Improving treatment geometries in total skin electron therapy: Experimental investigation of linac angles and floor scatter dose contributions using Cherenkov imaging

Abstract

Purpose:

The purpose of this study was to identify the optimal treatment geometry for total skin electron therapy (TSET) using a new optimization metric from Cherenkov image analysis, and to investigate the sensitivity of the Cherenkov imaging method to floor scatter effects in this unique treatment setup.

Methods:

Cherenkov imaging using an intensified charge coupled device (ICCD) was employed to measure the relative surface dose distribution as a 2D image in the total skin electron treatment plane. A 1.2m × 2.2m × 1 cm white polyethylene sheet was placed vertically at a source to surface distance (SSD) of 300cm, and irradiated with 6MeV high dose rate TSET beams. The linear accelerator coordinate system used stipulates 0° is the bottom of the gantry arc, and progresses counterclockwise so that gantry angle 270° produces a horizontal beam orthogonal to the treatment plane. First, all unique pairs of treatment beams were analyzed to determine the performance of the currently recommended symmetric treatment angles (±20° from the horizontal), compared to treatment geometries unconstrained to upholding gantry angle symmetry. This was performed on two medical linear accelerators (linacs). Second, the extent of the floor scatter contributions to measured surface dose at the extended SSD required for TSET were imaged using three gantry angles of incidence: 270° (horizontal), 253°(−17°), and 240°(−30°). Images of the surface dose profile at each angle were compared to the standard concrete floor when steel plates, polyvinyl chloride (PVC), and solid water were placed on the ground at the base of the treatment plane. Post processing of these images allowed for comparison of floor material-based scatter profiles with previously published simulation results.

Results:

Analysis of the symmetric treatment geometry (270°±20°) and the identified optimal treatment geometry (270°+23° and 270°−17°) showed a 16% increase of the 90% isodose area for the latter field pair on the first linac. The optimal asymmetric pair for the second linac (270°+25° and 270°−17°) provided a 52% increase in the 90% isodose area when compared to the symmetric geometry. Difference images between Cherenkov images captured with test materials (steel, PVC, and solid water) and the control (concrete floor) demonstrated relative changes in the 2D dose profile over a 1m x 1.9m region of interest (ROI) that were consistent with published simulation data. Qualitative observation of the residual images demonstrates localized increases and decreases with respect to the change in floor material and gantry angle. The most significant changes occurred when the beam was most directly impinging the floor (gantry angle 240°, horizontal−30°), where the PVC floor material decreased scatter dose by 1-3% in 7.2% of the total ROI area, and the steel plate increased scatter dose by 1-3% in 7.0% of the total ROI area.

Conclusions:

An updated Cherenkov imaging method identified asymmetric, machine-dependent TSET field angle pairs that provided much larger 90% isodose areas than the commonly adopted symmetric geometry suggested by Task Group 30 Report 23. A novel demonstration of scatter dose Cherenkov imaging in the TSET field was established.

1. INTRODUCTION

Total skin electron therapy (TSET) has long been an accepted form of palliative care to manage the cutaneous T-cell lymphoma known as mycosis fungoides.^1–6 The goal of the treatment is to deliver homogeneous superficial dose (to a depth of 1-5mm) to the entire patient skin surface, where the disease primarily manifests.⁷ In spite of the intention for dose uniformity, high deviations of in vivo dose distributions have been published. Anacak et al.⁸ report a mean deviation of 7.7 ± 7.4% in readings on the trunk, and a mean deviation of 19.7 ± 17.7% in extratrunk readings using a cohort of 67 TSET patients treated at a single institution, while noting similar deviations in patients treated with and without beam spoilers. While some variation in surface dose readings is expected, the extent of the variation can be minimized by ensuring the planar uniformity of the field before treatment.

Task Group 30 Report 23 (TG30) outlines the current recommendation by the American Association of Physicists in Medicine for commissioning and field dose parameters.⁹ The commonly adopted six position, dual field approach known as the modified Stanford technique places the patient on a raised treatment platform at extended source to surface distances between 3-8m, and irradiates with high dose rate electrons of energies between 4-12 MeV.^7–14 Using two fields (an upper and a lower beam) not only helps accommodate the large treatment area, but also aids in the reduction of x-ray contamination at the patient. A beam spoiler is sometimes introduced to degrade the beam energy to better match the required penetration depth at the patient, without affecting dose rate or levels of x-ray contamination,¹⁰ however there is some evidence that beam uniformity suffers.⁸

Typical dose measurement techniques used for TSET commissioning are limited to sparse point measurements using thermoluminescent dosimeters (TLDS), diodes, optically stimulated luminescent dosimeters (OSLDs), ionization chambers, or more regional 2D film measurements.^{15, 16} As a result of this, in conjunction with the large field sizes and extended source to surface distances (SSDs) required for this treatment, there are limited techniques to quickly evaluate the dose distribution in the treatment field as a whole at high resolution.

Previous work introduced a Cherenkov imaging method capable of capturing the full treatment region rapidly and accurately (Cherenkov intensity was linear with ionization chamber measurements with R²=0.93).¹⁷ This method utilizes an intensified charge coupled device (ICCD) camera to visualize surface dose on a large polyethylene sheet in the patient treatment plane, in addition to in vivo patient imaging during irradiation. The research presented herein aims at expanding the understanding of the TSET dose fields using similar Cherenkov imaging techniques, where the methods of analysis has been further developed and improved to make more extensive conclusions regarding optimal treatment setup.

Two main hypothesis are explored. The first hypothesis states that an asymmetric selection of gantry angles (unequal theta above and below the horizontal) for TSET setup provides a larger, more inclusive 90% isodose area than the recommended symmetric gantry angle pair (equal theta above and below the horizontal). This deviates from the original presentation of the Cherenkov imaging method, which evaluated treatment angle pairs based solely on the coefficient of variation within the test plane.¹⁷

In addition, interpretation of the dose distributions during TSET requires some consideration of dose from floor scatter.¹⁸ The second hypothesis states that the remote Cherenkov imaging method will exhibit sensitivity to dose deposited in the treatment plane as a result of the scattering interactions between the ionization radiation beam and the floor, which will be tested by introducing different scattering materials to the field and measuring differences between the collected Cherenkov images.

2. MATERIALS & METHODS

For this study, a Cherenkov imaging technique was adopted to image an analog for relative surface dose on a large 1.2m x 2.2m x 1cm white polyethylene sheet at the extended treatment plane used for TSET (SSD=300cm). Following the recommendation of previous publications,^{17, 19} an intensified charge coupled device (ICCD) camera (PIMAX4 1024i, Princeton Instruments, Trenton, NJ, USA), equipped with a 24mm Canon F1.8L lens, was placed approximately 4m away from the polyethylene sheet. No beam spoiler was used between the beam and the treatment plane, consistent with the adopted 6MeV TSET treatment protocol. As shown in Figure 1.a, Black-out fabric (Thorlabs, Newton, NJ, USA) was affixed to the posterior face of, as well as on the floor in front of, the plastic sheet to eliminate optical reflections on the surface of interest.

Fig. 1.

a) Experimental setup. b) Test image from camera: yellow box denotes size and shape of analyzed ROI in floor scatter experiment, and red line designates the height (30cm) of the wooden patient support structure used during TSET. c) PVC test material. d) Steel test material. e) Solid water test material. Each test material in c)-e) was covered with blackout fabric prior to Cherenkov imaging.

Cherenkov image acquisition was triggered to the linac radiation pulses using the current signal out of the high voltage power supply of the linac. A 3μs gate delay and a 5μs long gate window was used to time synchronize the ICCD acquisition with the radiation pulses. To ensure a strong signal to noise ratio, each image frame consisted of 500 radiation pulses (i.e. on-chip accumulations), which resulted in a readout frame rate of approximately 4 frames per second (fps).

Linac angular reference conventions can vary between models and clinics. In this work, the linac coordinate system used stipulates 0° is the bottom of the gantry arc with the beam pointed towards the ceiling, and progresses counterclockwise so that gantry angle 270° produces a horizontal beam orthogonal to the TSET treatment plane. Gantry angle combinations are listed in reference to the horizontal, 270°; angles greater than 270° are inclined towards the ceiling (upper treatment beam), and angles less than 270° are slanted towards the floor (lower treatment beam).

2.A. TSET Geometry Optimization Experiment Design

Building on the previously published technique in TSET geometry optimization using Cherenkov imaging,¹⁷ the ICCD setup described above was used to collect Cherenkov images corresponding to the 2D relative dose profile on the uniform polyethylene sheet placed at SSD=300cm, for 43 gantry angles (240°-260°, 270°, and 280°-300°), where 270 corresponds to the horizontal. The experiment was repeated on two linacs: 1) a Varian 2100CD series linac, and 2) a Varian Clinac iX Trilogy (Varian Medical Systems, Palo Alto, CA). The former was operated in high dose-rate TSET mode, delivering 6MeV electrons at dose rate 888 MU/min (calibrated at 3cGy/MU), and field size 36cm x 36cm at isocenter. The latter was operated in standard mode, delivering 6MeV electrons at dose rate 1000 MU/min (calibrated at 1cGy/MU), and field size 40cm x 40cm at isocenter.

Through post processing carried out in MATLAB (Mathworks, Natick, MA USA), all unique angle pair combinations were summed into composite images of relative surface dose, then quantitatively analyzed to determine the optimized treatment setup to provide the largest uniform field. The numerical analysis of the composite treatment images consisted of extracting the >90% isodose region of the image. In this context, 100% dose is the maximum intensity in the Cherenkov image, rather than the prescription point at the center of the field. Given that the Cherenkov image provides relative dosimetry, this distinction simplifies the comparison for analysis; any geometry producing a field with the prescription point outside of the 10% tolerance will be rejected after calculating the optimization metric α, to be defined later. The Cherenkov images can be renormalized after analysis to provide the 10% spread about the central prescription point.

The coefficient of variation (CoV, defined as the standard deviation divided by the mean), height (h), width (w), and total area of each >90% isodose region (A90) along the central axes, were each normalized by the best observed value of each measurement. These numbers were then summed into a single optimization metric, α, defined by the equation:

$$\alpha =\frac{{\mathit{CoV}}_{\mathit{measured}}}{{\mathit{CoV}}_{\mathit{min}}}+\frac{w_{\mathit{max}}}{w_{\mathit{measured}}}+\frac{h_{\mathit{max}}}{h_{\mathit{measured}}}+\frac{{A90}_{\mathit{max}}}{{A90}_{\mathit{measured}}}$$ (1)

Using this equation, the theoretical “best” treatment geometry would have α=4, the minimum possible value. In practice, the angle pair producing the smallest value of α is considered the optimized treatment geometry producing the largest, most uniform dose profile in the flat plane of the TSET setup.

2.B. Floor Scatter Experiment Design

To further understand the results of the TSET geometry optimization, an experiment was designed to test the sensitivity of the Cherenkov imaging technique to radiation dose from floor scatter. By imaging the 2D profile of surface dose on the plastic sheet with various materials on the floor at the foot of the polyethylene, the differences in relative dose observed could be inferred as differences in floor scatter.

Three gantry angles were tested, with increasing incidence of floor scatter: 270° (horizontal), 253°, and 240°. Each test case was imaged while irradiated by 250 monitor units (Mus) of the 6MeV high dose rate TSET beam; 9MeV electrons were not available for testing. In addition to the change in gantry angle, three unique materials, shown in Fig. 1.c–e, were placed at the foot of the vertical polyethylene sheet to be compared to the usual concrete floor: polyvinylchloride (PVC), solid water, and steel. Blackout fabric was used to cover the floor and material sample, to eliminate signal or optical reflections from the changing floor materials. The thicknesses of the three test materials were selected to exceed the 50% dose deposition range (R50) of 6MeV electron beam of each respective material at isocenter, and the final dimensions of each material are listed in Table 1. Values of R50 were calculated based on the R50 of water (2.2cm²⁰)

Table 1.

List of properties of floor scatter materials tested.

Material	Density (g/cm3)	R50 (cm)	Length (cm)	Width (cm)	Thickness (cm)
PVC	1.4121	1.6	121.92	60.96	3.81
Solid Water	1.04	2.1	60	30	3
Steel	7.6822	0.3	60.96	60.96	0.318

Post processing of the images was performed in MATLAB. Each test case produced 31 images as a result of the 250 MUs delivered. To eliminate random salt-and-pepper noise caused by stray radiation in the room hitting the camera, a temporal median filter was performed on each set of 5 consecutive images, resulting in a final stack of 27 images for each gantry angle and floor material combination. Background image subtraction was performed on each image to eliminate any influence of ambient light (minimal, given all images were collected with room lights off and instrumentation lights on the linac and couch were covered). The mean pixel value at each location was then used to generate a single frame of data for each test case. Each image was then normalized to the maximum pixel value observed in all of the data sets combined, so that relative comparison would be possible.

To visualize the changes in radiation floor scatter due to the modification of floor material, residual images were calculated by subtracting the control image (concrete floor covered in blackout fabric) from the experimental image (solid water, PVC, and steel, each covered in blackout fabric). An 11x11 pixel averaging kernel (corresponding to a roughly 2.5cm x 2.5cm area) was then applied to each residual image to smooth the results, and eliminate the influence of noise.

3. RESULTS

3.A. TSET Geometry Optimization Results

The numerical TSET optimization described above was performed on the composite Cherenkov images using two criteria. First, limiting the search to symmetric angle pairs that have an equal angle above and below the horizontal gantry position (270°). Second, this restriction was lifted and all possible angle pairs were analyzed.

The two optimized results are shown in Figure 2. By adopting an asymmetric angle pair (270°+23° and 270°−17°), the area of the 90% isodose region shown in Fig.2.b increased by 16% for the first machine tested (Varian 2100CD), to the area presented in Fig.2.d; at the midline axes, this corresponds to a 90% isodose region that is approximately 3cm wider, and 16 cm taller. The results were even more pronounced for the second machine (Varian Clinac iX Trilogy); the 90% isodose region grew by approximately 52% by adopting an asymmetric angle pair (Fig.2.h, 270°+25° and 270°−17°) versus a symmetric angle pair (Fig.2.f), as a result of the 47cm increase in the height at midline, in spite of the 12cm of width lost at midline where the profile exhibits an hourglass shape.

Fig. 2.

Composite images of optimized TSET geometry: a) on linac 1, restrained to symmetric gantry angles (270°±20°); b) 90% isodose area of (a); c) on linac 1, asymmetric gantry angles 270°+23° and 270°−17°; d) 90% isodose area of (c); e) on linac 2, restrained to symmetric gantry angles (270°±20°); f) 90% isodose area of (e); g) on linac 2, asymmetric gantry angles 270°+25° and 270°−17°; h) 90% isodose area of (g).

3.B. Floor Scatter Experiment Results

The results of the floor scatter experiment are shown in Figure 3 for visual interpretation. The first row displays the residual images (concrete image control subtracted from the test material image) corresponding to a gantry angle of 270°, where the primary beam is parallel to the floor. The second row contains the residual images for gantry angle 253°, where the field light for the radiation beam is completely on the white polyethylene sheet, but near the floor. The final row shows the residual images for gantry angle 240°, where the field light projects the most appreciably onto the floor. The residual images for each test scatter material are grouped into columns, arranged in order of increasing material density: solid water, PVC, and steel. All images are presented on the same color scale, and were calculated using the same normalization factor.

Fig. 3.

Residual images of each test material compared to concrete. The three gantry angles tested are shown in each row of images, with increasing incidence of the primary beam on the floor. The three test materials are grouped into the labeled columns in order of increasing material density (solid water, PVC, steel).

The box and whisker plots in Figure 4 expression present more directly the quantitative characteristics of the residual images in Figure 3, but eliminate the spatial information afforded by the 2D plots. The maximum change in percent dose was observed for the case where the primary beam was directed most obliquely at the floor (240°); peak single-pixel differences were +3.3% (maximum), −0.91% (minimum) and −2.3% (minimum) for steel, PVC and solid water, respectively. However, when averaging the changes across the vertical axis (±10cm from the central line) as shown in Figure 5, the magnitude of the differences decreases to roughly 2.4%, 0% and −1.6%, respectively.

Fig.4.

Box plots showing distributions of the normalized dose differences from the residual images in Figure 3 for: a) gantry angle 270°; b) gantry angle 253°; c) gantry angle 240°. The whisker lengths correspond to 1.5 times the interquartile range, and outliers, as the primary points of interest, are marked in red.

Fig. 5.

Averaged vertical profile of change in relative dose for gantry angle 240°.

4. DISCUSSION

4.A. TSET Geometry Optimization Discussion

The accuracy of the Cherenkov-based method for analyzing TSET surface dose has been previously studied and reported as linear with ionization chamber measurements (R²=0.93).¹⁷ The innovation of this approach for optimizing TSET geometry lies in calculation of a novel optimization metric, focused on maximizing the 90% isodose area, for quantitative comparison of each field angle pair. Establishing the ROI to the 90% isodose area was intended to draw parallels to the TG-30 TSET commissioning guidelines described briefly below.

Previous work, limited to analyzing coefficient of variation only, came to the conclusion that the homogeneity of the treatment field was approximately the same when optimizing the fields between symmetric gantry angle pair and asymmetric gantry angle pairs.¹⁷ This deeper investigation to new data, similarly acquired and corroborated at a second institution, incorporates the features of the 90% isodose field height, width, and area. Nontrivial gains in the 90% isodose region area were accomplished (16% and 52% for the two linacs at two separate institutions) when adopting an asymmetric gantry angle pair.

When confined to the principle of a symmetric treatment geometry, the results of the expanded Cherenkov image-based TSET optimization method shown in Fig.2 agree with the suggested angles for TSET outlined in TG-30: at SSD=300cm, employ angles that are ±20° from the horizontal.⁹ However, the new Cherenkov imaging optimization method allows for the rapid measurement of relative treatment field dosimetry from many gantry angle pairs; measuring many pairs of gantry angles using ionization chambers, films, thermoluminescent dosimeters, or diode-based dosimetry devices would be extremely time consuming. Analysis of all gantry angle pairs, irrespective of angle symmetry, supports the hypothesis that an asymmetric treatment geometry provides a larger 90% isodose region than currently adopted symmetric treatment angles; the optimized asymmetric pair increased the area 90% isodose region by over 16% for one machine, and a striking 52% for a second machine.

In practice TG-30 recommends “a vertical uniformity of ± 8% and a horizontal uniformity of ± 4% over the central 160 cm x 60 cm area of the treatment plane.”⁹ This means most commissioning takes place by assessing points only along the vertical and horizontal axes about the prescription point at midline; very little is done to investigate the integrated shape of the fields, despite the fact that the patient is treated in all four quadrants beyond these two orthogonal axes. An additional point of concern is that the defined ROI height is less than the average height of an adult male in the United States (176cm),²³ and the 60cm width scarcely covers the breadth of an average torso, particularly neglecting the wide spread of patient extremities whilst standing in the six positions required of the typical modified Stanford technique.

Often the question arises regarding how a flat treatment plane translates to uniform dose when the realities of curved patient anatomical features come into play. Given that TSET treatment itself is generic, with only clinically managed regional dose boosts prescribed at the physician’s discretion, it stands to reason that providing the largest isodose area in the flat treatment plane would translate to the most uniform surface dose to the average patient. This concept is the premise of the TG-30 requirements already quoted above, and is the immediately achievable goal when simulation CT scans or other patient surface information is not considered for patient-specific treatment planning. Methods for rapidly assessing the entire flat treatment plane, such as the Cherenkov-imaging based method used here, could help TSET geometries adopted clinically improve dose homogeneity delivered to patients.

The clinical implications of the hourglass shape in the optimized asymmetric gantry angle pair, particularly on the second linac tested, need further study. The six patient positions during treatment establish a complex aggregate of dose that is influenced by inter-fraction setup variations, and more broadly by disparity in gross anatomy between patients. However, the hourglass isodose region is logically matched to the patient positions where the arms are raised and the legs are spread for even dose coverage. In addition, given that TSET response is clinically managed with compensating boost fields when necessary, the hourglass shape is not expected to incur negative overall consequences.

A final interpretation of this data, particularly in the notable differences in the 90% isodose field shapes in Figure 2 between the two linacs tested, is that the linac models and treatment setups, down to room geometries and construction materials, can affect the ultimate shapes of the composite treatment fields. The flatness and symmetry of beams from of each machine were independently verified through traditional means as well as through Cherenkov imaging, so these differences cannot be written off as irregularities in the machine output. This reinforces the notion that there is not a one-size-fits-all solution, even if translating between two linacs and bunkers within the same clinic. Each commissioned setup must be rigorously tested under the same conditions as patient treatment to ensure the protocol adequately achieves the dosimetric goals of TG-30.

4.B. Floor Scatter Experiment Discussion

Definite qualitative changes in the relative dose pattern were observed when the floor material was altered to solid water, PVC or steel for gantry angles 253° and 240°, with minor changes between the three test materials at 270°, as shown in Figure 3. The scatter dose from these fields would be the result of inelastic processes with the floor material, producing a spectrum of lower energy electrons deflected towards the treatment plane.

With that in mind, these changes reflect the expectation that the dense, high-Z material of the steel plate would have more near-surface scatter events, and thus increase scatter in the orthogonally arranged treatment plane. The incoming electrons would stochastically travel deeper into the lower density, low-Z materials solid water and PVC before scattering, and would be more likely to be absorbed after scattering. Figure 4 allows for further interpretation to how the materials and angles affect the differences in observed dose, in that the whiskers and outliers to the residual image pixel values follow the expected trends towards hot or cold, depending on scattering material.

The Monte Carlo simulation work of Nevelsky et al.¹⁸ reported peak floor scatter contributions at 20cm above the floor to be +3%, −3% and −5% for iron, PVC and water, compared to concrete, given a SSD=400cm and gantry angle 17° above the horizontal. These magnitudes are larger than what was measured with the Cherenkov imaging method (+2.4%, 0.0%, −1.6%), even with a 30° change in gantry angle at the shorter SSD (300cm versus 400cm). This suggests that there is a small amount of scatter dose which is not detected via Cherenkov imaging, most likely due to the energy dependence of the optical emission.

Nevelsky et al. stated “the spectrum of the scattered electrons had a distribution which was almost uniform between a few hundred keV to 4 MeV, then decreased linearly to 6 MeV.”¹⁸ While most of these electrons should be above the threshold for Cherenkov light generation, the energy-dependence of the number of photons emitted would deflate the contributions of lower energy scattered electrons. For this reason, the Cherenkov imaging method provides a novel qualitative presentation of scatter dose distribution that follow anticipated and simulated trends, however the accuracy of the quantitative results are limited to within a few percent, depending on the material (and subsequently the energy spectrum of the scattered electrons).

It is worth noting that although the Cherenkov imaging method is less sensitive to lower energy electrons, these electrons would be responsible only for very superficial dose (which is the intent of the treatment). Moreover, the simulated scatter doses reported become trivial at a height of 50cm from the floor, and the patient is typically positioned on a wooden platform 20-30cm above the concrete, further diminishing the patients’ exposures to the scatter.

Taking into consideration the already broad ranges of dose reported with in vivo dosimetry (±15%),⁸ it is not expected that the menial scatter dose magnitudes¹⁸ are the primary cause of heterogeneities in dose received by the patient. Rather, these dose variations could be an indication of a possible weakness in the treatment field setup, in conjunction with patient positioning, which limits the achievement of appropriate dose coverage. Therefore, techniques to optimize the size and uniformity, such as the Cherenkov imaging method described herein, could be useful tools in evaluating and improving current clinical TSET setups.

This work also illustrates the importance of careful consideration of material selection in patient support structure construction. In TSET delivery, whether using a modified Stanford technique or a rotation technique, introducing metal components to the patient support structure could increase scatter dose, as was seen in this work with the steel plate. This also suggests that it would be ideal to perform setup dosimetry measurements within the confines of the support structure, to ensure scatter dose can be captured in the readings.

5. CONCLUSIONS

The results presented support the two tested hypotheses. First, an expanded method of Cherenkov image analysis showed that an asymmetric selection of gantry angles for TSET setup provided an appreciably larger, more inclusive 90% isodose area than the recommended symmetric gantry angle pair (270°±20°). Second, the Cherenkov imaging method was shown to exhibit sensitivity to dose from floor scatter during TSET.