Tumour perfusion is a universal cancer biomarker (1), with potential value in characterizing the aggressiveness of primary prostate tumours and separating significant prostate cancer from insignificant cancer (2–5). [¹⁵O]H₂O positron emission tomography (PET) is the gold standard for non-invasive quantitative perfusion imaging. A potential application for tumour perfusion imaging could be for evaluation of treatment response through serial measurements before, during and after a specific treatment. To evaluate whether a change in perfusion is significant or not, the reproducibility of the measurement must be known. Previous test-retest studies found a high reproducibility of perfusion measurements with [¹⁵O]H₂O PET in various non-prostatic tumours (6–8). Previously, our group assessed the repeatability of perfusion measurements in primary prostate cancer using ⁸²Rb PET (9), however no data on [¹⁵O]H₂O PET in prostate cancer exists. Hence, the aim of the present study is to determine the repeatability of [¹⁵O]H₂O tumour perfusion in prostate cancer.

All patients underwent two scan sessions, each consisting of a 5-minute dynamic pelvic [¹⁵O]H₂O PET scan and a dynamic [¹⁵O]H₂O PET heart scan for obtaining an image-derived input function. All scans were performed on a 3 Tesla GE Signa PET/MRI Quant Works scanner (GE Healthcare, Waukesha, Wisconsin, USA).

A standardized 400 MBq [¹⁵O]H₂O bolus was injected via an automatic injection pump at the beginning of each bed position, followed by 5 minutes dynamic PET acquisition.

Frame structure 1x10s, 1x5s, 15x3s, 5x5s, 2x10s, 5x15s, 4x30s. Voxel size 2.8x2.8x2.8 mm3. Reconstruction algorithm ordered subset expectation maximization (OSEM) with point-spread-function (PSF) and time-of-flight (ToF) (VuePoint FX SharpIR). The images were filtered with a 3 mm transaxial Gaussian filter and a light axial filter [1:6:1].

Tumour volumes of interest (VOIs) of thirteen well-defined MRI PI-RADS lesions from ten patients were drawn directly on the T2 MRI sequence by visual guidance, taking all available sequences into account. All VOIs were drawn using Hermes Affinity viewer version 3.0.1 (Hermes Medical Solutions, Stockholm, Sweden).

Tumour VOIs were subsequently transferred to the parametric [¹⁵O]H₂O PET K₁ images and [¹⁵O]H₂O PET image series for direct reading of tumour perfusion and extraction of time-activity-curves (TACs), respectively.

Heart image-derived input functions (HIDIF) were extracted automatically from the separate dynamic [¹⁵O]H₂O cardiac scan series by cluster analysis to identify venous and arterial clusters (10). These blood input functions were used for both parametric image calculation and VOI-based analysis. VOI-based analysis was performed with correction for delay and both with and without correction for dispersion (estimated using the image-derived input function from pelvic arteries (PIDIF) (5) (Table 1).

Parametric K₁ images were constructed using either HIDIF with correction for delay and dispersion or using HIDIF with voxel wise delay calculated from the Leading-Edge method (11) (Table 1).

[¹⁵O]H₂O wash-in (K₁) (mL/min/mL) and wash-out (k₂) (mL/min/mL) was calculated using a single-tissue compartment model. Kinetic analyses and blood input function extractions were performed using the aQuant Research Package (MedTrace, Hørsholm, Denmark).

Patient characteristics are presented in Table 2, while test and retest measurements using different reconstructions and corrections are found in Table 3.

In general, the different reconstructions performed almost equally and correlated excellently both between VOI methods and between VOI and parametric methods. There was a tendency towards larger variation when applying the correction for dispersion (VOI_PIDIF) for the VOI method and when using the Leading-Edge method for delay correction in the parametric images (Param_LE) (Fig. 2).

Measures of ICC and repeatability coefficients are listed in Table 4. Repeatability for k₂ was poor (repeatability coefficient 67.9% – 71.4%), probably explained by poor signal to noise ratio in small lesions with relatively low perfusion (Supplementary Fig. 1).

Sample size calculations for a longitudinal study on change in tumour perfusion based on the repeatability from the present study are found in Table 5.

The main result of the present study is that tumour perfusion in primary prostate cancer can be measured with a repeatability of around 30%, meaning that increase or decrease above 30% in the individual patient is likely to represent actual changes in tumour physiology.

As tumour perfusion correlated to International Society of Urological Pathology (ISUP) grade and hence aggressiveness in previous studies it is a relevant measure in prostate cancer (2–5). Tumour perfusion is a marker of nutrient agnostic growth potential; hence it might represent a more relevant aspect of tumour biology compared to prostate-specific membrane antigen (PSMA) expression. FDG is not an obvious alternative as prostate cancer is often not FDG-avid.

Potential clinical applications could be for monitoring small tumours with low ISUP grade followed by urologists in active surveillance, as a more aggressive treatment approach might be considered for tumours with high K₁. Besides, repeated tumour perfusion measurement might be a good method for monitoring the effect of local treatment of prostate cancer or for assessing effects of systemic neoadjuvant therapy prior to local treatment.

As these potential indications usually concern small tumours, it is crucial to show that tumour perfusion can be measured with high repeatability even in small tumours below 1 cm³.

In contrast to our previous repeatability study on ⁸²Rb PET, this study assessed the repeatability of the method unaffected by the day-to-day variability, which will inevitably affect longitudinal clinical scans during treatment. The patients were scanned on PET/MRI and instructed not to move between scans to ensure optimal alignment between MRI-derived VOIs and perfusion images. The present cohort of 10 patients with 13 primary lesions is comparable with most previous repeatability studies on PET tumour perfusion (8, 9, 14), and half the size of the cohort examined by Lodge et al. (7).

The repeatability of 30% found in the present study is comparable with those from Lodge et al. (7) and our previous study on ⁸²Rb PET (9), while the studies from de Langen et al. (14) and van der Veldt et al. (8) found somewhat lower repeatability of 16–18%. Explanations for this deviation could be the method of input function and that we included much smaller tumours than previous studies, even below 1 cm³.

Lodge et al. discussed the possibility of reducing the repeatability to approximately 26% by averaging two repeated measurements at each time point (7). Another possibility to increase repeatability could be to utilise the superior sensitivity of long axial field-of-view PET scanners, which would also obviate the need for a separate heart scan. However, the repeatability of [¹⁵O]H₂O tumour perfusion on long axial field-of-view PET scanners remains to be determined. Using PET/MRI has the obvious advantage of localising the tumour anatomically, which is needed for VOI definition in subjects with concurrent benign hyperplasia.

As gold standard for non-invasive measurement of tumour perfusion, [¹⁵O]H₂O PET is a robust tool for assessing tumour biology. With increasing availability of [¹⁵O]H₂O generator systems and other perfusion agents, routine PET-assessment of quantitative tumour blood flow is now within reach at multiple PET-centres worldwide.

Regarding primary prostate tumours, benign hyperplastic nodules with increased perfusion is one of the main challenges as shown in Fig. 1, row B. This illustrates both that PET perfusion is merely suited for tumour characterization than for tumour detection and the importance of a separate modality for VOI definition.

Monitoring of oncological treatment response in patients with metastatic prostate cancer could potentially be a promising application, especially with the introduction of long axial field-of-view PET scanners that allow quantitative assessment of tumour perfusion in patients with metastatic disease (11). On this topic, the repeatability of [¹⁵O]H₂O perfusion measurements in prostate cancer metastases is also unknown.

The potential clinical applications of PET tumour perfusion imaging for characterization and monitoring prostate cancer patients needs to be explored in future studies.

[¹⁵O]H₂O perfusion PET is a repeatable method for measurement of tumour perfusion in localized prostate cancer, showing comparable performance when using parametric K₁ perfusion maps and VOI-based analysis. For longitudinal evaluation, changes above 30% are likely to represent actual changes in tumour perfusion, for example as treatment response.