Quantitative 3T multiparametric MRI of benign and malignant prostatic tissue in patients with and without local recurrent prostate cancer after external‐beam radiation therapy

Abstract

Background

Post‐radiotherapy locally recurrent prostate cancer (PCa) patients are candidates for focal salvage treatment. Multiparametric MRI (mp‐MRI) is attractive for tumor localization. However, radiotherapy‐induced tissue changes complicate image interpretation. To develop focal salvage strategies, accurate tumor localization and distinction from benign tissue is necessary.

Purpose

To quantitatively characterize radio‐recurrent tumor and benign radiation‐induced changes using mp‐MRI, and investigate which sequences optimize the distinction between tumor and benign surroundings.

Study Type

Prospective case–control.

Subjects

Thirty‐three patients with biochemical failure after external‐beam radiotherapy (cases), 35 patients without post‐radiotherapy recurrent disease (controls), and 13 patients with primary PCa (untreated).

Field Strength/Sequences

3T; quantitative mp‐MRI: T₂‐mapping, ADC, and K^trans and k_ep maps.

Assessment

Quantitative image‐analysis of prostatic regions, within and between cases, controls, and untreated patients.

Statistical Tests

Within‐groups: nonparametric Friedman analysis of variance with post‐hoc Wilcoxon signed‐rank tests; between‐groups: Mann–Whitney tests. All with Bonferroni corrections. Generalized linear mixed modeling to ascertain the contribution of each map and location to tumor likelihood.

Results

Benign imaging values were comparable between cases and controls (P = 0.15 for ADC in the central gland up to 0.91 for k_ep in the peripheral zone), both with similarly high peri‐urethral K^trans and k_ep values (min⁻¹) (median [range]: K^trans = 0.22 [0.14–0.43] and 0.22 [0.14–0.36], P = 0.60, k_ep = 0.43 [0.24–0.57] and 0.48 [0.32–0.67], P = 0.05). After radiotherapy, benign central gland values were significantly decreased for all maps (P ≤ 0.001) as well as T₂, K^trans, and k_ep of benign peripheral zone (all with P ≤ 0.002). All imaging maps distinguished recurrent tumor from benign peripheral zone, but only ADC, K^trans, and k_ep were able to distinguish it from benign central gland. Recurrent tumor and peri‐urethral K^trans values were not significantly different (P = 0.81), but k_ep values were (P < 0.001). Combining all quantitative maps and voxel location resulted in an optimal distinction between tumor and benign voxels.

Data Conclusion

Mp‐MRI can distinguish recurrent tumor from benign tissue.

Level of Evidence: 2

Technical Efficacy Stage: 2

J. Magn. Reson. Imaging 2019;50:269–278.

ONE OF THE MAIN treatment options for localized prostate cancer (PCa) is external‐beam radiotherapy (EBRT). Depending on the risk group, 5‐year disease‐free survival varies from 67–80% for PCa patients who are treated with whole‐gland radiotherapy (RT) to 78 Gy.1 A small but significant proportion of patients failing the primary treatment will harbor locally recurrent disease only,2, 3 for which focal treatment strategies, targeting the tumor region while sparing the surrounding uncompromised tissue, might offer a curative treatment option. To this end, multiparametric magnetic resonance imaging (mp‐MRI) with both anatomical and functional properties is attractive for locoregional evaluation of recurrent PCa and for tumor boundary definition. However, RT‐induced tissue changes pose a challenge to MRI interpretation. T₂‐weighted (T₂w) MRI is the reference anatomical sequence; however, it has limited performance in this setting, as tumor conspicuity is decreased and zonal anatomy is often lost.4 The use of functional sequences such as 1.5T MR spectroscopy5 as well as diffusion‐weighted imaging (DWI) and dynamic contrast‐enhanced (DCE)‐MRI6, 7, 8 were found to surpass T₂w‐MRI in the detection of recurrent disease.

To successfully develop focal salvage strategies for recurrent PCa, an accurate detection and localization of the tumor as well as distinction from radiation‐affected benign tissue is necessary.

The objectives of this study were to: 1) investigate RT‐induced tissue changes and how these impacted the distinction of recurrent tumor from surrounding benign tissue; and 2) use quantitative sequences to characterize both benign tissue and recurrent PCa after RT, and investigate which sequences resulted in optimal tumor localization.

Materials and Methods

Patient Recruitment and Matching

This prospective single‐institution study was approved by the Institutional Review Board and took place between December 2015 and July 2018. All patients signed informed consent prior to participation in the study.

An overview of the study design can be seen in Fig. 1.

Figure 1.

Study design with an overview of the three analyzed cohorts. BCR, biochemical recurrence; bx, biopsy; EBRT, external‐beam radiation therapy.

CASES (PATIENTS WITH RECURRENT PCA)

Patients with a biochemical recurrence according to the Phoenix definition9 were enrolled in the study. Patients could only be included more than 24 months after completion of EBRT and if they qualified for salvage treatment. A ⁶⁸Ga‐PSMA‐PET (prostate‐specific membrane antigen / positron emission tomography) scan was acquired to exclude metastatic disease. Exclusion criteria were hormonal therapy in the past year, the use of anticoagulants that could not be stopped temporarily to acquire biopsies, contraindications for an MRI exam, and other treatments for cancer in the pelvis. Patients received an mp‐MRI. When a tumor‐suspected region was identified, an MR‐fused ultrasound‐guided biopsy was performed targeting this region. If no tumor‐suspected region could be identified, either only prostate‐specific antigen (PSA) follow‐up would be continued, or a standard systematic biopsy could be performed. This decision was made by the treating physician. Patients with positive imaging and biopsy findings for locally recurrent PCa constituted the cohort of case patients.

An inconclusive diagnosis was given to patients for whom a local or metastatic recurrence could not be detected on MRI, or the biopsies were negative, or both. These patients were analyzed further but separately from the cohort of cases.

CONTROLS (PATIENTS WITHOUT RECURRENT PCA)

Matched controls (patients without evidence of recurrent disease) were recruited and received an mp‐MRI scan. The matching criteria were: time since primary RT (years), use of hormonal therapy, and the risk group of the primary tumor. Risk stratification was based on the definition proposed by the European Association for Medical Oncology (ESMO) in 2010,10 incorporating changes suggested in the review by Rodrigues et al11 on the importance of the amount of high‐grade cancer, differentiating between Gleason 3 + 4 and 4 + 3. Risk stratification was defined as: low‐risk: T1‐T2a, Gleason ≤6 and PSA ≤10 ng/ml; intermediate‐risk: T2b‐T2c, Gleason = 7 (3 + 4) and PSA ≤20 ng/ml, not otherwise low‐or high‐risk; high‐risk: T3‐T4 or Gleason ≥7 (4 + 3) or PSA >20. The exclusion criteria were identical to those applied for the cases.

If a patient developed a biochemical recurrence during the study period, he was moved to the inconclusive cohort.

UNTREATED (PATIENTS AT PRIMARY DIAGNOSIS)

A previously reported cohort of 13 primary PCa patients,12, 13 scanned prior to treatment with an mp‐MRI protocol, was used as reference. As part of the study, these patients underwent two mp‐MRI scanning sessions. Following the second MRI examination, the patients were treated with a radical prostatectomy. This cohort is referred to as the untreated cohort.

MRI Protocol

All patients were scanned with a 3T Achieva dStream (A) or Ingenia (B) scanner (Philips Healthcare, Best, The Netherlands). Twenty patients were scanned in system A with a 6‐channel cardiac coil and 48 were scanned in system B with a 16‐channel anterior and 12‐channel posterior coil. With the exception of four patients (three scanned in system A), all were scanned with the use of an endorectal coil.

The mp‐MRI protocol consisted of an axial, sagittal, and coronal T₂w turbo spin echo; a 3D gradient echo T₁‐weighted (T₁w); a balanced steady‐state free‐precession (bSSFP); an axial multiecho spin echo k‐t‐T₂ sequence,12 with 12 echoes acquired at a spacing of 16 msec, starting at 32 msec; a transversal DWI and a DCE sequence. DWI was acquired using a single‐shot spin‐echo echo‐planar imaging sequence with diffusion encoding b‐values ranging from 0 to 800 s/mm². DCE was acquired with a 3D T₁w spoiled gradient echo sequence at a temporal resolution of 2.6 seconds over 5 minutes, before, during, and after intravenous administration of 15 ml of Dotarem (0.5 mM Dotarem, Guerbet, Roissy, France). The contrast agent was administered using a power injector (MedRad, Warrendale, PA), followed by a 30‐ml saline flush at a flow rate of 3 ml/s. Further sequence specific details can be found in the Supplementary Materials 1.

Apparent diffusion coefficient (ADC) maps were derived from the DWI sequence using b = 200 and 800 s/mm². The standard Tofts model14 was used to calculate the pharmacokinetic maps K^trans and k_ep. This was done using a T₁ map generated based on the variable flip angle method,15 using flip angle = 3°, 6°, 10°, 20°, and 30°, to convert signal intensity into concentration, and an arterial input function (AIF) with parameters derived from an in‐house study population. T₂ maps were created by converting the signal to the logarithmic scale and performing a weighted linear fit.

All sequence‐derived functional maps were generated using MatLab R2017b (MathWorks, Natick, MA).

Visual inspection was used to assess possible displacements between the functional sequences (and respective quantitative maps) and the T₂w‐MRI scans. Rigid registration based on mutual information was used whenever necessary to correct for these displacements. All images were resampled to the T₂w grid.

Region of Interest (ROI) Segmentation

The prostate, the peripheral zone (PZ), the central gland (CG), and the region surrounding the urethra—peri‐urethral tissue (PU)—were delineated. The PU region was delineated using both the sagittal and transversal T₂w images. The tumor‐suspected regions were delineated by a radiologist (14 years of experience) based on the MRI, PET, and biopsy report information. As the Prostate Imaging Reporting and Data System 2 (PI‐RADS v2)16 is not applicable to recurrent prostate cancer, tumor was defined as a region with low signal intensity (SI) on T₂w‐MRI, high SI in the b = 800 DWI scan, low SI on the ADC map, and increased enhancement in the K^trans and k_ep maps. For the untreated cohort, tumor delineations were originally performed on the histopathology and propagated to imaging. Figure 2 illustrates the ROIs as delineated for a case patient.

Figure 2.

Illustration of the delineated ROIs. In the left image the eroded prostate contour, the peripheral zone, and the tumor can be seen delineated on the transversal T₂w image. On the right, the peri‐urethral contour is presented overlaid on the sagittal T₂w scan.

The bSSFP sequence was used to identify and exclude from further analysis the implanted fiducial markers. Prostate delineations were eroded in all directions by 1 mm to ensure that only prostate tissue was analyzed.

ROI analysis was performed using Python 3 (Python Software Foundation, Delaware).

Statistics

A t‐test for continuous variables (PSA) and a chi‐square test for categorical variables (Gleason and TNM stage) were used to compare the clinical characteristics of the cohorts.

Normality was checked using the Shapiro–Wilk test. In a case of nonnormality, a nonparametric Friedman's analysis of variance (ANOVA) for dependent groups was used to compare the different regions within the groups (untreated, cases, and controls). If significant, the differences were further evaluated using a post‐hoc Wilcoxon signed‐rank test. To compare the same region between groups, a Mann–Whitney test was used. To correct for multiple testing, a Bonferroni correction was applied to the significance level of α = 0.05.

Univariate and multivariate generalized linear mixed‐effect modeling (GLMM) was applied to assess the predictive value of imaging on the voxelwise likelihood of tumor. Only case patients were used in this analysis. Voxels were grouped into benign (resulting from a combination of unaffected PZ and CG) and tumor (based on the radiologist delineations). To obtain the likelihood of tumor on a voxel level, fixed and random effects were included. The quantitative imaging maps and voxel anatomical location (PZ or CG) were accounted as predictive variables and therefore included as fixed effects. Random effects accounted for spatial clustering by incorporating voxel location within the prostate (the relative distance in x, y, and z from the prostate center of mass) and patient identifiers. When the association between model parameters and tumor probability was nonlinear, the parameters were grouped in quartiles and regression coefficients were estimated for each group considering the first quartile as reference. The model fit was assessed using the Bayesian Information Criterion (BIC) (a decrease of 10 points reflects an improved fit) and by evaluating the residual random error. Statistics were performed with the Statistical Package for Social Sciences, v. 22.0 (SPSS, Chicago, IL). Analysis was performed in R17 using the lme4 package.18

Results

Patient Characteristics

In total, 33 case patients and 35 controls were prospectively included in the study. Adjuvant hormonal therapy was given for a maximum period of 3 years to 82% of the cases and to 94% of the controls upon treatment of the primary PCa. However, none of the patients received hormonal therapy within 1 year prior to the MRI exam.

CASES (PATIENTS WITH RECURRENT PCA)

Ten of the 33 included patients had either regional metastases without evidence of intraprostatic recurrence (n = 5) or an inconclusive diagnosis (n = 5). Twenty‐three patients remained that fitted all the inclusion criteria for cases. For 3 of the 23 case patients, the first biopsy was negative and a repeated navigated biopsy confirmed the MRI suspected tumor location.

In 2 out of 23 patients the k‐t‐T₂ sequence was not acquired and for another the ADC map was not used due to poor image quality.

For the 23 cases a total of 27 suspected tumor foci were delineated, with an average tumor volume of 1.37 cm³. From these, four were located in CG. The median time to recurrence for the 23 cases was 7 years since primary EBRT treatment.

CONTROLS (PATIENTS WITHOUT RECURRENT PCA)

One patient was initially recruited as a control but had a biochemical failure during the study period. This patient was moved to the inconclusive cohort. The remaining 34 matched controls (patients without evidence of locally recurrent PCa) had a median time of 6 years since the EBRT treatment.

UNTREATED (PATIENTS AT PRIMARY DIAGNOSIS)

The 13 untreated patients had a median time of 20 (range, 5–65) days between the MRI and prostatectomy. None of these patients received hormonal therapy before imaging. For one patient, the small dimensions of the histopathological tumor resulted in a subvoxel tumor ROI in MRI. For this patient only benign tissue was analyzed.

Table 1 describes the patient characteristics for cases, controls, and untreated patients. For the untreated patients, the characteristics reported in the table are based on the biopsy report and clinical TNM stage.

Table 1.

Patient Characteristics

	Cases		Controls	Untreated
	Primary tumor	Recurrence	Primary tumor	Primary tumor
TNM stagea
T1	—	—	5	5
T2	11	13	6	7
T3	11	4	23	1
T4	1	—	—	—
Gleason scoreb
≤ 6	4	—	6	7
7	12	—	12	6
≥ 8	6	—	16	—
PSA (ng/ml) c
≤ 10	10	22	11	12
10 < PSA ≤ 20	6	1	14	1
≥ 20	6	—	8	—

Recurrent tumors were not assigned a Gleason score as radiation induced atypia can be a confounder for pathological interpretation.

^aIn six case patients TNM stage of recurrent tumor was not reported.

^bFor one case patient only the Anderson score (=2) was available.

^cIn one case and one control PSA of primary tumor could not be retrieved.

The clinical characteristics of the cohorts were compared. TNM stage was grouped in two (T1+T2 and T3+T4) and the Gleason score in three groups (Gleason ≤6, 7, ≥8). Cases and controls had comparable clinical characteristics (PSA: P = 0.789; TNM stage: P = 0.393; Gleason score: P = 0.271). Both case and control patients had significantly higher‐risk primary PCa when compared with the untreated patients (PSA: P = 0.008 and P = 0.003; TNM stage: P = 0.004 and P < 0.001; Gleason score: P = 0.03 and P = 0.004 for cases and controls, respectively).

Figure 3 illustrates the values of T₂, ADC, K^trans, and k_ep for untreated, controls, and case patients. Median imaging values for the three cohorts are presented in Table 2.

Figure 3.

Boxplots with median values for all regions, imaging modalities, and the three cohorts of patients: untreated, controls (nonrecurrent), and cases (recurrent). Edges of the boxes are the first (25^th) and third (75^th) quartiles, and whiskers represent the 5^th and 95^th percentile. PZ, peripheral zone; CG, central gland; PU, peri‐urethral tissue.

Table 2.

Quantitative Imaging Values for the Three Cohorts in the Different Prostatic Regions. Median of all individual patient's median values, with the minimum and maximum presented in parentheses

		PZ	CG	Tumor	PU
T2 (msec)	Untreated	129 (67–258)	97 (81–114)	85 (70–128)	94 (84–135)
	Controls	93 (76–129)	79 (66–99)	—	92 (77–121)
	Cases	91 (80–114)	79 (70–94)	81 (66–97)	91 (77–103)
ADC (x10−3 mm2/sec)	Untreated	1.40 (0.70–1.90)	1.20 (1.10–1.50)	1.00 (0.70–1.60)	1.32 (1.20–1.80)
	Controls	1.32 (0.06–1.57)	1.20 (0.07–1.44)	—	1.41 (0.07–1.70)
	Cases	1.28 (0.45–1.51)	1.14 (0.52–1.32)	0.95 (0.41–1.29)	1.40 (0.67–1.62)
Ktrans(min−1)	Untreated	0.16 (0.08–0.41)	0.29 (0.14–0.42)	0.29 (0.17–0.70)	0.28 (0.13–0.51)
	Controls	0.11 (0.06–0.17)	0.15 (0.10–0.26)	—	0.22 (0.14–0.36)
	Cases	0.11 (0.05–0.16)	0.14 (0.08–0.24)	0.22 (0.09–0.40)	0.22 (0.14–0.43)
kep(min−1)	Untreated	0.63 (0.32–1.20)	0.85 (0.31–1.18)	0.73 (0.39–3.89)	0.50 (0.26–0.96)
	Controls	0.33 (0.17–0.49)	0.37 (0.26–0.58)	—	0.48 (0.32–0.67)
	Cases	0.32 (0.19–0.48)	0.36 (0.24–0.49)	0.58 (0.40–1.36)	0.43 (0.24–0.57)

Characteristics of Irradiated Benign Tissue

Figure 4 illustrates the quantitative maps and suspected tumor regions for a representative case (Fig. 4a) and control (Fig. 4b) patient.

Figure 4.

Representative example of a case (a) patient with recurrence and a control (b) without recurrent disease, with all anatomical and functional maps. Thin and thick white lines represent prostate and tumor delineations, respectively. Black and white arrows in the K^trans maps point to the increased enhancement in the peri‐urethral tissue, present in both cases and controls.

The values for benign PZ and CG were comparable between cases and controls (P = 0.15 for ADC in the CG up to 0.91 for the k_ep in the PZ). Values in the PU were also similar between the groups (with the lowest P = 0.05 for k_ep >0.05/3). Both cases and controls presented with PU enhancement.

For controls, significant differences were found between PZ and CG and between CG and PU (P < 0.001 for all comparisons between imaging values and regions).

The values obtained for the PZ, CG, and PU in the inconclusive cohort were not significantly different from those of the cases (with the lowest P = 0.03 for T₂ in the PZ >0.05/3) or controls (with the lowest P = 0.04 for T₂ in the CG >0.05/3). The quantitative imaging values for this specific cohort are presented in the Supplementary Materials 2.

Radiation‐Induced Changes

To investigate radiation‐induced tissue changes, the values from the untreated patients were used as a population reference against which the values of the cases were compared.

The different prostate regions were more homogeneous after RT and the range of values was in general smaller than before treatment. With the exception of ADC (for which P = 0.03 > 0.05/4), T₂, K^trans, and k_ep values in the PZ were significantly lower (T₂, k_ep P < 0.001, and K^trans P = 0.002) after RT. The CG imaging values were for all maps significantly lower after RT (all with P ≤ 0.001). In the PU, despite the higher K^trans values seen before treatment, the imaging values for all maps remained comparable before and after treatment (P = 0.19 for T₂ up to P = 0.56 for ADC).

Similar characteristics were seen for all maps for recurrent and primary tumors (P = 0.17 for k_ep up to P = 0.22 for ADC).

Tumor Localization

To compare recurrent tumor with the remaining irradiated benign gland, the cohort of case patients was assessed. With all imaging maps, median tumor values were significantly different from benign tissue within the PZ (P < 0.001 for all imaging modalities). No differences were seen between the T₂ values of suspected tumor and benign CG tissue (P = 0.08). However, values of ADC, K^trans, and k_ep from suspected tumor regions were significantly different (all with P < 0.001) from those of the CG. When compared with the surrounding benign tissue, heightened K^trans and k_ep values were seen in the PU. The PU K^trans values were comparable with those of suspected tumor (P = 0.81), but with k_ep a significant distinction (P < 0.001) could be made between the two regions. The PZ and CG ROIs were significantly different for all maps except k_ep (P = 0.05 > 0.05/6).

Using the GLMM, we observed that univariately all MRI parameters were significantly associated with tumor likelihood (P < 0.001). Location within the gland was also significantly associated with tumor likelihood. The association between T₂ and tumor probability was nonlinear, and therefore this parameter was grouped in quartiles with univariate regression coefficients β = 0 (reference group), 0.19, –0.59, and –1.83. The regression coefficients were for ADC: β = –5.06, K^trans: β = 17.27, k_ep: β = 8.42 and location: β = –0.64. The multivariate model combining all mp‐MRI parameters and location obtained the best fit and was the most predictive for tumor likelihood (BIC of 392,212, P < 0.001). The values for the regression coefficients and standard errors are presented in Table 3. A reconstructed probability map for a case patient is presented in Fig. 5.

Table 3.

Model Parameters Obtained When Combining T₂ Map, ADC, K^trans, and k_ep Imaging Values to Predict Tumor at the Voxelwise Level

Fixed effects (MRI)	Regression coefficients (β)	Std. Error	P
T2 map reference group (24 – 73 msec)	0	—	—
T2 map 2nd quartile (74 – 83 msec)	0.08	0.01	<0.001
T2 map 3rd quartile (84 – 97 msec)	–0.40	0.01	<0.001
T2 map 4th quartile (98 – 1188 msec)	–1.50	0.02	<0.001
ADC	–2.86	0.02	<0.001
Ktrans	6.52	0.09	<0.001
kep	4.85	0.04	<0.001
Location	–0.98	0.01	<0.001
Intercept (β0)	–5.55	0.46	<0.001
Random effects	Variance (σ2)
Patients	1.41
Distance CMx	0.67
Distance CMy	8.50
Distance CMz	13.30
Residuals (ε0)	Median	First quartile	Third quartile
	–0.07	−0.13	−0.03

Imaging values and location are included as fixed effects; patient and spatial coordinates as random effects.

Figure 5.

A reconstructed probability map based on the generalized linear mixed model (a). In red, overlaid on the T₂w scan (b), is the same probability map thresholded at 0.5 and in (c), in yellow, the original tumor delineation. The anatomical and functional maps are presented in (d), together with the prostate and tumor delineations in white.

Discussion

In this study mp‐MRI was used to quantitatively describe and distinguish benign prostate tissue from recurrent PCa after RT. The use of two patient cohorts, with and without recurrent PCa, allowed us to show that tumor unaffected areas were similar in both groups. Compared with the pretreatment setting, and with the exception of ADC in the PZ, all other imaging values (T₂, K^trans, and k_ep) were lower, but also more homogeneous between the PZ and CG. This is in agreement with literature describing the loss of zonal distinction in the post‐RT setting.4 Still, significant differences between the imaging values of PZ and CG were seen in both recurrent and nonrecurrent patients. As in primary disease, recurrent tumor occurred predominantly in the PZ. This is not surprising, as recurrence usually happens at the site of primary PCa.19 For all maps, values of recurrent tumor were not significantly different from those of primary PCa.

With all imaging maps, we found that the PZ was significantly different from recurrent tumor, illustrating that all quantitative maps seem to have the potential to distinguish tumor in this region. However, tumor detection in the CG is challenging using solely the T₂ map, as the values for benign CG and tumor are not significantly different. This distinction is possible when evaluating ADC, K^trans, and k_ep. The PI‐RADS v2¹⁶ is the standard radiological guideline for the diagnosis of primary PCa. These guidelines do not apply to the detection of suspected recurrent PCa following treatment. Yet, in the absence of further guidance, PI‐RADS v2 is often used as a starting point for image assessment. The guideline describes T₂w as the dominant mp‐MRI sequence for the diagnosis of primary PCa in the transition zone. Our results suggest that this sequence is of less relevance compared with the other mp‐MRI sequences when diagnosing recurrent PCa in this region. The results also suggest that DCE‐MRI has an important role in the diagnosis of recurrent disease. In nonirradiated prostate tissue, the presence of benign prostatic hyperplasia (BPH) in the central gland is an important imaging confounder for the use of DCE‐MRI in this region. In a previous study, pathology slides from salvage prostatectomy specimens of recurrent PCa patients after RT were registered to mp‐MRI acquired prior to the surgery. In the analyzed pathology, no BPH could be found.20 This suggests that BPH vanishes after RT. The enhancement in the peri‐urethral area remains and needs to be considered. It has been hypothesized that the fibrotic changes observed in irradiated prostates,21 and the decreased microvasculature associated with atrophic tissue, enhance the neovascularity of recurrent tumor relative to the remainder benign atrophic tissue.22

With overall median values below 100 msec, T₂ is decreased for the prostate as a whole after RT. Measured T₂ values (msec) were in agreement with a previous study observing the effects of RT in prostate tissue23 (at week 8 of the RT treatment, mean ± SD: PZ = 89 ± 13, CG = 76 ± 5, tumor = 75 ± 9) and they provide evidence that the PZ and CG values do not increase after these 8 weeks.

The ADC values (in 10⁻³ mm²/sec) obtained for recurrent tumor were similar to those reported in other studies (mean ± SD: 1.0 ± 0.1 and 0.98 ± 0.23).24, 25 Tumor values in the untreated cohort were also comparable to those previously described (tumor PZ, mean ± SD: 1.08 ± 0.39).26 Values for benign prostate tissue in our case and control cohorts were slightly lower than those reported by others (mean ± SD: 1.6 ± 0.2 and 1.60 ± 0.21; ×10⁻³mm²/sec).24, 25 We note that these studies did not distinguish PZ from CG and reported values for a composite of both regions. In our untreated cohort, benign prostate tissue showed slightly lower values than those seen by Sato et al (mean ± SD: 1.80 ± 0.41 for PZ and 1.58 ± 0.37 for the transition zone).26

In controls, the PU region exhibited high K^trans and k_ep values without signs of malignancy. Similar high values were found in cases and this enhancement strongly overlapped with tumor values, which might represent a confounder for tumor detection in the CG. Yet with k_ep significant differences were seen between tumor and PU, suggesting that this could be a relevant parameter when assessing the CG. This effect has been previously described in the recurrent setting,6, 27 potentially as a result of remaining BPH or peri‐urethral vasculature. Donati et al28 found no additional benefit of DCE when complementing T₂w and DWI, which could have been the result of benign CG enhancement.

The GLMM analysis revealed that tumor distinction was optimized with the use of mp‐MRI, highlighting the importance of using a combination of anatomical and functional sequences when identifying and localizing tumor for salvage treatment strategies. Our study did not include spectroscopic data. Nonetheless, spectroscopic data has been suggested as a valuable tool for recurrent tumor distinction.24

Quantitative tissue characterization enables a more straightforward comparison between subjects, studies, and centers. It also conveys important information to adapt the scanning protocols to better image this study population. As an example, for patients with suspected recurrence the sampling echo times for T₂ mapping should be shortened, as T₂ decay is expected to be shorter compared with untreated patients.

We aimed to circumvent the limitations of a 2D standard transrectal ultrasound (TRUS)‐guided systematic biopsy by using a 3D MRI‐TRUS fusion approach to target the tumor‐suspected region. With this technique's improved spatial information, it has been described that fewer cores were necessary to detect more clinically significant cancer,29 often leading to reclassification of primary tumors in higher‐risk categories.30 Nonetheless, for a subset of our patients, cancer was only confirmed with repeated biopsy. Even when navigated, the biopsy procedure is limited by possible coregistration errors and in delivering the biopsy needle to the intended point.31 Tumor delineations performed using 1.5T T₂w were shown to underestimate the true tumor volume when compared with prostatectomy samples.19 As a result, the detection rates of MRI‐TRUS fusion biopsies might also be affected by errors in defining the lesion to target.

Our study has several limitations. First, it was a single hospital study with a limited number of patients. The controls were recruited to match the characteristics of cases as best as possible, and no significant differences were seen between their clinical parameters. However, our cohort of cases and controls had significantly higher‐risk primary PCa when compared with the untreated patients. Thus, the reported imaging differences for benign tissue could theoretically arise from baseline imaging dissimilarities. No significant imaging differences were seen between the tumors in the untreated and recurrent cohorts.

Ideally, using salvage prostatectomy specimens as the standard of reference would be preferred to accurately identify cancer location, but this procedure is only sporadically performed. Thus, patients with an inconclusive diagnosis were analyzed separately to avoid a potential bias. The diagnosis of recurrent disease is in practice made with a combination of mp‐MRI and PSMA‐PET findings. In this study we opted to individually investigate the properties of MRI which, due to its higher spatial resolution, is better suited for the purpose of tumor localization. A combination of both modalities can certainly help better select patients suitable for salvage treatment. Lastly, our GLMM was not tested in an independent cohort. Thus, further validation is required to establish applicability to other populations. The model is also potentially reflecting delineation practices.

In conclusion, tumor unaffected areas were similar between recurrent and nonrecurrent patients, and generally with lower and more homogeneous values than before RT. The PU region presents with increased enhancement in the pharmacokinetic maps for both cases and controls. Analysis of mp‐MRI, with the quantitative maps T₂, ADC, K^trans, and k_ep together with location information resulted in optimal distinction between tumor and benign voxels. The ability to accurately localize tumor and distinguish it from benign tissue with the use of mp‐MRI will help in the design of focal salvage treatment strategies for locally recurrent PCa.

Supporting information

Supplementary Table 1 – Sequence specific parameters.

Supplementary Table 2 – Median of all patient's median values for each region and imaging sequence for the inconclusive cohort. Minimum and maximum values are presented between brackets.