Evaluation of a surface imaging system's isocenter calibration methods

Abstract AlignRT is a surface imaging system that has been utilized for localizing and tracking patient position during radiotherapy. AlignRT has two calibration procedures that can set the system's isocenter called “Monthly Calibration” (MC) and “Isocentre Calibration” (IC). The MC utilizes a calibration plate. In addition to the calibration plate, the IC utilizes a cubic phantom that is imaged with the linac treatment beam to aid in aligning the AlignRT and treatment‐beam isocenters. This work evaluated the effects of misaligning the calibration plate during the calibration process. The plate was intentionally shifted away from isocenter ±3.0 mm in the longitudinal and lateral directions and ±1.0 mm in the longitudinal, lateral, and vertical directions. A mock stereotactic radiosurgery (SRS) treatment was used to evaluate the effects of the miscalibrations. An anthropomorphic head phantom was placed in an SRS treatment position and monitored with the AlignRT system. The AlignRT‐indicated offsets were recorded at 270°, 315°, 0°, 45°, and 90° couch angles for each intentional misalignment of the calibration plate during the MC. The IC was also performed after each miscalibration, and the measurements were repeated and compared to the previous results. With intentional longitudinal and lateral shifts of ±3.0 mm and ±1.0 mm of the calibration plate, the average indicated offsets at couch rotations of ±90° were 4.3 mm and 1.6 mm, respectively. This was in agreement with the theoretical offset of √2*(shift‐of‐the‐calibration plate). Since vertical shifts were along the rotation axis of the couch, these shifts had little effect on the offsets with changing couch angle. When the IC was applied, the indicated offsets were all within 0.5 mm for all couch angles for each of the miscalibrations. These offsets were in agreement with the known magnitude of couch walkout. The IC method effectively removes the potential miscalibration artifacts of the MC method due to misalignments of the calibration plate.


| INTRODUCTION
AlignRT (Vision RT Ltd, London, UK) is a 3-camera, non-invasive, nonradiographic, optical surface imaging system that provides the user with translational and rotational offsets from a reference surface as well as the total displacement (i.e., the vector sum of the translations). 1 The reference surface can be obtained in two ways, either from a computed tomography (CT)-defined body contour that has been imported from the treatment planning system (TPS) or from a reference image that is captured by the AlignRT system at the time of treatment when the patient is in the final treatment position.
AlignRT has two calibration procedures that can set the imaging system's isocenter. These are called "monthly calibration" (MC) and "isocentre calibration" (IC). The MC procedure involves placing a calibration plate centered at the linac isocenter to set the imaging system's isocenter. The IC procedure involves imaging a cubic phantom with implanted ceramic spheres with the linac treatment beam to aid in aligning the imaging system's isocenter with the treatment beam's isocenter.
AlignRT has been utilized for a number of treatment sites, including: breast, 2-5 extremities, 6 head and neck, 7,8 and frameless stereotactic radiosurgery (SRS). 9,10 Due to the tighter tolerances typically required for SRS treatments and the use of couch rotations, this work focused on the SRS treatment procedure. However, the results are applicable to other treatment sites.
When using AlignRT with SRS, the patient is typically immobilized in an open-face mask. Prior to treatment, a region-of-interest (ROI) is defined consisting of the patient's face. The typical workflow involves: initially aligning the patient based on room lasers, using AlignRT to finely adjust the patient position (translations and rotations) based on the reference CT-defined body contour, using radiographic analysis (e.g., orthogonal kV image pair and CBCT) to shift the patient into the final treatment position, capturing a new reference surface with the AlignRT system, and monitoring the intrafractional motion during treatment based on this new reference surface. 9,10 For treatment plans that utilize couch rotations, like SRS, the planned couch angles are available in a dropdown list in the AlignRT user interface. Selecting a different couch angle will rotate the reference surface relative to the AlignRT isocenter and allow tracking the patient at these positions. However, misalignments of the linac treatment beam and AlignRT isocenters may propagate as falsely indicated offsets with the AlignRT system when couch rotations are used. This work investigated the effects of these potential misalignments between the treatment beam isocenter and the AlignRT isocenter for each of the isocenter calibration methods.

2.A | Calibration procedures
The MC procedure utilizes a calibration plate that is provided by the manufacturer (see Fig. 1). The plate consists of a 2D array of high-contrast circles with known dimensions and location. Four of the circles are labeled with the numbers 1 to 4, allowing the different camera pods to correctly identify the orientation of the calibration plate.
During the calibration process, the center of the calibration plate is placed as close to the linac isocenter as possible. The AlignRT manual suggests aligning the calibration plate cross-hairs with the linac field cross-hairs or the room lasers and placing the surface of the calibration plate at 100 cm source-to-surface distance (SSD). Next, the numbered circles are located on images taken from each of the three camera pods. This allows for a spatial correlation between the cameras that allows the system to triangulate objects in space. The AlignRT system's isocenter is set at the center of the cross-hair on calibration plate.
The IC procedure still requires the spatial calibration with the calibration plate, so the MC must still be completed. In addition, the IC uses a cube phantom with five-embedded ceramic spheres (see Fig. 2). One of the spheres is located at the center of the cube, while the other four are arranged asymmetrically around the center sphere. During the calibration process, the phantom is

2.C | Displacements with intentional miscalibrations
To evaluate the potential problems with the calibration method that only utilizes the calibration plate, intentional miscalibrations were applied. Using table shifts, the calibration plate was displaced in one direction from the linac isocenter prior to the MC to cause a miscalibration. The one-dimensional miscalibrations investigated were: AE3.0 mm in the longitudinal (lng) and lateral (lat) directions and AE1.0 mm in the vertical (vrt), lng, and lat directions. This resulted in a total of 10 investigated miscalibrations. After each miscalibration, the head phantom was returned to the position initially indicated using CBCT. AlignRT surface tracking was started using the CTdefined reference surface and the indicated offsets were recorded at each of the couch rotational positions from the treatment plan.
In addition, to investigate the procedure used clinically for SRS, a new reference surface was captured with AlignRT with the head phantom at the position initially indicated using CBCT (at a couch rotation of 0°). Surface tracking was started and again, the couch was rotated to the planned couch rotations and the indicated displacements were recorded.    however, the direction of the displacement changes from lng to lat.

2.D. | Displacements after isocenter calibration
That is, the direction of the indicated offset remains in the same direction relative to the treatment room.

3.B | Displacements after isocenter calibration
A graph of the AlignRT-indicated displacements vs. couch angle while using the AlignRT-captured reference surface for various intentional miscalibrations after IC is shown in Fig. 8. F I G . 6. Graph of the AlignRT-indicated displacements vs. couch rotation while using the AlignRT-acquired reference surface for various intentional miscalibrations before the isocenter calibration was applied. AlignRT users should evaluate the potential need for the IC before utilizing the system for these non-coplanar SBRT treatment techniques.
It is worth noting that an offset in calibration plate (intentional or not) during the daily QA of the AlignRT system will not result in a displacement of the AlignRT isocenter since the daily QA is simply confirming the spatial correlation between each of the camera pods is still intact (within a tolerance). In addition, the MC has the option to take additional images of the calibration plate raise 75 mm from isocenter. Again, this does not affect the isocenter of the AlignRT system, but is used to help the accuracy of the system in the situation where the isocenter is located posteriorly in the patient which results in the monitored surface being farther from isocenter.
F I G . 7. The AlignRT display of the head phantom relative to the AlignRT-acquired reference surface with a À3.0 mm lng intentional miscalibration at a couch angle of 0°(left) and at 270°(right).
F I G . 8. Graph of the AlignRT-indicated displacements vs. couch rotation while using the AlignRT-acquired reference surface for various intentional miscalibrations after the isocenter calibration was applied.
F I G . 9. Image demonstrating the geometric effect of an isocenter misalignment of distance X between the linac and AlignRT system when an AlignRT-acquired reference surface is used. With the couch at 0°, the reference and monitored surface match. However, when the couch is rotate AE90°, the magnitude of the indicated offset between the reference and monitored surface is given as (X √2).