Fifteen healthy male volunteers (mean age 35.9 years, median 34, range 27–53) were included in this prospective, institutional review board-approved study (Reference: CUH 15/YH/0570). Participants were recruited through posters on campus between March and June 2016, with written informed consent obtained in all cases. MR imaging was performed on four consecutive days. The subjects were instructed to abstain from ejaculation for at least 3 days prior to the first MRI, to perform ejaculation after scan 1 and prior to scan 2, then refrain from ejaculation until the completion of the study. Scans 2, 3 and 4 were therefore performed 1–24 h, approximately 48 h and 72 h post-ejaculation, respectively.
Magnetic resonance imaging
MR imaging was performed on a 3-T MR750 magnet (General Electric Healthcare, Waukesha, USA) using a 32-channel phased array body coil. Sequences included high resolution axial T2-weighted fast recovery fast spin echo (FRFSE) imaging, TR/TE of 3663/102 ms field of view (FOV) 22 × 22 cm, 3 mm slice thickness with no gap, in-plane resolution 0.85 × 0.57 mm, and 3 signal averages; sagittal T2 cube sequence, 1 mm slice thickness with no gap, in-plane resolution 1.0 × 0.8 mm. Axial diffusion-weighted imaging (DWI) was matched to the T2 axial sequence, using a dual spin-echo planar pulse sequence with TR/TE of 3775/70 ms, FOV 28 × 28 cm, resolution 2.2 × 2.2 mm, 6 signal averages and b values of 150, 750, 1000 and 1400 from which automated ADC maps were generated (Table 1).
Table 1 Sequences in MRI protocol
MR segmentation
Seminal vesicle (SV) and prostate volumes were calculated using whole volume segmentation on T2-weighted images using in-house software programmed in Matlab (Supplemental data 1). Whole-gland ADC measurements were acquired after outlining the prostate using the relevant T2-weighted axial sequence as an anatomical reference. ADC values were recorded from each voxel with the regions of interest (ROIs) and, after summation of values from all slices, a whole-gland ADC mean value was derived. Whole-gland T2 signal intensity was recorded from prostate outlines normalised to muscle signal intensity as an internal reference. Three ROIs were drawn within the left obturator internus muscle (≥0.5 cm3) on consecutive slices, with a ratio of median whole-prostate to muscle T2 signal intensity recorded for each study. All outlining was performed by a single uro-radiologist (T.B.) with 7 years’ experience reporting prostate MRI. Outlines were drawn in a random order and blinded to the clinical information of study number and time pre- or post-ejaculation.
Organ volumes (in cubic millimetres) were calculated from a sum of the drawn ROIs (in square millimetres) multiplied by the spacing between acquired slices (in millimetres). An alternative method was also employed incorporating quadratic interpolation in the form of Simpson’s rule applied in the slice direction when summing ROI areas on adjacent slices. The two methods were compared for approximate equality of results as a quality control check. The subsequent analysis was then performed on the interpolated volume measurements. For SV volume, a separate calculation was made of fluid volume by excluding the wall volume which is expected to remain constant, but contributes proportionately more to the overall volume when the seminal vesicles are collapsed or underfilled. As the seminal vesicles contain fluid and organ wall in a convoluted pattern which is difficult to outline manually, a thresholding method was applied to yield a non-connected subset of areas which excluded the wall volume. The threshold on image intensity was set to a fraction (f) of the maximum pixel intensity (S
max) calculated as the 95% percentile in the SV ROI intensity histogram. Thus, for any given organ, the maximum signal in the central outlined slice was found and the threshold for inclusion of pixels set to S > (S
max × f). The fraction f was varied manually at limits of 0.7 or 0.8 (depending on the presence of motion-induced ‘blurring’) to achieve optimal segmentation as evaluated through visual inspection by a radiologist (T.B.) blinded to the clinical information (Fig. 1). Fluid volume was then calculated in the same way as described above, multiplying the thresholded total of ROI subareas by the spacing between slices.
Statistics
A linear mixed-effects model was fitted to the seminal vesicle fluid volume and whole-prostate ADC observations. Each model included fixed effects for the time of the observation (baseline, day 1, day 2 or day 3) and random effects for each individual to account for correlation within individuals. A log transform was applied to the data to ensure homoscedastic error variance. A paired two-tailed t test was used to compare the ratio of T2 signal intensity to muscle between studies in a stepwise manner. Spearman’s correlation was performed to assess the relationship between age and both prostate and seminal vesicle volume and between time post-ejaculation and change in SV volume from baseline to day 1. Statistical analyses were performed using Stata®14 (StataCorp LP, Texas, USA). p values of less than 0.05 were considered to be statistically significant.