Cherenkov emissions for studying tumor changes during radiation therapy: An exploratory study in domesticated dogs with naturally-occurring cancer

Abstract

Purpose: Real-time monitoring of physiological changes of tumor tissue during radiation therapy (RT) could improve therapeutic efficacy and predict therapeutic outcomes. Cherenkov radiation is a normal byproduct of radiation deposited in tissue. Previous studies in rat tumors have confirmed a correlation between Cherenkov emission spectra and optical measurements of blood-oxygen saturation based on the tissue absorption coefficients. The purpose of this study is to determine if it is feasible to image Cherenkov emissions during radiation therapy in larger human-sized tumors of pet dogs with cancer. We also wished to validate the prior work in rats, to determine if Cherenkov emissions have the potential to act an indicator of blood-oxygen saturation or water-content changes in the tumor tissue-both of which have been correlated with patient prognosis.

Methods: A DoseOptics camera, built to image the low-intensity emission of Cherenkov radiation, was used to measure Cherenkov intensities in a cohort of cancer-bearing pet dogs during clinical irradiation. Tumor type and location varied, as did the radiation fractionation scheme and beam arrangement, each planned according to institutional standard-of-care. Unmodulated radiation was delivered using multiple 6 MV X-ray beams from a clinical linear accelerator. Each dog was treated with a minimum of 16 Gy total, in ≥3 fractions. Each fraction was split into at least three subfractions per gantry angle. During each subfraction, Cherenkov emissions were imaged.

Results: We documented significant intra-subfraction differences between the Cherenkov intensities for normal tissue, whole-tumor tissue, tissue at the edge of the tumor and tissue at the center of the tumor (p<0.05). Additionally, intra-subfraction changes suggest that Cherenkov emissions may have captured fluctuating absorption properties within the tumor.

Conclusion: Here we demonstrate that it is possible to obtain Cherenkov emissions from canine cancers within a fraction of radiotherapy. The entire optical spectrum was obtained which includes the window for imaging changes in water and hemoglobin saturation. This lends credence to the goal of using this method during radiotherapy in human patients and client-owned pets.

Fig 1. Setup of canine and camera for multibeam radiotherapy.

A) A close-up image of the tumor on the shoulder of this dog demonstrates where the Oxylite probes were placed within the tumor and normal tissue. On top of the tumor is a Cherenkov-transparent bolus. B) The camera was placed on the table behind the subject so the gantry would not interfere with the camera, and any table movements would maintain the same camera-tumor position. C) From the camera’s point-of-view, the tumor is visible.

Fig 2. RT and Cherenkov-imaging workflow.

One of our aims was to determine if integration of this camera system with the RT workflow is achievable. We found the workflow to be simple. Once the camera was setup, the imaging required just two extra steps: (1) turning off the room lights for the duration of the RT fraction; and (2) pressing the acquisition button in the software. For this study, we simply observed biology that occurred during a standard radiation dose fraction; however, to maximize the amount of data that could be acquired per dog; the individual dose fraction was broken into several equally-sized subfractions that were each given at a standard dose rate (600 monitor units per minute) but temporally separated by 5 minutes.

Fig 3. Anatomical, Cherenkov and CT images of the first canine subject.

A) Representative low-light anatomical (non-Cherenkov) image of canine subject 1 (see Table 1), acquired during the second subfraction. The tumor, foreleg and the caudal ventral edge of bolus are labeled in red. B) Here, the same dog is depicted but with the Cherenkov images. The gantry was angled at 40˚. C) The Cherenkov image corresponding to the anatomical image in A). The gantry angle is at 290˚, and there is a clear dependence of Cherenkov signal on the gantry angle.

Fig 4. Example data from canine subject #1.

A) For the whole tumor ROI, the Cherenkov intensity was measured for each frame, normalized to 1/R² and plotted against time. The legend denotes fraction 1 (F1), subfractions 1–6 (S1-S6), and gantry angles 40° and 290° (G40 and G290). Over time, each of the frames is consistent, suggesting that average Cherenkov information for each subfraction is a good representative of the data. B) For canine 1, fraction 1, subfraction 1, gantry angle 40°, the ROIs are drawn in red around the whole tumor, normal tissue, tumor edge and tumor center. C) For the same dog, the averaged value for each subfraction is recorded with the standard error of the mean as the error bars. The vertical black bars denote different fractions on different days. D) Similarly, the Oxylite data is shown for both the tumor and normal tissue over the course of the entire treatment. Each data point represents the mean Oxylite value for a subfraction.

Fig 5. First fraction data for three canine subjects.

The Cherenkov intensity data (A, C, E) has been normalized to the first normal tissue subfraction. The intensity for the whole tumor, normal tissue, tumor edge and tumor center is shown, all for the same gantry angle. Using a one-way ANOVA that compared all groups followed by Tukey’s multiple comparisons test, normal tissue is consistently different than the remainder of the tumor areas, as is the tumor edge (* p<0.05, ** p<0.01, *** p<0.001, **** p<0.0001). The intensity difference between the adjacent subfractions as well as the difference between the first and last subfractions are shown (B, D, F). No significant difference between ΔS_X2-S_X1 was found nor are there any trends in the data. A) Normalized Cherenkov intensity for canine 1, fraction 1 B) Inter-subfraction difference for canine subject 1, fraction 1 C) Normalized Cherenkov intensity for canine subject 2, fraction 1 D) Inter-subfraction difference for canine subject 2, fraction 1 E) Normalized Cherenkov intensity for canine subject 3, fraction 1 F) Inter-subfraction difference for canine subject 3, fraction 1.

Fig 6. Summary of percent change from fraction baseline for included canine subjects.

For each fraction FN, the magnitude of the percent change was calculated for three canine subjects and combined into a single graph. A) The Oxylite data for normal tissue and tumor tissue are significantly different according to a two-tailed, paired T-test with p = 0.0039. Though the Oxylite data is only a point measurement, it is clear that tissue oxygen levels are changing more in the tumor than in the normal tissue. B) The Cherenkov data shows that the absolute percent change for the tumor edge is the largest. The normal tissue also changes from baseline while the whole tumor and tumor center does not change as much, relative to the other tissue areas. C) The coefficient of variation was calculated for each whole tumor over the course of the entire radiation plan for a single gantry angle (standard deviation / mean). Not only is there high pixel-pixel variation, but that variation changes for a single fraction.