Cerenkov Luminescence Imaging in Prostate Cancer: Not the Only Light That Shines

DISCUSSION

The primary aim of the current study was to assess the concordance in margin status between intraoperative CLI and postoperative histopathology. This study showed that agreement between CLI and histopathology ranged from 31% to 83% depending on definition of PSM and addition of image interpretation by LS. The explorative aim of the study was to characterize the observed chemiluminescence. This signal was found to be unrelated to ⁶⁸Ga-PSMA-11, to have a mean SBR of 4.9 ± 1.1, and to have a half-life of 48 ± 11 min. This chemiluminescence indeed proved clinically relevant as it hampered easy identification of 4 of 10 PSMs on CLI, all 4 located at the base.

When the histopathologic PSM definition is considered as tumor on ink, the false-positive rate found in the current study can be considered high, with 24 of 30 hot spots (LS ≥ 4) not having a PSM on histopathology (agreement, 31%). Comparable false-positive rates were also described by Darr et al., though their workflow and CLI protocol were slightly different (13). Several explanations for these false-positives findings can be considered. First, in 19 false-positives, the tumor–to–specimen edge distance was within 1 mm on histopathology. Because of the average positron range of 2.8 mm for ⁶⁸Ga (20), the CLI technique has inherent difficulties in the detection of PSMs defined as tumor on ink. If instead of tumor on ink a PSM is defined as a margin of 1 mm or less, agreement between CLI and histopathology increases to 83%. The clinical relevance of a CSM is a matter of debate; some argue that compared with an NSM, a CSM results in a higher rate of biochemical recurrence, though it is not an independent predictor (21). A CSM was found to have a role in biochemical recurrence by Izard et al. as well, though those investigators defined a CSM as being less than 0.1 mm (22). CLI might have added value in guiding surgeons toward areas with an increased risk on PSMs and CSMs. The positron range might also influence the correlation between the extent of the PSM on CLI and histopathology; in 5 of 10 PSMs, the deviation between the extent on CLI and histopathology was more than 1 mm (Table 2). This difference can also be caused by shrinkage of the specimen during fixation and histopathologic processing. Still, it is worth further exploring, as the extent of a PSM is an important prognostic factor (23).

Second, the presence of the chemiluminescence signal also contributed to the false-positive  findings,  especially  in  the  base. In all included  patients, chemiluminescence was visible in the base in varying degrees, even when the primary tumor was not located there, and it was observed in nonradioactive specimens as well. The latter proved that this signal does not originate from ⁶⁸Ga-PSMA-11. Extensive rinsing with sodium chloride did not make the signal disappear, thus eliminating the effect of radioactive urine and blood contamination, as was initially suggested by Darr et al. (13). Our hypothesis is that this signal is chemiluminescence, caused by the use of diathermy. The sudden heat from electrosurgery is known to result in the production of reactive oxygen species (24). After a cascade of reactions, decomposition of the molecule occurs, resulting in emission of optical photons. These photons partially overlap with the Cerenkov spectrum (9,13,14). In our institute, diathermy is performed mainly when separating the prostate from the bladder, though it is also used in other areas. Therefore, chemiluminescence is most prevalent at the base, explaining why the correspondence between CLI (hot spot, yes or no) and histopathology was only 20% (Fig. 2). Despite the fact that the addition of image interpretation with an LS improved agreement, 4 PSMs at the base were false-negative on CLI with an LS of 3 (Fig. 3C). To facilitate better detectability of PSMs, the need for a method to distinguish between Cerenkov and the diathermy-induced chemiluminescence is eminent.

The mean SBR of the chemiluminescent signal and the mean number of PSM-suspected hot spots on unfiltered images were comparable: 4.9 ± 1.1 and 5.3 ± 2.3, respectively (Fig. 3). In 550-nm filtered images, the SBR of the PSM was larger, at 9.9 ± 5.5, than the chemiluminescent signal, at 4.2 ± 0.3, though they still overlapped. Next, filtered images are more susceptible to noise, resulting in high SBRs. A study by Spinelli found a heat-induced chemiluminescent half-life of 4–6 min in chicken breast tissue (15), suggesting a potential distinction between Cerenkov and chemiluminescence based on half-life. However, the half-lives of the 2 signals in prostatic tissue had a large overlap; chemiluminescence half-life was 48 ± 11 min, compared with 68 min for ⁶⁸Ga. The discrepancy in chemiluminescence half-life compared with the findings of Spinelli may be explained by the differences in tissue composition and methods used to induce thermal damage. A focused ultrasound transducer (60 W) was used by Spinelli, whereas in our institution a monopolar diathermy device (Intuitive) at 40 W was used.

Another solution may be to look for alternative forms of monopolar diathermy that have a lower wattage resulting in less thermal induction and spread (25,26), such as the PEAK PlasmaBlade (Medtronic). Next, a theoretic method to reduce the presence of free radicals in tissue, thereby preventing lipid peroxidation, might be the use of antioxidants such as vitamin C or E (27), or the influence of rapid cooling might be investigated. Whether these solutions are viable requires further investigation, and for clinical use it is of key importance that they not interfere with the surgical outcome or histopathologic evaluation of the specimens.

FAR, an opaque film that largely blocks optical photons while allowing positrons to be detected directly, was used to investigate whether such blocking is less susceptible to chemiluminescence as CLI. The signal originating from ⁶⁸Ga-PSMA-11–containing areas (tumor and benign) were enhanced approximately 3.2 times, compared with these areas on CLI (Fig. 5; Supplemental Fig. 3), whereas chemiluminescence was reduced to 70%. Although this technique might help to distinguish between ⁶⁸Ga-PSMA-11 and chemiluminescence, there are some practical drawbacks to consider. First, because the FAR obscures the white-light image, anatomic and spatial information about the specimen is absent. Second, accurate positioning of the FAR onto the specimen proved challenging, and more work is needed to improve ease of use. Third, the FAR does not block all optical photons originating from the chemiluminescence (Fig. 5). Finally, contamination of the scintillator might lead to false-positive hot spots (Fig. 5).