Introduction

Approximately 11–38% of patients treated with radical prostatectomy have a positive surgical margin (PSM) on final histopathology [1, 2]. Although conflicting results on the long-term oncological effects on survival have been published, adjuvant local radiotherapy was found to reduce biochemical recurrence rate and is therefore offered to men with PSM [1, 3, 4]. Complete surgical excision is challenging as the surgeon currently depends on visual and tactile appearance to distinguish cancerous from normal tissue. Additional modalities for intraoperative margin assessment can enable more radical excision. NeuroSAFE is a histopathological technique that applies fresh-frozen section evaluation on areas near the neurovascular bundle during surgery. This technique is currently used in several clinics to reduce PSM rates while improving functional outcomes [5,6,7]. As this technique aims at conserving the neurovascular bundle, rather than actual margin assessment of the prostate surface, only the posterolateral part of the prostate will be sampled. Next, the NeuroSAFE technique is labour intensive, time consuming (~ 45 min) and prone to sampling errors as only a small part of the prostate is assessed [7, 8].

Other imaging-based technologies to assess margin status are being developed, one of which is Cerenkov luminescence imaging (CLI). Cerenkov radiation is produced when a charged particle (e.g., beta particle from isotopes such as fluorine-18, gallium-68, zirconium-89, yttrium-90) travels faster than the speed of light trough a dielectric medium such as tissue [9,10,11]. When the locally polarized medium along the path of that particle returns to its ground state, broad-spectrum electromagnetic radiation known as Cerenkov radiation is emitted. Because of the weak light intensity of Cerenkov radiation, it can only be detected by means of highly sensitive optical imaging systems in a dark environment. Roughly one decade ago, CLI became of interest to the biomedical field [12], after publications that modern sensitive CCD detectors are able to visualize Cerenkov radiation in mice. The initial studies using fluorine-18 and zirconium-89 showed that accurate visualization and quantification of Cerenkov radiation is possible [10,11,12,13,14,15,16]. The first clinical study of CLI was published in 2013 using Iodine-131 to image the thyroid gland [17], and in 2017, the feasibility of CLI for intraoperative margin assessment was demonstrated in breast cancer [18]. In the latter study patients received a preoperative injection of 18F-fluorodeoxyglucose (18F-FDG), followed by breast-conserving surgery and intraoperative imaging of excised wide local excision specimens in a dedicated CLI system to assess margin status.

Applicability of 18F-FDG in prostate cancer (PCa) is unfortunately limited, since there is a low uptake of glucose due to the low rate of glycolysis [19, 20]. With the introduction of tracers that target the prostate-specific membrane antigen (PSMA), the role of molecular imaging in PCa changed rapidly. Positron emission tomography (PET) using gallium-68 (68Ga) or 18F-labelled PSMA-ligands are increasingly used for primary staging in high-risk patients, with excellent diagnostic performance [21]. An important aspect for CLI is the initial positron energy of an isotope; the higher the initial energy, the longer the path length through tissue, and hence, the higher the Cerenkov yield. So, in addition to the receptor-targeted accumulation of PSMA-ligands in PCa lesions, the advantage of 68Ga-labelled radiopharmaceuticals for CLI is the high Cerenkov yield [16]. In our previous study, we demonstrated that the Cerenkov yield from 68Ga is approximately 22× higher compared to 18F, which should enable clinical translation of CLI in PCa [22].

The aim of this first-in-man clinical study was to assess the feasibility of 68Ga-PSMA CLI for intraoperative margin assessment in PCa. Here clinical feasibility of the technology is defined as the ability to distinguish between a PSM and a negative surgical margin (NSM) by imaging the entire prostate within 45 min (typical duration of the NeuroSAFE procedure) and with low radiation exposure to staff.

Material and methods

Patient inclusion

Present study was approved by the local Medical Ethics Committee (NL8256). Six high-risk primary PCa patients scheduled for robot-assisted radical prostatectomy (RARP) were enrolled in this ongoing trial after written informed consent was obtained. The main inclusion criteria were a tumour lesion > 1.5 cm on MRI, and a PSMA-positive intra-prostatic lesion on PSMA PET/CT. In all patients, CLI images were not used for surgical or histopathological decision-making.

Surgery and intraoperative CLI

An intravenous injection of ~ 100 MBq [68Ga] Ga-HBED-CC-PSMA (ABX GmbH, Radeberg, Germany) was given immediately after the da Vinci® surgical system was docked. After removal, the prostate was prepared for Cerenkov imaging by rinsing it with sodium chloride to clear any potential radioactive contamination from blood or urine.

Images from six sides of the prostate (left/right, anterior/posterior, basal/apical) were acquired using the LightPath® CLI system (Lightpoint Medical Ltd., Chesham, UK) (Fig. 1). This device consists of a light-tight specimen chamber, and acquires a photographic reference image and a Cerenkov image which are automatically overlaid [23]. To reduce the noise from high-energy gamma photons (called gamma strikes), the images are immediately processed by the LightPath® software by applying a Gaussian filter (σ = 3 pixels) and a median filter (3 × 3 pixels) and then displayed. In our previous in vitro study, it was found that images with 120 s exposure, 2 × 2 binning and no filter were sufficient for detection of clinically relevant 68Ga-PSMA activity levels [21]. This imaging protocol formed the basis for the current ex vivo clinical study. To account for potential differences in signal levels that can be expected when converting in vitr results to an ex vivo clinical setting, in the current study, images were acquired with different acquisition times (30, 60, 150, 300 s), pixel binning (2 × 2 [234 μm], 4 × 4 [469 μm], 8 × 8 [938 μm]) and optical filter (no filter and 550 nm shortpass filter) with the aim to identify an acquisition protocol that provided sufficient sensitivity within a time-window feasible for intraoperative use (< 45 min).

Fig. 1
figure 1

Workflow of the current CLI study, a A preoperative 68Ga-PSMA PET-CT and MRI scan are acquired 3–4 weeks prior to surgery as per standard of care, based on the result of these images patients are included. b During surgery 68Ga-PSMA is administered i.v. after the da Vinci® surgical system is docked. Radiation dose to all surgical staff is monitored. Note the position of the scrub nurse in close proximity to the patient. Once removed, the prostate is rinsed with NaCl to clear any radioactive urine and blood that could be present on the surface. c The prostate is positioned in a disposable specimen tray. d Images of all six sides of the prostate using the different settings are acquired in the CLI device. e Unfiltered Cerenkov image of the intact prostate specimen. f Upon image completion of the intact prostate, the prostate is inked and cleaved ~ 1 cm from the apex by a trained person. g Image of the cleaved prostate and the corresponding CLI image acquired with the same settings as the intact specimen to confirm tumour uptake and quantify the intensity in benign and cancerous tissue

If areas of increased signal on CLI (called “hotspots”) were visualised on the intact images acquired without an optical filter, the acquisition was repeated with a 550 nm shortpass filter. As shorter-wavelength light has a higher attenuation and scattering in tissue, a persistent hotspot on the shortpass-filtered image indicates presence of PSMA-containing cancer cells near the excision margin. Based on visual assessment of the intact images, hotspots were marked with a suture for accurate comparison with histopathology.

Following CLI of the intact prostate specimen, the surface of the prostate was inked and cleaved 1 cm from the apex according to standard histopathological protocol. The cleaved prostate was subsequently imaged with CLI to visualize the primary lesion, to confirm presence of 68Ga-PSMA in the primary lesion and to quantify the signal in benign and cancerous tissue. The workflow of the CLI-procedure is displayed in Fig. 1.

Radiation safety monitoring

As 68Ga-PSMA is not routinely used in the operating room, the radiation exposure to the operating room staff was monitored using electronic personal radiation dosimeters (MGP Instruments DMC 2000; Mirion Technologies, Ltd.). Measurements included the scrub nurse positioned directly next to the patient, the surgeon positioned behind the da Vinci surgical robot at ~ 3 m from the patient, and the anaesthetist, periphery nurse and researcher positioned more than 1 m away from the patient.

Histopathology

After completion of CLI, the specimens were taken to the pathology department. Histopathological examination was performed as per standard of care by an experienced Uropathologist (EMB). A PSM was defined as cancer extending into the inked surface [24].

Image analysis

In addition to visual assessment of the intact prostate image to qualify if the margins are positive or negative, explorative quantitative assessment was performed after surgery using MATLAB R2017b (The MathWorks, Natrick, WA). The researcher (JoH) who performed this post hoc evaluation was blinded to the histopathology results, but had access to the patient’s PSMA-PET/CT and MRI. First, the images were post-processed by subtracting a background image (i.e. a CLI image of an empty specimen tray acquired with the same acquisition settings) from the clinical CLI images. This was performed to remove any camera noise in the image as well as the artefact from a defect pixel in the camera (Fig. 2). Measurements of the mean radiance (photons (p)/s/cm2/sr) were performed by manually selecting regions of interest (ROI) on the post-processed images of both the intact and cleaved prostate images around areas showing hotspots (tumour) and in areas with no increased signal (tissue background), and based on the latter ROIs, the tumour-to-background ratios (TBRs) were calculated. The quantitative analysis of the cleaved prostate images was performed with the aim to eventually identify a quantitative threshold for discriminating between a PSM and NSM. To enable comparison of signal levels between images of each patient individually, the signals (tissue background and tumour) were corrected for radioactive decay. Additionally, decay-corrected radiances were also corrected for injected activity to enable comparison of images between patients (corrected radiance). The process of image analysis is displayed in Fig. 2.

Fig. 2
figure 2

Overview of the image analysis and quantification methods. a Intraoperative Cerenkov image of the specimen after median and Gaussian filtering. This image is used for visual identification of PSMs and NSMs. b A Cerenkov image of an empty background. Note the presence of the defective pixel (green arrow). c Post-processed Cerenkov image after background subtraction. Note that the background subtraction removes the defective pixel on the post-processed image. d The PSMA PET/CT scan. The green line on the maximum intensity profile (MIP) image indicates the location where the prostate is cleaved. The transverse image of the fused PET/CT scan highlights the location of the tumour (red ROI) and benign tissue (blue ROI). e Intraoperative CLI image of the prostate from patient 2 after the specimen was cleaved at the apex, showing the tumour (red ROI) and the benign tissue (blue ROI). f CLI image of intact prostate specimen from patient 2. Two hotspots, corresponding with a histopathological PSM, can be identified (ROI 1 and ROI 3). Areas with no increased signal, corresponding with benign tissue from a NSM, can also be seen (ROI 2). For clarity, each ROI is also highlighted with a green arrow. For this patient the TBR was calculated by dividing the radiance from ROI 1 and ROI in 2. The dotted white line represents the location where the prostate was cleaved. g Histopathology image of the area corresponding with ROI 3 on the intact CLI image. A PSM can be identified (green arrow)

Results

Six patients were initially included into the study, but one patient had to be excluded due to unsuccessful 68Ga-PSMA labelling on the day of surgery. The mean injected activity was 83 MBq ± 19, and the mean time between injection and CLI acquisition of the prostate was 73 ± 14 min. An overview of the patient characteristics, 68Ga-PSMA activity, CLI results and histopathological results can be found in Table 1. Two out of five patients had a PSM according to histopathology (patient 1 and 2).

Table 1 Patient characteristics and CLI and histology results of all 5 patients

Figure 3 displays images with variable binning, filters and exposure times, which were used to evaluate and optimize the CLI acquisition settings on the intact prostate. An exposure time of 150 s was sufficient to detect cancerous lesions on CLI on both the intact prostate specimen (Figs. 3 and 4. Based on the intact prostate images, a 150-s exposure was favoured over a 300-s acquisition as the TBR for the latter did not sufficiently improve to outweigh the increase in acquisition time (TBR 1.85 vs 1.98). An 8 × 8 pixel binning improved the TBR compared with 2 × 2 and 4 × 4 binning, respectively (TBR 1.85, 1.26, and 1.06). Acquisitions without a filter had a higher TBR (4.33) compared to the ones with filter (TBR 1.85).

Fig. 3
figure 3

Results of protocol optimisation process with different acquisition settings. The images in the upper row are from patient 1 who had a histopathological PSM at the apex. All images were acquired with 8 × 8 pixel binning and 550 nm shortpass filter but with different exposure times of 30 s (TBR:1.10), 60 s (TBR:1.18), 150 s (TBR:1.85) and 300 s (TBR:1.98). The images in the centre row are also from patient 1and were acquired with 150 s exposure time and filter, but different pixel binning 2 × 2 (TBR:1.06), 4 × 4 (TBR:1.26) and 8 × 8 (TBR:1.85). The bottom row images are from patient 3 and were acquired with 150 s exposure time and 8 × 8 binning, but without a filter (TBR:4.33) and with filter (TBR:1.85), respectively. Note that the area of increased signal with the green arrow appears in all non-background subtracted images; this is an artefact from a defect pixel in the camera of the CLI system, and can be ignored. The colour bar indicates the scale of counts

Fig. 4
figure 4

An overview of the PET/CT scans and CLI images of all five patients. a The MIP of the PSMA PET-scan. Note that these scans were obtained using 68Ga-HBEDCC-PSMA (patient 5) and 18F-DCFPyl (patient 1, 2, 3, 4), hence the difference in PSMA-tracer distribution. b The transverse image of the fused PSMA PET/CT-scan at the level of the prostate lesion. c Non-filtered CLI images of intact prostate specimen. Note that patient one did not have a non-filtered image taken. d Filtered CLI images of intact prostate specimen. e Non-filtered CLI images of the cleaved prostate. The prostate specimen was cleaved at ~ 1 cm from the apex. Note that in patient 3, it was not located in the apex region but in the base, and therefore no tumour signal is visible on the CLI image. Scaling of the PET-scans was performed based on the intensity of the liver; CLI images were scaled visually based on the benign background signal. The displayed CLI images are the intraoperative images that are not corrected for background which explains the presence of the defective pixel. All CLI images were acquired with 150 s exposure time and 8 × 8 binning. *The patient numbers marked with a star had a PSM based on histopathology

Data from the CLI images of the cleaved prostate showed that the tumour had a higher radiance as compared with the benign tissue, both on the filtered (mean TBR 3.6 ± 1.4) and non-filtered images (mean TBR 2.1 ± 0.4), Table 1. As expected, the use of a filter decreased the radiance in both tumour and tissue background areas. In patient 3, the tumour could not be identified on the cleaved prostate image, as the tumour was not located in the apex region where the incision was made but in the base.

In the two patients with a PSM (patient 1 and 2), hotspots were identified in corresponding areas on both the non-filtered and the filtered intact prostate image, thus enabling successful identification of PSM on CLI (Fig. 4). Three out of five patients had a histopathological NSM. In one of these patients (patient 3), the NSM was correctly identified on CLI as the hotspot was present on the non-filtered images but not on the filtered image, potentially indicating PSMA-containing cells further away from the resection margin (Fig. 4). In the other two patients (patient 4 and 5), a hotspot was identified on the non-filtered and in one patient as well on the filtered CLI image, and in both cases, histopathology found tumour cells < 0.1 mm from the margin. In patient 5, the hotspot was not clearly visible on the filtered intact image; however, the quantitative TBR was high at that location.

The sterile scrub nurse received the highest dose because of their close proximity to the patient: 0.016 mSv. The average dose to the anaesthetist, surgeon, periphery nurse and researcher were substantially lower: 0.001 mSv, 0.005 mSv, 0.002 mSv and 0.001 mSv, respectively.

Discussion

In the current study, we present the first-in-man usage of intraoperative 68Ga-PSMA CLI in primary PCa. The results from the first patients enabled optimisation of the imaging protocol, and demonstrated the technical ability to correctly identify PSMs. The use of optical filters improves the ability to visually distinguish between PSM and NSM on CLI, a finding that has not been reported before. Importantly, the performed workflow was considered clinically feasible and safe for surgical staff in terms of radiation exposure levels.

These initial results show that by using an administered activity of ≥ 65 MBq 68Ga-PSMA, a 150-s acquisition time with 8 × 8 pixel binning, and specimen imaging with and without 550 nm shortpass filter, a PSM can be successfully detected. Compared with the published 18F-FDG CLI results in breast cancer, the current protocol uses a 3 times lower activity and a 2 times shorter acquisition time which is possible due to the higher tracer uptake and Cerenkov yield of 68Ga-PSMA [18]. Despite the longer duration of prostate cancer surgery, the lower injected activity and shorter half-life resulted in a roughly 2 times lower staff exposure. The radiation exposure to personnel, a maximum of 0.016 mSv per procedure, was within the International Commission on Radiological Protection limits [27]. Based on these values, a single scrub nurse can perform a minimum of 62 CLI procedures before exceeding the limits; a surgeon can perform 200 CLI procedures. All 6 sides of the prostate could be assessed with the CLI within approximately 20 min, which is well within the time window acceptable for intraoperative use. Although CLI may still delay surgery to some extent, the procedure is approximately twice as fast than a typical NeuroSAFE procedure that is only performed on the posterolateral surface of the prostate. Furthermore, CLI enables whole-specimen assessment in the operating theatre without the need for a dedicated pathology department, contrarily to NeuroSAFE. Next, when a lymph node dissection is performed as well, the surgery can continue with the dissection while CLI imaging of the prostate occurs.

In 3 out of 5 patients, the histopathological PSMs and NSMs were correctly detected on CLI. In two patients, a hotspot on the Cerenkov image from the intact prostate was obtained from a tumour deposit at < 0.1 mm from the inked surface. These false-positive results can be explained by the physical properties of 68Ga. The CLI signal is not produced in one spatial location, but instead along the positron trajectory in tissue (± 2.8 mm on average for 68Ga, which is the same as the PET range) [28]. Although the use of an optical filter restricts the detected signal to enable visualisation of activity on the surface [29], it remains hard to quantify from what depth this activity actually originates. As identified in this first-in-man study, depth estimation is a key challenge of CLI and an important limitation compared with NeuroSAFE. The spatial accuracy that is achieved with histopathological evaluation is far greater than will ever be achieved using CLI. This aspect is further accomplished by the fact that it is difficult to estimate the real amount of activity present in the prostate after administration based on the Cerenkov images alone. In the future, the amount of activity present in the excised prostate specimen could be measured using dose calibrators.

A quantitative approach could help to discriminate between benign and tumour based on a threshold value, rather than on visual interpretation alone, since visual assessment is inherently prone to subjective window-level settings, and may result in erroneous identification of hotspots. In this study, each prostate was cleaved to enable visualisation and quantification of benign and cancerous tissue in the apex. This resulted in a mean TBR > 2, showing that the technology is able to sufficiently differentiate between tumour and benign tissue. Still, the process of cleaving is not preferred in the eventual work up, as it is time intensive and requires additional training. Our ongoing study will explore the possibility to define a TBR based on preoperative PET/CT-scans to recognize and to quantify potential areas that have a higher risk for PSMs on Cerenkov images based on radiance values. However, in the current feasibility study, the numbers were too small to make quantitative conclusions from the radiance levels. For quantification, the time between injection and imaging, administered activity, can be used to normalise the data. Additional variables that influence the CLI signal that are more challenging to account for are the PSMA expression of the tumour, and the fact that CLI uses optical imaging which means that scattering and attenuation from superficial tissue and blood can alter the Cerenkov signal intensity.

Though the current histopathological definition of a PSM is 0 mm (i.e. tumour on ink) [24], it is suggested that CLI could still provide valuable information on margin status, not only to the surgeon but also to inform the pathologist about areas at risk to guide more detailed histopathological evaluation. Another possibility is to guide the NeuroSAFE procedure itself by indicating suspicious areas, so that only those areas could be sampled for NeuroSAFE assessment, thereby reducing the risk of sampling errors and the duration of the NeuroSAFE procedure.

The current feasibility study has several limitations. Firstly, this feasibility study only addresses a small selective population. The patients were selected based on a high chance of a PSM, thus not representative for the general population that will undergo RARP. Still, the included patients are the ones that could benefit the most from intraoperative margin assessment [29]. Secondly, the administered activity varied between the patients. The 68Ga-PSMA was produced locally at specific time slots each day due to the fact that this is a generator product. Thus, the actual time of surgery needs to correspond to the 68Ga-PSMA production time. Batches for multiple patients are retrieved from one production; therefore, once the syringe is prepared for the study patient, and surgery is delayed, the activity cannot be adjusted accordingly. This is further complicated by the relatively short half-life of 68Ga (68 min). To account for lower 68Ga-PSMA activity levels due to delays in the start of surgery, the exposure time could be increased, thereby increasing TBR and ultimately aiding the visual identification of PSM on CLI. Still, variation in activity present in the prostate is normal considering a different PSMA expression between patients, and even within one prostatic lesion. This could be correlated to the SUV from preoperative PET/CT scans.

Conclusion

This first-in-man study demonstrates that 68Ga-PSMA CLI is a promising and safe technique for intraoperative margin assessment in PCa. The validated acquisition settings enable detection of PSMs within an acceptable time window for intraoperative use, and acceptable radiation exposure to staff. The ongoing trial will further evaluate the diagnostic accuracy of CLI in a larger population and will assess the ability to establish a quantitative threshold to discriminate PSM from NSM, in addition to visual inspection alone.