Study Design
This prospective clinical study (NCT04009174) was approved by the Institutional Research Ethics Board. All participants in this study provided written informed consent. Subjects who had untreated biopsy-proven localized PCa were prospectively enrolled in this study. Inclusion criteria were: aged 18 years or older; biopsy confirmed PCa; suitable for and consenting to radical prostatectomy for treatment. Exclusion criteria were: prior therapy including hormone therapy for PCa, use of 5-alpha reductase inhibitors - finasteride or dutasteride - within 6 months of study date; unable to comply with all pre-operative imaging; prostate size exceeding the dimensions of whole-mount pathology slides; allergy to CT contrast agent; sickle cell disease or other anemias; impaired renal function (estimated GFR < 60 mL/min/1.73m2); residual bladder volume > 150 cc (determined by post-void ultrasound); hip prosthesis and/or vascular graft that are MRI incompatible or other metallic objects within the pelvis; contraindication to MRI, such as a pacemaker or other electronic implants, known metal in orbits and cerebral aneurysm clips.
Participating patients underwent the pre-operative imaging session consisting of dynamic [18F]DCFPyL PET/CT, CTP, and PET/MR imaging. Dynamic [18F]DCFPyL PET/CT was performed first, followed by CTP using a hybrid PET/CT scanner (Discovery VCT, GE Healthcare, Waukesha, WI, USA). At 2-hour post-injection of [18F]DCFPyL, static PET and MR imaging were performed with a hybrid PET/MR scanner (3T Biograph mMR, Siemens, Malvern, PA, USA). All patients underwent radical prostatectomy. The explanted gland with implanted fiducial markers was imaged with T2w MRI then processed using spatially accurate whole-mount sectioning [17]. Cross-sections of the explanted gland spanning the entire mid-prostate were stained with hematoxylin and eosin (H&E) imaging with digital histopathology.
22-min Dynamic [18F]DCFPyL PET/CT Imaging
Dynamic [18F]DCFPyL PET imaging was performed on a Discovery VCT (GE Healthcare, Waukesha, WI, USA) PET/CT scanner. A CT scan was taken with patients lying supine on the patient couch for localization of the prostate and attenuation correction of PET images. The dynamic [18F]DCFPyL PET scan covered the whole prostate up to the iliac crest, image-derived arterial time-activity curve required for kinetic analysis of dynamic PET data was acquired from an internal iliac artery to generate parametric maps. Starting at the injection of 325 MBq of [18F]DCFPyL as a bolus into an antecubital vein, the dynamic PET scan acquired over 22-min the following number of volumes at each of seven framing intervals: 11 at 10 s, 5 at 20 s, 4 at 40 s, 4 at 60 s and 4 at 180 s. Each volume comprised of forty-seven 3.27 mm thick slices. Early time standardized uptake (SUVEarly) in g/mL was measured as the average of the last four dynamic PET volumes (10-22 min post-injection).
The acquired dynamic volumes were analyzed to generate parametric maps of the whole prostate. We calculated influx rate constant (K1) in mL/min/g, efflux rate constant (k2) in min-1, binding rate constant (k3) in min-1, dissociation rate constant (k4) in min-1, net uptake rate constant from plasma (Ki) in mL/min/g and distribution volume (DV) in g/mL maps by deconvolving the arterial time-activity curve from tissue time-activity curve using F2TC model [15]. The tissue time-activity curve of each voxel was smoothed by a 3-by-3 mean filter with those from the immediate neighboring pixels in the same slice.
CT Perfusion (CTP) Imaging
Free-breathing CTP scan was performed immediately after the dynamic PET scan without moving the patient. The CTP Images were acquired over 3 min using a shuttle mode where two contiguous 4 cm sections of the pelvis covering the prostate and both internal iliac arteries, identified from the CT localization and attenuation correction scan, were alternately scanned starting 6 seconds after a bolus injection of contrast agent (Isovue 370, Bracco Diagnostic Inc., NJ, USA) at a rate of 3 ml/s and a dosage of 0.7 ml/kg into an antecubital vein. The CTP images were acquired using 5 mm thick slices, 120 kVp and 50 mAs in two phases: the first phase, at intervals of 2.8 s for first 1 min; the second phase, every 15 s for the next 2 min.
The acquired dynamic CT volumes were co-registered using non-rigid image registration (GE healthcare) to minimize misregistration from breathing motion before functional maps, including blood flow (BF) in mL/min/100g, blood volume (BV) in mL/100g, mean transit time (MTT) in second, vessel permeability surface product (PS) in mL/min/100g and contrast delay time (T0) in second, were generated with CT Perfusion software (GE Healthcare).
2-hour post-injection PET/MR
After the CTP scan, the patient was allowed to rest, lying supine on a couch, in a quiet injection waiting room. At 2-hour post-injection, after emptying the bladder, a static PET image of the prostate using limits established with the prior dynamic PET/CT scan was acquired over 15 min. on a 3T Biograph mMR (Siemens, Malvern, PA, USA) hybrid PET/MR scanner [18]. PET images were reconstructed with an ordered subset expectation maximization (OSEM) algorithm with 21 subsets, 3 iterations and a 4 mm Gaussian filter. Reconstructed resolution was 2.09 x 2.09 x 2.03 mm3. MRI was acquired simultaneously with a prostate endorectal receive coil (Medrad), flexible body array (mMR Body) and spine array (mMR Spine). Attenuation correction maps were generated from a 2-point Dixon acquisition segmented for water, fat, and air. T2-weighted MRI included 2D transverse, sagittal and coronal acquisitions as well as a 3D acquisition. 2-hour post-injection SUV (SUVLate) maps in g/mL were measured from the acquired images.
Image Registrations and Analysis
After surgical resection, the prostate specimen was placed in 10% buffered formalin and marked with fiducial markers [17]. The specimen was then temporarily placed in an perfluoropolyether oil (Christo-Lube, Lubrication Technology, Franklin Furnace, OH, USA) filled container for imaging. Ex-vivo T2-weighted (T2w) MR imaging was performed with a 3T Discovery MR750 scanner (GE Healthcare, Waukesha, WI, USA). The oil provided a black background to minimize boundary artifacts in the MR images of the explanted prostate. After fixing in formalin for 48 hours, the specimen was processed as per the standard pathology grossing protocol at our institution and submitted for routine processing, paraffin embedding, and reporting. Whole-mount 4 μm-thick H&E-stained histology cross-sections spanning the entire mid prostate were then digitized for contouring and histopathological scoring by pathologists.
In order to correlate all PET, CT and MR maps to cancerous nodules identified in digital histopathology images of the explanted prostate, a registration pipeline, as shown in Fig. 1, was developed. The ex-vivo T2w MR was used as the bridge to co-register ex-vivo histopathology images to in-vivo maps. Histology, ex-vivo MR and in-vivo MR co-registration was done using fiducial markers and anatomical structures/landmarks [17]. Co-registration of CTP average map and ex-vivo T2w-MRI was done manually with ITK-SNAP (www.itk-snap.org) using anatomical landmarks in the excised prostate gland [19]. PET-CTP co-registration was achieved by registering the CT images acquired for attenuation correction of dynamic PET and CTP average maps with the automated 3D registration module - General Registration - in 3D slicer (www.slicer.org) [20]. After co-registration of all PET and CTP parametric maps with histopathology images, regions of interest were drawn on the latter images, to encompass the DIL and the entire prostate outside the DIL (non-DIL) by pathologists (contoured by MG and verified by MM and JAG). The DIL was identified as the largest cancerous lesion within the histopathology images. These ROIs were superimposed on all corresponding PET and CTP maps to generate voxel and volumetric data for DIL and non-DIL prostatic tissue.
Statistical Analysis
Statistical analyses were performed in SPSS Statistics 27 (IBM Analytics, Armonk NY), using two-sided statistical testing with significance accepted at P < 0.05. Descriptive statistics were presented as mean and SD or as median and range (min to max). Differences between DIL and non-DIL imaging parameters were compared using the Wilcoxon signed-rank test. Multivariable linear logistic regression with backward elimination was used to determine the most accurate model of [18F]DCFPyL PET and CTP parameters from a subset, each of which attained P < 0.05 in univariable testing, to distinguish DIL from non-DIL tissue. We performed leave-one-patient-out cross-validation to validate the selected logistic regression model for voxel-wise analysis and reported the average and standard error of error rate (ER), false positive rate (FPR) and false negative rate (FNR), area under the receiver operating characteristic curve (AUC) of all folds.