Direct measurement and correction of both megavoltage and kilovoltage scattered x‐rays for orthogonal kilovoltage imaging subsystems with dual flat panel detectors

Abstract Purpose To measure the scattered x‐rays of megavoltage (MV) and kilovoltage (kV) beams (MV scatter and kV scatter, respectively) on the orthogonal kV imaging subsystems of Vero4DRT. Methods Images containing MV‐ and kV‐scatter from another source only (i.e., MV‐ and kV‐scatter maps) were acquired for each investigated flat panel detector. The reference scatterer was a water‐equivalent cuboid phantom. The maps were acquired by changing one of the following parameters from the reference conditions while keeping the others fixed: field size: 10.0 × 10.0 cm2; dose rate: 400 MU/min; gantry and ring angles: 0°; kV collimator aperture size at isocenter: 10.0 × 10.0 cm2: tube voltage: 110 kV; and exposure: 0.8 mAs. The average pixel values of MV‐ and kV‐scatter (i.e., the MV‐ and kV‐scatter values) at the center of each map were calculated and normalized to the MV‐scatter value under the reference conditions (MV‐ and kV‐scatter value factor, respectively). In addition, an MV‐ and kV‐scatter correction experiment with intensity‐modulated beams was performed using a phantom with four gold markers (GMs). The ratios between the intensities of the GMs and those of their surroundings were calculated. Results The measurements showed a strong dependency of the MV‐scatter on the field size and dose rate. The maximum MV‐scatter value factors were 2.0 at a field size of 15.0 × 15.0 cm2 and 2.5 at a dose rate of 500 MU/min. The maximum kV‐scatter value was 0.48 with a fully open kV collimator aperture. In the phantom experiment, the intensity ratios of kV images with MV‐ and kV‐scatter were decreased from the reference ones. After correction of kV‐scatter only, MV‐scatter only, and both MV‐ and kV‐scatter, the intensity ratios gradually improved. Conclusions MV‐ and kV‐scatter could be corrected by subtracting the scatter maps from the projections, and the correction improved the intensity ratios of the GMs.

During beam irradiation, the target is tracked in real time by IR markers at four or five positions and a preconstructed correlational model between the IR marker positions and the three-dimensional (3D) positions of the target, which are indicated by radiopaque markers. [13][14][15] The predicted 3D target position is the average of the 3D target positions calculated from the IR markers. This approach also involves a four-dimensional (4D) model, which is a quadratic polynomial equation. Two to four gold sphere markers (Olympus, Tokyo, Japan) and one flexible linear marker Visicoil (IBA dosimetry, Louvain-la-neuve, Belgium) are used for lung cancer and liver and pancreatic cancer, respectively. [7][8][9][10] Just before the first beam irradiation on each treatment day, the IR markers are monitored for 20 s, while the IR camera and radiopaque markers are detected by calculating the ratio between the intensity of the radiopaque marker and that of its surroundings in two orthogonal kV images simultaneously. 16 The detected two-dimensional (2D) radiopaque marker positions are converted into 3D positions using predefined camera parameters. 17 Then, a 4D model is constructed by fitting datasets representing the IR marker and radiopaque marker positions into the equation.
At the time of beam irradiation, the 3D position of the marker is predicted by the IR marker position with the 4D model after 25 ms.
Then, the gimbaled x-ray head swings to the predicted position. 18 During beam irradiation, the radiopaque markers are also detected on two orthogonal kV images in 1 s intervals. Both kV subsystems are always turned on during DTT treatment. The detected 2D positions are converted into 3D positions, and the differences between the converted and predicted 3D positions are calculated, using those positions visualized based on the concurrent kV images. If the difference exceeds a tolerance depending on the breathing pattern of the patient, Vero4DRT interrupts the MV beam irradiation automatically.
In addition, the converted 3D radiopaque marker data can be used to rebuild a 4D model if necessary on the treatment day.
As mentioned above, IR marker-based DTT treatment requires concurrent kV imaging during MV beam irradiation. The concurrent kV images consist of primary and scattered kV x rays in addition to scattered x rays from the MV beam (MV-scatter) and from the kV beam irradiated by the other kV source (kV-scatter), which are scattered by the body of the patient. Thereby, the image contrast of the concurrent kV images is degraded by the MV-and kV-scatter. Image contrast degradation can cause detection errors or lack of detection of the radiopaque marker, which leads to failure of the auto beam-off system and online rebuilding of the 4D model. Figure 1 shows an example of the treatment console for DTT treatment in Vero4DRT. In Fig. 1(b), the concurrent kV images are degraded by MV-and kV-scatter. Thereby, the absolute difference between the detected and predicted radiopaque marker positions was not calculated, and MV-and kV-scatter in the Vero4DRT system have not been well investigated.
Thus, the objectives of this study were to measure and quantify the MV-and kV-scatter of two orthogonal kV imaging subsystems directly under various MV and kV beam parameters and to demonstrate MVand kV-scatter correction to improve radiopaque marker detection in a phantom study with intensity-modulated beam irradiation 2 | MATERIALS AND METHODS

2.A. | Vero4DRT specifications
The gantry of Vero4DRT can rotate ±185°around the lengthwise axis of the patient couch (gantry rotation) and ±60°around the vertical axis

2.B.2. | MV-scatter map acquisition with various MV beam parameters
Since both kV subsystems are always turned on during DTT treatment, the kV collimators were closed and two Pb plates (1 cm thickness) were deposited at the exits of each kV source to shield leaked kV x rays to acquire an image containing MV-scatter only, that is, an MV-scatter map. Ten orthogonal kV images were acquired by each kV imaging subsystem at a frame rate of 1 fps during MV beam irradiation. Those images were averaged, and the averaged image was used as the MV-scatter map.
Each MV beam parameter was varied from its reference value while the other parameters were fixed to assess the dependency of each parameter, including the field size, dose rate, gantry angle, and ring angle. The details of the parametric variation are shown in Table 1 Herein, we call the relative MV-scatter value the "MV-scatter value factor." F I G . 1. Example of treatment console during dynamic tumor tracking treatment using Vero4DRT. The projection images from flat panel detectors1 (FPD1) and FPD2, kV x-ray and tracking parameters, infrared reflective marker motion, and absolute difference between the detected and predicted radiopaque marker positions are displayed. (a) If the radiopaque marker is detected without problems, the absolute difference can be calculated and shown. (b) If not, the absolute difference cannot be calculated and shown on the console (orange arrows). In addition, the contrast of the radiopaque marker is degraded by noise, especially in the image from FPD2.

2.B.3. | kV-scatter map acquisition with various kV beam parameters
To acquire an FPD1 (or FPD2) image containing kV-scatter from kV Source 2 (or 1), the kV collimators of kV Source 1 (or 2) were closed and the Pb plate was deposited at the exit of kV Source 1 (or 2).
Ten orthogonal kV images were acquired by each kV imaging subsystem at a frame rate of 1 fps without MV beam irradiation. The images from FPD1 (or FPD2) were averaged, and the averaged image was used as the kV-scatter map for FPD1 (or FPD2). As in the procedure for the MV-scatter maps, an ROI of 100 × 100 pixels at the center of each kV-scatter map was defined, and the pixel values in that ROI were averaged (kV-scatter value). Thereafter, the kVscatter values were normalized to that obtained with the reference MV beam parameters (a field size of 10.0 × 10.0 cm 2 , a dose rate of 400 MU/min, and gantry and ring angles of 0°). Herein, we refer to the relative kV-scatter value as the "kV-scatter value factor." Each kV beam parameter was varied from its reference value while the other parameters were fixed to assess the dependency of each parameter, including the tube voltage, exposure, kV collimator aperture size, gantry angle, and ring angle. The details of the parametric variations are shown in Table 2.   Six intensity-modulated MV beams were used for the experiment.

3.B.2. | Dose rate dependency
The dose rate dependencies with various field sizes for FPD1 and

3.C.3. | Gantry and ring angle dependency
The gantry and ring angle dependencies of FPD1 and FPD2 are shown in Figs. 6(g) and 6(h), respectively. As in the MV-scatter case, the trend of FPD2 is symmetrical to that of FPD1 with respect to 180°rotation. The maximum difference between the MV-scatter values at ring angles of 0°and ±20°is 0.07.

3.D. | MV-and kV-scatter correction experiment using intensity-modulated beams
The kV only, MV + kV, kVScorr, MVScorr, and MVkVScorr images for Thus, the kV-scatter values were smaller than the MV-scatter values.
The kV-scatter value factor was 0.12 under the reference kV beam conditions. The dominant kV beam parameter in determining the kVscatter was the kV collimator aperture size.
For the MV-and kV-scatter correction experiment, a clinical treatment scenario was assumed. As shown in Fig. 7, a larger field size (segment 11) yielded more MV-scatter in the MV + kV images, as supported by the direct measurements (Fig. 5). In addition, not only MV-scatter correction but also kV-scatter correction is necessary since the intensity ratio was improved in the kVScorr images, even though the amount of kV-scatter was less than that of MVscatter.
According to the obtained results, MV-and kV-scatter have greater effects on monitoring images during DTT treatment when the field and kV collimator aperture are large and the dose rate is high. The monitoring images of a patient who has a large target or an implanted marker movement are potentially affected by MV-and kV-scatter, as the field size or kV collimator aperture size may be large. Thus, care must be taken in DTT treatment, particularly that for pancreatic cancer, because the field size and kV collimator aperture size in pancreatic cancer treatment are larger than those in lung or liver cancer treatment.
Degradation of the intensity ratio in concurrent orthogonal kV images for monitoring in terms of both MV-and kV-scatter would cause failure of the automatic beam-off system, which is based on the detected radiopaque marker positions in the monitoring images.
In case the marker positions cannot be detected in the monitoring image, the automatic beam-off system is turned off, and an operator