen on histopathology, and a further 4 tumour foci were identified that had not been identified on imaging. This resulted in a sensitivity of 13/17 (76%) and a positive predictive value of 100% for DCE-MRI. In a similar way to the qualitative radiological assessment in our methodology, the values for sensitivity and positive predictive value given here relate to correct identification of a tumour focus, but not necessarily to correct assessment of the size of a given lesion. The most comparable studies to our qualitative radiological assessment method are those by Girouin et al [17] and Futterer et al [16]. In 46 patients and using a similar method to ours, Girouin et al [17] divided the prostate into 20 geographical zones, also distinguishing the central gland from the peripheral zone and seminal vesicles. Three observers interpreted DCE-MRI images by visual assessment of the enhancement pattern. In general terms, malignant lesions were those considered to enhance earlier than the surrounding gland. In similar results to ours, a sensitivity of 46вЂ“60% and specificity of 91вЂ“94% was found for DCE-MRI images when compared with 17.6вЂ“24% sensitivity and 94вЂ“98% specificity for T2W images. Futterer et al [16] divided the prostate into 14 zones, and two radiologists scored T2W images and DCE-MRI parameter maps for Ktrans, ve and kep on a five-point scale. DCE-MRI parameter maps scored 69вЂ“95% sensitivity and 80вЂ“96% specificity compared with 52вЂ“67% sensitivity and 73вЂ“74% specificity for T2W images. The parameter ve resulted in the greatest accuracy for both readers. 154 Our study, as well as others, has shown that prostate tumour is associated with a higher ve, i.e. the fractional EES, than surrounding gland. Diffusion-weighted MRI, however, has shown prostate tumours to have a lower apparent diffusion coefficient than the surrounding gland [29вЂ“32], suggesting a more compact cellular tissue with less EES. The reason for this apparent contradiction is not clear, but it is possible that the pharmacokinetic model used in DCE-MRI, when assuming that the intravascular volume and contrast within it are negligible, fails in prostate cancer which may have a nonnegligible vascular volume component. In keeping with other studies, we have demonstrated that there are differences in the enhancement properties of prostate cancer when compared with the benign peripheral gland. This study does, however, have several limitations. The вЂ�вЂ�pixel-by-pixelвЂ™вЂ™ analysis is reliant on specialist software and is therefore not widely available. Pixels are considered independently, and patterns within groups of pixels are not taken into account. In reality, a group of adjacent pixels of a parameter above a given threshold might be considered more suspicious for malignancy than single pixels, a property not considered by this method of analysis, and which may reduce the specificity of this method of analysis. Additionally, this point-by point technique is highly susceptible to small differences in orientation of the prostate between imaging and sectioning. The effect of any such misalignment would be particularly marked for small tumours. For radiologist-reviewed images, errors in orientation and subsequent slice assignment would produce a less marked effect than in pixel-by-pixel analysis, as larger regions of the prostate were considered. A criticism of this method of analysis is that it represents the ability of one individual radiologist to identify tumour. In addition, the numerical description of accuracy described refers only to the ability of the imaging technique to correctly identify tumour in a particular region of the prostate, as defined in the study. It takes no account of the ability to correctly identify the size of lesions. Not all slices of the prostate were fully analysed for six patients because the slice taken for RNA extraction was not subject to routine histopathological processing. However, even if the least favourable scenario for DCEMRI and the best for T2W images were assumed вЂ” namely that tumour was present in all quadrants of the non-analysed slice and that DCE-MRI images failed to demonstrate tumour but T2W images did вЂ” the calculated sensitivity of DCE-MRI still remained superior to that of T2W images (0.48 vs 0.23, respectively; p50.02). In this study, we took meticulous care to acquire images at an angulation and slice thickness that would The British Journal of Radiology, February 2009 Dynamic contrast-enhanced MRI in prostate cancer allow a direct comparison with histology. Indeed, the method of sectioning the unfixed prostate has led to the granting of a UK patent. In addition, we used novel software to correct for size and shape distortions of pathology specimens to allow as accurate a correspondence of pixels as possible for the quantitative analysis. This study was carried out to determine whether DCEMRI would prove a useful tool for radiotherapy target localization. Ideally, such a technique would have sensitivity and specificity values of 100% in terms of correct spatial and volumetric estimation of tumour. The consequences in terms of improved tumour control vs normal tissue toxicity for a hypothetical prostate lesion localized with DCE-MRI and radiation dose escalated would depend upon the balance between sensitivity and specificity, and also on the anatomical location of the lesion within the prostate. Dose-escalating true- or falsepositive lesions would have different normal tissue implications for, say, the rectum or urethra depending upon whether they were located in the peripheral zone or central gland. In this study, the diagnostic accuracy of dynamic images by a radiologistвЂ™s impression are arguably sufficient to make a case for boosting вЂ�вЂ�regionsвЂ™вЂ™ of the prostate, as defined here. With these levels of sensitivity and specificity, approximately half of involved regions would be detected, and less than 10% of regions would be boosted needlessly. These approximations may under- rather than over-estimate the potential value of DCE-MRI, as patients treated with external beam radiotherapy will, in general, have larger, more detectable cancers than those studied here, who were treated with prostatectomy. Further work will model tumour control probability and the normal tissue complication probability based on dose-escalating tumour nodules localized by the qualitative radiological assessment method of analysis. Other functional MRI techniques such as diffusion-weighted MRI [31вЂ“35] and MR spectroscopic imaging [36вЂ“38] show promise in prostate cancer imaging, and future work should also seek to assess the value of these techniques as tools for target localization in prostate cancer. Conclusions We have analysed the ability of DCE-MRI to correctly localize prostate tumour in two ways. We have shown that, for one radiologist using a commercial software package, DCE-MRI results in superior tumour localization compared with T2W scans. We have demonstrated that tumour appears to have different enhancement properties to benign peripheral zone and that, according to ROC analysis, this results in a diagnostic test which would be considered of fair discriminatory value by conventional criteria. Further work will establish whether a clinical benefit would be derived by using this method of tumour localization for prostate cancer radiotherapy dose escalation. Acknowledgments This work was undertaken in The Royal Marsden NHS Foundation Trust, which received a proportion of its The British Journal of Radiology, February 2009 funding from the NHS Executive; the views expressed in this publication are those of the authors and not necessarily those of the NHS Executive. This work was supported by The Department of Health New and Emerging Applications of Technology (NEAT) program grant number B132, the Institute of Cancer Research, the Bob Champion Cancer Trust and Cancer Research UK Section of Radiotherapy [CUK] grant number C46/ A2131. References 1. Dearnaley DP, Sydes MR, Graham JD, Aird EG, Bottomley D, Cowan RA, et al. Escalated-dose versus standard-dose conformal radiotherapy in prostate cancer: first results from the MRC RT01 randomised controlled trial. Lancet Oncol 2007;8:475вЂ“87. 2. Peeters ST, Heemsbergen WD, Koper PC, van Putten WL, Slot A, Dielwart MF, et al. Dose-response in radiotherapy for localized prostate cancer: results of the Dutch multicenter randomized phase III trial comparing 68 Gy of radiotherapy with 78 Gy. J Clin Oncol 2006;24:1990вЂ“6. 3. Pollack A, Zagars GK, Starkschall G, Antolak JA, Lee JJ, Huang E, et al. Prostate cancer radiation dose response: results of the M. D. Anderson phase III randomized trial. Int J Radiat Oncol Biol Phys 2002;53:1097вЂ“105. 4. Shipley WU, Verhey LJ, Munzenrider JE, Suit HD, Urie MM, McManus PL, et al. Advanced prostate cancer: the results of a randomized comparative trial of high dose irradiation boosting with conformal protons compared with conventional dose irradiation using photons alone. Int J Radiat Oncol Biol Phys 1995;32:3вЂ“12. 5. Zietman AL, DeSilvio ML, Slater JD, Rossi CJ Jr, Miller DW, Adams JA, et al. Comparison of conventional-dose vs highdose conformal radiation therapy in clinically localized adenocarcinoma of the prostate: a randomized controlled trial. JAMA 2005;294:1233вЂ“9. 6. Nutting CM, Corbishley CM, Sanchez-Nieto B, Cosgrove VP, Webb S, Dearnaley DP. Potential improvements in the therapeutic ratio of prostate cancer irradiation: dose escalation of pathologically identified tumour nodules using intensity modulated radiotherapy. Br J Radiol 2002;75:151вЂ“61. 7. Pickett B, Vigneault E, Kurhanewicz J, Verhey L, Roach M. Static field intensity modulation to treat a dominant intraprostatic lesion to 90 Gy compared to seven field 3dimensional radiotherapy. Int J Radiat Oncol Biol Phys 1999;44:921вЂ“9. 8. Parker GJ, Suckling J, Tanner SF, Padhani AR, Revell PB, Husband JE, et al. Probing tumor microvascularity by measurement, analysis and display of contrast agent uptake kinetics. J Magn Reson Imaging 1997;7:564вЂ“74. 9. Jager GJ, Ruijter ET, van de Kaa CA, de la Rosette JJ, Oosterhof GO, Thornbury JR, et al. Dynamic TurboFLASH subtraction technique for contrast-enhanced MR imaging of the prostate: correlation with histopathologic results. Radiology 1997;203:645вЂ“52. 10. Preziosi P, Orlacchio A, Di Giambattista G, Di Renzi P, Bortolotti L, Fabiano A, et al. Enhancement patterns of prostate cancer in dynamic MRI. Eur Radiol 2003;13:925вЂ“30. 11. Rouviere O, Raudrant A, Ecochard R, Colin-Pangaud C, Pasquiou C, Bouvier R, et al. Characterization of timeenhancement curves of benign and malignant prostate tissue at dynamic MR imaging. Eur Radiol 2003;13:931вЂ“42. 12. Engelbrecht MR, Huisman HJ, Laheij RJ, Jager GJ, van Leenders GJ, Hulsbergen-Van de Kaa CA, et al. Discrimination of prostate cancer from normal peripheral zone and central gland tissue by using dynamic contrastenhanced MR imaging. Radiology 2003;229:248вЂ“54. 155 A S N Jackson, S A Reinsberg, S A Sohaib et al 13. Kiessling F, Lichy M, Grobholz R, Heilmann M, Farhan N, Michel MS, et al. Simple models improve the discrimination of prostate cancers from the peripheral gland by T1weighted dynamic MRI. Eur Radiol 2004;14:1793вЂ“801. 14. Turnbull LW, Buckley DL, Turnbull LS, Liney GP, Knowles AJ. Differentiation of prostatic carcinoma and benign prostatic hyperplasia: correlation between dynamic GdDTPA-enhanced MR imaging and histopathology. J Magn Reson Imaging 1999;9:311вЂ“6. 15. van Dorsten FA, van der Graaf M, Engelbrecht MR, van Leenders GJ, Verhofstad A, Rijpkema M, et al. Combined quantitative dynamic contrast-enhanced MR imaging and (1)H MR spectroscopic imaging of human prostate cancer. J Magn Reson Imaging 2004;20:279вЂ“87. 16. Futterer JJ, Heijmink SW, Scheenen TW, Veltman J, Huisman HJ, Vos P, et al. Prostate cancer localization with dynamic contrast-enhanced MR imaging and proton MR spectroscopic imaging. Radiology 2006;241:449вЂ“58. 17. Girouin N, Mege-Lechevallier F, Tonina Senes A, Bissery A, Rabilloud M, Marechal JM, et al. Prostate dynamic contrastenhanced MRI with simple visual diagnostic criteria: is it reasonable? Eur Radiol 2007;17:1498вЂ“509. 18. Jhavar SG, Fisher C, Jackson A, Reinsberg SA, Dennis N, Falconer A, et al. Processing of radical prostatectomy specimens for correlation of data from histopathological, molecular biological, and radiological studies: a new whole organ technique. J Clin Pathol 2005;58:504вЂ“8. 19. Jager GJ, Ruijter ET, van de Kaa CA, de la Rosette JJ, Oosterhof GO, Thornbury JR, et al. Local staging of prostate cancer with endorectal MR imaging: correlation with histopathology. AJR Am J Roentgenol 1996;166:845вЂ“52. 20. Schiebler ML, Tomaszewski JE, Bezzi M, Pollack HM, Kressel HY, Cohen EK, et al. Prostatic carcinoma and benign prostatic hyperplasia: correlation of high-resolution MR and histopathologic findings. Radiology 1989;172:131вЂ“7. 21. Sommer FG, Nghiem HV, Herfkens R, McNeal J, Low RN. Determining the volume of prostatic carcinoma: value of MR imaging with an external-array coil. AJR Am J Roentgenol 1993;161:81вЂ“6. 22. Lencioni R, Menchi I, Paolicchi A, Carini M, Amorosi A, Bartolozzi C. Prediction of pathological tumor volume in clinically localized prostate cancer: value of endorectal coil magnetic resonance imaging. Magma 1997;5:117вЂ“21. 23. dвЂ™Arcy JA, Collins DJ, Padhani AR, Walker-Samuel S, Suckling J, Leach MO. Informatics in Radiology (infoRAD): Magnetic Resonance Imaging Workbench: analysis and visualization of dynamic contrast-enhanced MR imaging data. Radiographics 2006;26:621вЂ“32. 24. Tofts PS, Brix G, Buckley DL, Evelhoch JL, Henderson E, Knopp MV, et al. Estimating kinetic parameters from dynamic contrast-enhanced T(1)-weighted MRI of a diffusable tracer: standardized quantities and symbols. J Magn Reson Imaging 1999;10:223вЂ“32. 156 25. Walker-Samuel S, Leach MO, Collins DJ. Evaluation of response to treatment using DCE-MRI: the relationship between initial area under the gadolinium curve (IAUGC) and quantitative pharmacokinetic analysis. Phys Med Biol 2006;51:3593вЂ“602. 26. Fritz-Hansen T, Rostrup E, Larsson HB, Sondergaard L, Ring P, Henriksen O. Measurement of the arterial concentration of Gd-DTPA using MRI: a step toward quantitative perfusion imaging. Magn Reson Med 1996;36:225вЂ“31. 27. McNeal JE, Bostwick DG, Kindrachuk RA, Redwine EA, Freiha FS, Stamey TA. Patterns of progression in prostate cancer. Lancet 1986;1:60вЂ“3. 28. Stamey TA, Yang N, Hay AR, McNeal JE, Freiha FS, Redwine E. Prostate-specific antigen as a serum marker for adenocarcinoma of the prostate. N Engl J Med 1987;317:909вЂ“16. 29. Choi YJ, Kim JK, Kim N, Kim KW, Choi EK, Cho KS. Functional MR imaging of prostate cancer. Radiographics 2007;27:63вЂ“75; discussion 75вЂ“7. 30. Reinsberg SA, Payne GS, Riches SF, Ashley S, Brewster JM, Morgan VA, et al. Combined use of diffusion-weighted MRI and 1H MR spectroscopy to increase accuracy in prostate cancer detection. AJR Am J Roentgenol 2007;188:91вЂ“8. 31. Hosseinzadeh K, Schwarz SD. Endorectal diffusionweighted imaging in prostate cancer to differentiate malignant and benign peripheral zone tissue. J Magn Reson Imaging 2004;20:654вЂ“61. 32. Sato C, Naganawa S, Nakamura T, Kumada H, Miura S, Takizawa O, et al. Differentiation of noncancerous tissue and cancer lesions by apparent diffusion coefficient values in transition and peripheral zones of the prostate. J Magn Reson Imaging 2005;21:258вЂ“62. 33. Charles-Edwards EM, deSouza NM. Diffusion-weighted magnetic resonance imaging and its application to cancer. Cancer Imaging 2006;6:135вЂ“43. 34. Shimofusa R, Fujimoto H, Akamata H, Motoori K, Yamamoto S, Ueda T, et al. Diffusion-weighted imaging of prostate cancer. J Comput Assist Tomogr 2005;29:149вЂ“53. 35. Haider MA, van der Kwast TH, Tanguay J, Evans AJ, Hashmi AT, Lockwood G, et al. Combined T2-weighted and diffusion-weighted MRI for localization of prostate cancer. AJR Am J Roentgenol 2007;189:323вЂ“8. 36. Kurhanewicz J, Swanson MG, Nelson SJ, Vigneron DB. Combined magnetic resonance imaging and spectroscopic imaging approach to molecular imaging of prostate cancer. J Magn Reson Imaging 2002;16:451вЂ“63. 37. Casciani E, Polettini E, Bertini L, Masselli G, Emiliozzi P, Amini M, et al. Contribution of the MR spectroscopic imaging in the diagnosis of prostate cancer in the peripheral zone. Abdom Imaging 2007, in press. 38. Squillaci E, Manenti G, Mancino S, Carlani M, Di Roma M, Colangelo V, et al. MR spectroscopy of prostate cancer. Initial clinical experience. J Exp Clin Cancer Res 2005;24:523вЂ“30. The British Journal of Radiology, February 2009