Value of Fused 18F-Choline-PET/MRI to Evaluate Prostate Cancer Relapse in Patients Showing Biochemical Recurrence after EBRT: Preliminary Results

Purpose. We compared the accuracy of 18F-Choline-PET/MRI with that of multiparametric MRI (mMRI), 18F-Choline-PET/CT, 18F-Fluoride-PET/CT, and contrast-enhanced CT (CeCT) in detecting relapse in patients with suspected relapse of prostate cancer (PC) after external beam radiotherapy (EBRT). We assessed the association between standard uptake value (SUV) and apparent diffusion coefficient (ADC). Methods. We evaluated 21 patients with biochemical relapse after EBRT. Patients underwent 18F-Choline-PET/contrast-enhanced (Ce)CT, 18F-Fluoride-PET/CT, and mMRI. Imaging coregistration of PET and mMRI was performed. Results. 18F-Choline-PET/MRI was positive in 18/21 patients, with a detection rate (DR) of 86%. DRs of 18F-Choline-PET/CT, CeCT, and mMRI were 76%, 43%, and 81%, respectively. In terms of DR the only significant difference was between 18F-Choline-PET/MRI and CeCT. On lesion-based analysis, the accuracy of 18F-Choline-PET/MRI, 18F-Choline-PET/CT, CeCT, and mMRI was 99%, 95%, 70%, and 85%, respectively. Accuracy, sensitivity, and NPV of 18F-Choline-PET/MRI were significantly higher than those of both mMRI and CeCT. On whole-body assessment of bone metastases, the sensitivity of 18F-Choline-PET/CT and 18F-Fluoride-PET/CT was significantly higher than that of CeCT. Regarding local and lymph node relapse, we found a significant inverse correlation between ADC and SUV-max. Conclusion. 18F-Choline-PET/MRI is a promising technique in detecting PC relapse.


Introduction
In patients affected by prostate cancer (PC) and treated with external beam radiation therapy (EBRT) relapse may occur in about 50% of cases within 5 years after treatment [1] and the recurrence rate of PC in these patients is reasonably higher than in those treated with surgery. Increasing levels of PSA after EBRT are often related to PC recurrence, especially when PSA rises above 2 ng/mL [2] and PSA doubling time is <6 months [3].
Several imaging techniques have been used to detect sites of PC recurrence, such as bone scintigraphy (BS), contrastenhanced computed tomography (CeCT), and ultrasound (US). CeCT and US do not provide high diagnostic accuracy in detecting relapse [4,5]. BS, which has low specificity, often requires further adequate investigation. In this field, 18 F-Fluoride-PET/CT may be helpful, especially in high risk PC patients, as its high sensitivity and specificity in detecting bone lesions overcome the intrinsic limitations of BS [6]. At present, the most promising tools are multiparametric MRI (mMRI) and PET/CT with radio-labelled choline derivatives. MRI can accurately detect residual prostate cancer and locoregional recurrence after EBRT [7]. Choline-PET/CT, being a whole-body imaging technique, may locate lymph nodes and bone metastases in the majority of patients treated with EBRT [8][9][10]. The simultaneous acquisition of PET and mMRI with new PET/MRI scanners may be helpful, especially in doubtful cases [11], but this integrated imaging modality is currently not easily available. To date, however, little can be said about the accuracy of combined 18 F-Choline-PET/MRI in PC [12] and few studies have compared the accuracy of mMRI and 18 F-Choline-PET/CT in patients affected by recurrent PC [13]. The aim of this study was firstly to assess the diagnostic performance of fused 18 F-Choline-PET/MRI in patients with suspected relapse of PC after EBRT. We also compared the accuracy of 18 F-Choline-PET/MRI with that of mMRI, 18 F-Choline-PET/CT, 18 F-Fluoride-PET/CT, and CeCT in detecting local recurrence, lymph node, and bone metastases. We finally assessed the association between the standard uptake value (SUV-max) of PET and apparent diffusion coefficient (ADC) value of local recurrence and lymph node metastases.

Patient Population.
We prospectively evaluated 21 PC patients who showed biochemical relapse after first-line treatment with EBRT. Sixteen patients were treated only with EBRT and the others were treated with EBRT and androgen deprivation therapy (ADT). We included in our study only patients with Gleason score ≥7, PSA ≥ 2 ng/mL, and PSA doubling time (DT) ≤6 months. All patients underwent integrated 18 F-Choline-PET/contrast-enhanced (Ce)CT, 18 F-Fluoride-PET/CT, and mMRI within ten days. The main characteristics of the patients are summarized in Table 1. The Local Ethics Committee approved the study. All patients had been treated exclusively with EBRT as primary treatment. Patients had received a dose range of 74-81 Gy to the prostate gland and 72-74 Gy to the prostatic fossa. Therapy was performed in accordance with EAU guidelines [14]. All patients at the time of EBRT were classified as N0. 18 F-Choline-PET/CeCT Protocol. 18 F-Choline-PET/CT was performed in the fasting state (at least 6 h). An 18 F-Choline activity of 3 MBq/kg (IASOCholine IASON Labormedizin Gesmbh & Co. Kg, Linz, Austria) was administered intravenously; data were acquired 10 after the injection by means of a dedicated PET/CT system (Discovery ST; General Electric Healthcare Technologies, Milwaukee, WI). Low dose CT was acquired for both attenuation correction and topographic localization. The CT parameters used for acquisition were 140 kV, 80 mA, and 0.5 s per rotation and pitch 6 : 1, with a slice thickness of 3.25 mm equal to that of PET. PET was acquired in 3D mode from the upper neck to the upper thighs, by means of sequential fields of view, each covering 12 cm (matrix of 256 × 256), over an acquisition time of 3 min. Finally, a whole-body CeCT (120 kVp, -axis tube current modulation with minimum/maximum mA: 200/600) was acquired during the portal venous phase, 50 s after reaching the threshold of 100 UH in the mid thoracic aorta using a bolus tracking technique. Iopromide 370 mg I/mL (Ultravist, Bayer HealthCare Pharmaceuticals Inc., Montville, NJ) was used as intravenous iodinated contrast medium in all patients, and it was administered at a fixed flow rate of 3 mL/s to guarantee a constant iodine delivery rate of 1.2 g I/s. 18 F-Choline-PET/CT, CeCT, and 18 F-Choline-PET/CeCT studies were visualized on Xeleris Workstation version 2.1753 (General Electric, Milwaukee, WI, USA) which allowed PET/CT and PET/CeCT fusion and CeCT multiplanar reconstructions. 18 F-Fluoride-PET/CT Protocol. PET acquisition started 60 min after intravenous injection of 370 MBq 18 F-labelled Na F (IASOflu-IASON GmbHGraz, Austria) and included 10-12 bed positions. All PET scans were acquired in 3D mode (3 min emissions per bed position) and were reconstructed by using an iterative reconstruction algorithm. CT acquisition data were optimized to obtain diagnostic CT images also with small image reconstruction thickness using the following data: 120 kV, 120-400 mA with an automatic system of dose optimisation (Auto-mA, General Electric Medical Systems), tube rotation time 0.5 s, 16 mm × 1.25 mm detector configuration, and table feed 13.75 mm per rotation. Overall, we acquired only one CT scan, obtaining two image reconstruction thicknesses 3.75 mm (for attenuation correction and anatomic localization) and 1.25 mm (for multiplanar reconstruction). No iodinated contrast medium was injected. CT acquisition was performed from the skull vertex to the distal femur, with the same field of view used for 18 F-Fluoride-PET.

Multiparametric MRI.
Multiparametric MRI (mMRI) was performed with a 1.5T MRI scanner (Signa HDxt, GE Healthcare, Milwaukee, WI) equipped with 8-channel pelvic phased-array coil, according to a standardized protocol. FSE T2-weighted (T2w) sequences (slice thickness 4 mm, BioMed Research International 3 interslice gap 0.4 mm, in plane resolution 0.6 × 0.6 mm) with a large field of view were first oriented in the three orthogonal planes (sagittal, axial, and coronal), without fat saturation, enabling identification of the prostate gland and pelvic lymph node stations. High-resolution oblique axial and coronal scans (slice thickness 3-3.5 mm, interslice gap 0.3 mm) were further oriented perpendicular and parallel to the rectoprostatic plane, in order to avoid misinterpretation due to partial volume effects. Diffusion-weighted imaging (DWI) sequence (b 1000 s/m 2 , TR 3400, TE 76.3, bandwidth 250 kHz, FOV 40 × 40, slice thickness 4.0 mm, interslice gap 0.4 mm, matrix 128 × 128, number of excitations 6, and in plane resolution 1.5 × 1.5 mm) was acquired in the axial plane, using same slice locations of the first FSE T2w sequence. Dynamic contrastenhanced MRI (DCE-MRI) was acquired during intravenous injection of the paramagnetic contrast medium with a flow rate of 3 mL/s. A 3-dimensional spoiled gradient echo fat saturated T1-weighted (LAVA) pulse sequence was repeated in the axial plane 27 times with a temporal resolution of 12 s, during the injection of a single dose of contrast agent. An axial STIR sequence (TR, 5000 ms; TE, 80.1 ms; inversion time, 150 ms; thickness, 5 mm; interslice gap 0.5 mm; matrix, 256 × 256) was performed to detect focal bone lesions and for comparing mMRI with PET techniques (i.e., 18 F-Choline-PET/CT and 18 F-NaF PET/CT) in the detection of bone recurrence. Antiperistaltic drugs were given for minimizing motion artifacts due to intestinal peristalsis at the beginning of each mMRI examination. 18 F-Choline-PET/MRI. Multimodal imaging coregistration, fusion, and synchronized navigation were performed using a novel dedicated software, developed for research purposes (Quanta Prostate, Camelot Biomedical Systems s.r.l., Genoa, Italy). This software was developed for aiding multiparametric MRI (mMRI) interpretation and reporting, as well as multimodal MRI/PET coregistration. Quanta Prostate is able to simultaneously display different MRI datasets, allowing calculation of color-coded ADC maps from DWI sequences and wash-in/wash-out rate maps from time-intensity curves of DCE-MRI. Color-coded ADC and perfusion maps can be overlapped on T2w images and examined with different levels of transparency. Using Quanta Prostate, a deformable registration technique (employing nonlinear transformation and spatially varying deformable models) is applied for MRI/PET coregistration, in order to compensate for changes in patient positioning and local deformations between different imaging datasets (e.g., due to varying degrees of filling of the urinary bladder). Once MRI/PET coregistration is performed, the operator can perform quantitative measurements (ADC and SUV-max values) by drawing a ROI on the currently selected ADC or SUV map. 18 F-Fluoride-PET/CT. 18 F-Choline-PET/CT and 18 F-Fluoride-PET/CT studies were interpreted visually and semiquantitatively using the maximum standardized uptake value (SUV-max), on a patient-bypatient and lesion-by-lesion basis using the above mentioned dedicated software to review fused PET/CT images. However, no SUV-max cut-off value have been introduced and SUV was calculated for each lesion just to support visual interpretation.

18 F-Choline-PET/CT and
On 18 F-Choline-PET/CT, any focal and nonphysiological uptake corresponding to prostate parenchyma, abdominal and pelvic lymph nodes, and bone was considered as pathological.
On 18 F-Fluoride-PET/CT bone lesions were considered to be malignant if a high and focal 18 F-Fluoride uptake was detected without any morphological changes on CT or a high and focal 18 F-Fluoride uptake was associated to typical morphological changes on CT.

Multiparametric MRI.
All mMRI were reviewed on a dedicated workstation on (Advantage Workstation 4.6, General Electric Medical Systems), using a tool for mMRI visualization and reporting. A definitive diagnosis of cancer recurrence (i.e., local, lymph nodal, and skeletal recurrence) was formulated after examining each MRI examination, reporting the number of lymph nodal and skeletal lesions. Findings indicative of local recurrence were identified according to ESUR prostate MR guidelines, 2012, considering that radiation-induced fibrosis and shrinkage of the gland may often hinder lesion detection [15]. DWI sequence was examined by generating an ADC map, searching for intraglandular nodular foci characterized by restriction of water diffusion relative to the surrounding prostate parenchyma. ADC value of the glandular area affected by recurrent disease was registered and compared with the corresponding SUV-max value of 18 F-Choline-PET/CT. On DCE-MRI sequences, prostate cancer recurrence was recognized as an early enhancing area in contrast to the surrounding glandular tissue. A quantitative wash-in perfusion map was generated for aiding the detection of hypervascularized foci within the prostate gland. The following morphological and size criteria were used to define a metastatic lymph node: short axis diameter >10 mm for an oval lymph node and diameter >8 mm for a round lymph node. The ADC value of the most PET-positive lymph node of each lymph nodal station involved by recurrent disease was registered and compared to the SUV-max value of 18 F-Choline-PET/CT. Axial STIR sequence was employed to search bone metastases that appear as hyperintense focal lesions within the spongy or cortical bone compartments.

Contrast-Enhanced CT.
On CT scans, the separation between the prostate and the levator ani muscle is poorly defined, and intraprostatic anatomy is not well demonstrated. Only the following morphologic criteria were applied for defining local recurrence: prostate enlargement with focal capsule bulging; nodular areas of inhomogeneous contrast enhancement within the peripheral gland; sign of extracapsular extension. Bone metastases were considered to have an osteosclerotic or mixed (i.e., osteolytic and osteosclerotic) appearance.

Multimodal Fusion
Imaging. On 18 F-Choline-PET/MRI, any focus of nonphysiological uptake corresponding to any suspected mMRI finding in prostate gland was considered suggestive for local recurrence. Any focal uptake corresponding to mMRI detectable abdominal and pelvic lymph nodes was considered as lymph node metastasis. Any focal 18 F-Choline bone uptake corresponding or not to a pathological finding on mMRI was considered as bone metastasis. 3.6. Statistical Analysis. No specific sample size calculation was performed, given the pilot nature of the study. Categorical data were summarised as number (percentage) of subjects/lesions; continuous data were summarised as mean, standard deviation, median and range The normal distribution of datasets of different variables was assessed by means of the D' Agostino-Pearson test. When datasets did not follow the normal distribution, nonparametric tests were used instead of parametric ones. The degree of correlation between ADC and -max values for local and lymph node recurrence was assessed by the Spearman rank test. The diagnostic performance of different diagnostic modalities (i.e., mMRI, CeCT, 18 F-Choline-PET/CT, and 18 F-NaF PET/CT) for determining the presence of recurrent disease was calculated using 2 × 2 tables. Sensitivity, specificity, positive and negative predictive values, and overall diagnostic accuracy were calculated. The diagnostic performances of different modalities were compared by means of the Fisher exact test for proportions with level of statistical significance set at 0.05.

Patient-Based
Analysis. Fused 18 F-Choline-PET/MRI was positive in 18 of 21 patients, with an overall detection rate (DR) of 86%. Local relapse was detected in 6 patients, one of whom was also affected by bone metastases. Lymph node metastases were detected in 6 patients, two of whom also had bone involvement. Six patients showed only bone metastases. The DR of 18 F-Choline-PET/CT, CeCT, and mMRI was 76% (16/21 patients), 43% (9/21 patients), and 81% (17/21 patients), respectively. The difference between the DR of multimodal fusion imaging 18 F-Choline-PET/MRI and that of CeCT was statistically significant ( = 0.016), while, comparing mMRI and 18 F-Choline-PET/CT with CeCT, it only tended to be significant ( = 0.093 for both modalities). No significant difference was found between 18 F-Choline-PET/MRI and 18 F-Choline-PET/CT, nor between 18 F-Choline-PET/MRI and mMRI ( = 0.67 for both comparisons). Table 2 shows the DRs recorded when the different sites of PC recurrence were considered separately. Considering all skeletal metastases, including those beyond the field-of-view (FOV) of mMRI and multimodal fusion imaging 18 F-Choline-PET/MRI, 18 F-Fluoride-PET/CT provided the highest DR (detecting all 9 patients affected by bone metastases), which was found to be significantly different from that of CeCT ( = 0.034).
The accuracy, sensitivity, and NPV of fused 18 F-Choline-PET/MRI were significantly higher than those of both mMRI ( < 0.05) and CeCT ( < 0.05). No differences in  terms of specificity ( > 0.05) and PPV ( > 0.05) were observed between the different diagnostic modalities. A lesion-based analysis to determine the sensitivity of each modality in detecting recurrent lesions according to different anatomical locations (i.e., local recurrence, lymph nodes, and bone metastases) was also performed; these results are summarized in Table 3. 18 F-Choline-PET/MRI was more useful in detecting local relapse, identifying more local recurrences (6/6, 100%), than 18 F-Choline-PET/CT (4/6, 67%). Two cases of local recurrence are illustrated in Figures 1 and 2. However, the statistical difference between these two modalities was not significant, owing to the low number of local recurrences ( = 0.45). Fused 18 F-Choline-PET/MRI showed significantly higher sensitivity in locating lymph node metastases than mMRI ( = 0.0002) and CeCT ( < 0.0001). One case is illustrated in Figure 3. On the other hand, multimodality fusion 18 F-Choline-PET/MRI and 18 F-Choline-PET/CT detected the same number of lymph node metastases. With regard to whole-body assessment, we compared 18 F-Choline-PET/CT, CeCT, and 18 F-Fluoride-PET/CT in terms of their ability to detect bone metastases. Two cases of skeletal metastases are illustrated in Figures 4 and 5. The data regarding sensitivity are summarized in Table 4.

Correlation between ADC Values and SUV-max
With regard to local recurrences, we found a high and statistically significant inverse correlation between ADC value (7.4 × 10 −4 ± 2 × 10 −4 mm 2 /s, median 9.

Discussion
In this pilot study, we used different morphological and functional imaging modalities (i.e., mMRI, CeCT, 18 F-Choline-PET/CT, 18 F-Fluoride-PET/CT, and fused 18 F-Choline-PET/MRI) to prospectively examine a small, homogeneous cohort of 21 PC patients with biochemical recurrence after first-line EBRT. The clinical implementation of 18 F-Choline-PET/CT and mMRI in assessing recurrent PC after EBRT is currently yielding encouraging results   [7,16], and this growing evidence is confirmed by the data of our study showing high DRs for both modalities (i.e., 76% and 81%, resp.). Multimodal fusion imaging between 18 F-Choline-PET/CT and mMRI (fused 18 F-Choline-PET/MRI) yielded an even better DR (i.e., 86%), thus underscoring that multimodal coregistration, synchronized navigation, and combined interpretation are more valuable than the individual, separate assessment of different diagnostic techniques. Only preliminary studies [11,12] are available on the simultaneous acquisition of PET and MRI with integrated PET/MRI scanners on PC, and, to date, only one paper has addressed the issue of multimodal fusion PET/MRI imaging (with the two modalities acquired at separate times with different scanners) in the assessment of 17 patients with primary prostate cancer [17]. To the best of our knowledge, our study was the first to investigate the diagnostic value of this hybrid technique in the assessment of PC recurrence after EBRT. Park et al. [17] performed an intermodality MRI/ 11 C-Choline-PET fusion process assisted by highresolution ex vivo MRI of the prostate specimen, which also offered the integration of registered histologic information.
They created a parameter represented by the simple quotient of SUV over ADC value in a volume of interest within the prostate parenchyma, finding that the SUV/ADC ratio significantly increased the lesion-to-benign background contrast for tumours with Gleason grades ≥3 + 4. In our paper, we adopted an easier and more feasible coregistration technique, assisted by the new software Quanta Prostate (Camelot Biomedical Systems s.r.l., Genoa, Italy), to integrate morphological, functional, and metabolic information from different imaging modalities. With regard to the lesion-based analysis, 18 F-Choline-PET/MRI showed significantly higher accuracy, sensitivity, and NPV than mMRI ( < 0.05) and CeCT ( < 0.05). Specifically, 18 F-Choline-PET/MRI detected more lymph node metastases than mMRI and CeCT. This result may be due to the fact that only morphological and size criteria were adopted for defining a metastatic lymph node (i.e., short axis diameter >10 mm for an oval lymph node and diameter >8 mm for a round lymph node) on both mMRI and CeCT. In a previous pilot study [13] on 14 PC patients, Beer et al. compared 11 C-Choline-PET SUV with ADC values of pelvic lymph nodes suspected for metastatic involvement. They found a moderate but significant inverse correlation between SUV and ADC values and a significant difference in mean ADC and SUV values between benign and malignant lymph nodes. Our results confirmed this significant inverse correlation between SUV and ADC values. However, a major limitation of DWI in PC is the lack of specificity of the ADC value in discriminating between benign and malignant lymph nodes. We did not find any significant difference in terms of diagnostic performance between 18 F-Choline-PET/MRI and 18 F-Choline-PET/CT, but this latter technique was able to identify only 4/6 cases of local relapse versus 6/6 with the former technique. Although the diagnostic value of T2w imaging is hampered by radiation-induced fibrosis and shrinkage of the prostate, functional MRI techniques (i.e., DWI sequences and DCE-MRI) are effective in the detection of local PC recurrence. When multimodal fusion imaging is used, perfect matching between a focus of 18 F-Choline uptake and a suspicious mMRI finding on fused PET/MRI images may improve the diagnostic confidence of each separate diagnostic modality. Despite the low number of patients included, a significant inverse correlation was also observed between ADC and SUV values in the case of local recurrent disease. This supports the relationship in PC between the high cellularity expressed by ADC and the metabolic activity of phospholipid turnover expressed by SUV. With regard to the detection of skeletal metastases in pelvic bones, fused 18 F-Choline-PET/MRI performed significantly better than CeCT (Table 4). In addition, on comparing CeCT, 18 F-Choline-PET/CT, and 18 F-Fluoride-PET/CT in the detection of bone metastases throughout the entire skeleton, we found that the highest sensitivity was obtained with 18 F-Fluoride-PET/CT. However, a significant difference emerged only between 18 F-Fluoride-PET/CT and CeCT. 18 F-Choline-PET/CT proved to be slightly less sensitive than 18 F-Fluoride-PET/CT, as previously reported [18], in which it failed to detect one small sclerotic rib metastasis.
Despite our encouraging results, some limitations should be noted. As the number of patients in our study was small, no specific sample size calculation was performed. In addition, some questions may arise with regard to the reference standard that we adopted. Indeed, no really independent tool for classifying lymph node and bone lesions was available, and, for logistic and ethical reasons, histological confirmation of metastases was not obtained in the majority of cases. However, as in two previous studies [10,16], clinical, laboratory, and follow-up data were collectively considered and used as a standard of reference. This type of multidisciplinary followup is generally accepted to confirm lymph node and bone metastases. Moreover, histological confirmation was available in 3/5 patients showing only local recurrence and in 1/6 patients affected by only lymph node metastases.

Conclusions
According to our preliminary results, 18 F-Choline-PET/MRI fusion imaging may be considered a feasible and promising diagnostic tool for detecting PC recurrence in patients showing biochemical relapse after first-line treatment with EBRT.