Real-time Cherenkov emission portal imaging during CyberKnife® radiotherapy

The feasibility of real-time portal imaging during radiation therapy, through the Cherenkov emission (CE) effect is investigated via a medical linear accelerator (CyberKnife®) irradiating a partially-filled water tank with a 60 mm circular beam. A graticule of lead/plywood and a number of tissue equivalent materials were alternatively placed at the beam entrance face while the induced CE at the exit face was imaged using a gated electron-multiplying-intensified-charged-coupled device (emICCD) for both stationary and dynamic scenarios. This was replicated on an Elekta Synergy® linear accelerator with portal images acquired using the iViewGT™ system. Profiles across the acquired portal images were analysed to reveal the potential resolution and contrast limits of this novel CE based portal imaging technique and compared against the current standard. The CE resolution study revealed that using the lead/plywood graticule, separations down to 3.4  ±  0.5 mm can be resolved. A 28 mm thick tissue-equivalent rod with electron density of 1.69 relative to water demonstrated a CE contrast of 15% through air and 14% through water sections, as compared to a corresponding contrast of 19% and 12% using the iViewGT™ system. For dynamic scenarios, video rate imaging with 30 frames per second was achieved. It is demonstrated that CE-based portal imaging is feasible to identify both stationary and dynamic objects within a CyberKnife® radiotherapy treatment field.


Introduction
Cherenkov radiation is emitted when charged particles travel faster than the speed of light in a given dielectric medium (Cerenkov 1934). During x-ray radiotherapy, secondary electrons are predominantly generated from Compton scattering, which have sufficient energy to produce Cherenkov emission (CE) in water or tissue. CE has been shown to be measurable with an optical imaging system, given that the main wavelength of emission is within the visible spectrum and shown to be proportional to radiation doses deposited in the medium .
The incorporation of Cherenkov emission in electronic portal imaging device (EPID) detectors has recently been investigated (Mei et al 2006), while Cherenkov radiation imaging has been utilised as a dosimetry technique during photon (Jang et al 2012, Zhang et al 2013c, electron , Helo et al 2014b and proton (Helo et al 2014a) radiotherapy. Furthermore, real-time superficial dosimetry has been demonstrated in both animal and human studies , Jarvis et al 2014.
In radiotherapy, EPID-based portal imaging is often employed for pre-treatment patient positioning or intra-fraction tumour tracking. While some commercial linear accelerators incorporate portal imaging equipment, the CyberKnife ® system does not offer such an option and therefore alternative techniques are desirable. However, due to the fact that the CyberKnife ® system utilises multiple non-coplanar and non-isocentric beams, a conventional portal imaging device would not be practicable.
In this study the feasibility of real-time CE portal imaging during CyberKnife ® radiotherapy is investigated. This would enable real-time verification of in-patient targeting and treatment delivery accuracy.

Methods
A medical linear accelerator (CyberKnife ® , Accuray, Sunnyvale, CA) irradiated a partially filled water tank (300 × 300 × 350 mm) using a 60 mm diameter circular beam of 6 MV nominal energy at 800 mm source-axis-distance (SAD) with a dose rate of 10 Gy min −1 at a depth of 15 mm. A number of attenuating materials were placed on a moving platform between the beam source and the water tank (figure 1(a)), while an opaque 10 mm thick white solid water slab (Solid Water HE, Gammex Inc, Middleton, WI) was placed at the beam exit face to ensure that only the CE due to radiation at the exit face is imaged. The use of a water slab is in this case (due to the transparent nature of water) critical to ensure that the imaged CE as emitted by the x-ray beam is that of the exit beam. A gated electron-multiplying-intensified-chargedcoupled device (emICCD) (PI-MAX4: 512 EM, Princeton Instruments) with a commercial lens (Canon EF 135 mm f/2L USM) positioned at the exit face at a small incident angle (~20° to reduce direct radiation to the camera as it was not shielded) imaged the induced CE at the surface of the solid water slab. Image acquisition was synchronised with the beam pulses using the trigger signal from the accelerator in order to maximise the signal-to-background ratio (Glaser et al 2012).
A number of attenuating materials were used to assess the resolution and contrast of the measured CE at the exit face. For assessment of the resolution a graticule was constructed from a number of lead sheets (2 × 60 × 20 mm) spaced at regular intervals using plywood separators, figure 1(b). Note that the two lead markers at either end contained two sheets, and therefore represent a thickness of 4 mm. To assess the contrast of the measured CE a number of 28 mm diameter tissue equivalent rods (RMI 467 phantom, Gammex Inc, Middleton, WI) were used, figure 1(d).
The water tank was filled with tap water to a depth of 110 mm, and the phantoms were placed such that the lower 50% of the radiation beam travelled through the phantom and the water, whereas the upper 50% of the beam travelled through the phantom and air, before being measured at the exit face through the solid water slab. CE images were acquired at 30 frames per second (fps) with each frame accumulating CE for 5 radiation pulses. All data were background-subtracted using images with no radiation. The background-subtracted images (30 frames for static and 3 frames for the dynamic) were then median-filtered to produce the CE images for data analysis.
CE-based static and dynamic portal images were visually inspected for an initial proofof-concept validation. Resolution and contrast were assessed by analysing the profile plots through the two regions of the static images (air and water), facilitating quantitative evaluation. Resolution was quantified by measuring the distances of the minima in the profile plots, associated with the position of the lead sheets. To measure relative contrast, the mean percentage decrease of signal in the profile plots due to the presence of the tissue equivalent rods was calculated with respect to the normalised maximum. Specifically, the mean values for a width of 10 mm centred at the peak/trough of the maximum and minimum values were used.  The experimental configuration for contrast assessment in static configuration was replicated on an Elekta Synergy ® (Elekta AB, Stockholm, Sweden) linear accelerator with portal images being acquired using the iViewGT ™ system, to allow direct comparison of CE-based portal imaging against current technology. For this experiment a 240 × 240 mm square beam of 6 MV nominal energy at a dose rate of 3.6 Gy min −1 at 1000 mm SAD was employed, with portal images captured for 50 monitor unit (or 0.33 Gy to 1000 mm SAD) acquisitions.

Results
A Cherenkov image using the stationary graticule is shown in figure 2 along with profile plots of the measured CE intensity. The intensity profiles through both air and water are selfnormalised to allow quantitative analysis of the detected resolution, corresponding to the distance between each lead sheet, figures 2(b) and (c) and table 1. The lowest separation of 3.40 mm (spacing 'v' in figure 1(c)) is clearly evident in the portal image (figure 2(a)) and is measured as 3.58 mm and 3.92 mm for radiation travelling through air and water, respectively. Larger errors are observed at the two edges of the detected signal, which is primarily due to data capture geometry and beam divergence. Each lead sheet with a width of 2 mm is visible (except those at the edges which have a width of 4 mm), highlighting the potentials of higher resolution imaging.
Portal images in the presence of tissue equivalent rods were analysed and the intensity profiles used to assess the measurable relative contrast for each phantom. Figure 3 shows example portal images of the 'SB3 Cortical Bone' phantom as well as the corresponding selfnormalised profile plots.  The use of tissue equivalent phantoms demonstrates that CE emission detected at the exit face is sensitive to small contrasts often seen in biological tissue. As summarised in table 2, for tissue equivalent rods with 28 mm diameter, a contrast in CE of up to 15% was observed with an electron density of 1.69 relative to water, with similar contrast seen for beam passing  through the air and water-filled part of the water tank. On the other hand, EPID based portal images revealed higher contrast for beam travelling through the air part of the water tank, with lower contrast for beam travelling through the water-filled part. Real-time video (30 fps) CE imaging of a moving graticule and a tissue equivalent rod can be seen in the supplementary material (stacks.iop.org/PMB/60/N419/mmedia), with a snapshot shown in figure 4. Even though of inferior image quality due to the substantially lower number of frames in each median-filtered stack, the periodic movement (0.25 Hz) of the phantoms can be observed.

Discussion and conclusions
The use of CE for entrance dose measurements have previously been demonstrated showing a direct correlation between measured CE intensity and applied dose. However the use of CE based portal imaging, directly from the exit face of an irradiated medium has not been previously examined.
The use of CE based portal imaging has been demonstrated here, using both a highly attenuating resolution phantom as well as tissue mimicking phantoms. It has been shown that objects with as little as 3.4 mm separation are easily detected, while tissues with relative electron density of greater than 4% (with respect to water) demonstrate CE contrast of 8% or greater. Although the data from the CE based portal images show more noise, this can be overcome through the use of more optimised imaging setup, including the number of measured frames, a more adequate lens and noise removing image processing. Video rate CE based portal imaging was achieved to monitor movements of the graticule and tissue equivalent phantom in real-time. Therefore, CE based portal imaging, a novel imaging technique during radiation therapy, can be potentially applied for systems that lack alternative options such as EPID.
Contrary to EPID-based portal imaging, CE portal imaging only utilizes visible optical photons produced at the exit surface through the CE effect. This minimizes the contribution of scattered radiation (where the scattered photons are at lower energy so are less efficient at producing CE than the transmitted 'primary' beam), resulting in higher measured contrast when radiation passes through water or tissue (as compared to EPID-based portal image), which highlights for the first time the additional benefits of CE portal imaging over EPIDbased techniques.