Independent evaluation of the effectiveness of IsoCal in improving image center accuracy on Varian TrueBeam and Clinac machines

Abstract Modern medical linear accelerators (linacs) are often equipped with image guidance systems that are capable of megavolt (MV), kilovolt (kV), planar, or volumetric imaging. On Varian TrueBeam linacs, the isocenter accuracies of the imaging systems are calibrated with a procedure named IsoCal. On Clinac series linacs from Varian, installation of IsoCal is optional and the effects of IsoCal on the imaging systems can be turned on or off after the IsoCal procedure is performed. In this study, we report on the effectiveness of IsoCal in improving the coincidence of the image centers with the radiation isocenter, using an independent Winston‐Lutz (WL) method to locate the radiation isocenter. A ball‐bearing phantom was imaged with 2D MV, 2D kV, and cone beam computed radiography systems on two TrueBeam and two Clinac machines. Using the same phantom, digital WL tests with 16 combinations of gantry and collimator angles were performed to locate the radiation isocenter. The offsets between the IsoCal‐calibrated image centers and the WL radiation isocenter were found to be within 0.4 mm on the four linacs in this study. When IsoCal was turned off, the maximal offsets of the image centers were greater than 1.0 mm on the two Clinac machines. The method developed in this study can be used as a vendor‐independent quality assurance tool to assess the isocentricity of the image centers and radiation central axes.

the lack of precision tools needed in evaluating the small, often submillimeter, deviations of the image centers from the radiation isocenter. The traditional IGRT QA method uses a cube phantom that is positioned at the mechanical isocenter using the room lasers. 5 The accuracy of this method is inherently limited due to the uncertainty in the room lasers. The recent Machine Performance Check (MPC) method is reported to have high accuracy; 12,13 however, it is developed by the same linac vendor (Varian) and thus cannot be regarded as an independent QA method. In this study, we independently assess the effectiveness of the Varian Iso-Cal technique. We employ the digital Winston-Lutz (WL) test method, which has been demonstrated to have submillimeter accuracy. 4,6,[14][15][16] This method measures the image center accuracies directly against the radiation isocenter. Unlike the traditional WL test, the digital WL test does not require a precision linear stage to adjust the phantom position iteratively to the radiation isocenter. Therefore, the digital WL test is simple and fast in terms of its phantom and the setup. A previous study used the digital WL test to verify the IsoCal effectiveness on Varian Clinac machines. 10 The study showed that IsoCal increased the coincidence between the 2D image centers and the radiation isocenter to within 0.6 mm.
The digital WL in that study used four gantry angles and a single collimator angle, that is, the collimator rotation was not considered.
In this study, we re-design the digital WL test by employing more gantry and collimator angles to achieve higher accuracy in localizing the radiation isocenter. We implement this method to evaluate the IsoCal effectiveness on Varian TrueBeam machines. Furthermore, we include the evaluation of CBCT image centers on both Clinac and TrueBeam machines in this study.

2.A | IsoCal on Varian clinac machines
The theory and procedure of IsoCal for Varian Clinac machines have been previously described. 10,11 First, a cylindrical phantom containing 16 BBs was positioned at the mechanical isocenter using the room lasers. An aluminum plate with a steel pin was inserted in the gantry accessory slot. Four MV images of the phantom were obtained at collimator angles (195°, 270°, 0°, 90°; Varian IEC 601-2-1 scale) and a fixed gantry angle (0°) to establish the central axis (CAX) of the radiation beam at the given gantry angle. Subsequently, 8 MV images of the phantom were acquired at various gantry angles (225°, 270°, 315°, 0°, 45°, 90°, 135°, 180°) and a fixed collimator angle (90°) to establish the radiation isocenter. Finally, 8 kV images and a CBCT scan of the phantom were obtained. Using these images and the Varian IsoCal software, the 2D MV, 2D kV, and CBCT images centers were localized relative to the radiation isocenter. The offsets of the image centers were used to create a system file in XML format. In subsequent imaging (MV, kV, or CBCT), the XML file was used to register an image center correction to the image's DICOM header. The correction was applied to the digital graticule when the image was displayed in the Varian OBI console, or in a third party software such as MOSAIQ (Elekta AB, Stockholm, Sweden).

2.B | IsoCal on Varian TrueBeam machines
The hardware of the IsoCal system on TrueBeam linacs is identical to that on Clinac linacs. Two automatic procedures, IsoCal calibration and IsoCal verification, are associated with the IsoCal system. In both procedures, we acquired 4 MV images of the phantom at different collimator angles (195°, 270°, 0°, 90°) and a fixed gantry angle (180°). We then acquired a series of MV images and kV images of the IsoCal phantom with a full gantry rotation, approximately one image every 3°gantry angle. Similar to the IsoCal system on Clinac, the IsoCal calibration was to determine the corrections required to align the centers of the kV and MV images to the radiation isocenter. At the end of IsoCal calibration, the correction data were stored in a configuration file in the Varian on-board imaging system. After the IsoCal calibration was completed, the IsoCal corrections were subsequently applied by correcting the lateral and longitudinal MV or kV imager positions as a function of gantry angles before the images were taken, and thus no shifts of the digital graticules were needed. This was in contrast to procedures used with the IsoCal on Clinac where imagers were not shifted but the corrections were applied by shifting the digital graticules of the acquired images.
On TrueBeam linacs, the IsoCal verification procedure was performed to validate the IsoCal calibration. The IsoCal verification procedure was the same as the IsoCal calibration procedure except that the IsoCal corrections were applied to MV and kV imagers. The residual corrections from the IsoCal verification should be close to zero.

2.C | Digital Winston-Lutz test
The WL phantom included a tungsten sphere of 6.5 mm in diameter glued on an acrylic rod, which was screwed into an acrylic block ( Fig. 1). The phantom was placed on the treatment table and kept stationary during the entire image acquisition. The BB was placed near the linac isocenter (within ±3 mm in each direction) using the guidance of room lasers. The center of the BB was used a reference point to which the radiation isocenter and the image centers were localized. Thus, there was no need to place the BB exactly at the radiation isocenter. Two coordinate systems were used in this study. The first coordinate system x-y-z was static with the origin defined at the BB center. The second coordinate system u-v was defined for the 2D MV and 2D kV images (Fig. 1). The u-v coordinate system rotated with the MV source or the kV source. The origin of u-v coordinates was defined at the projection of the BB center on the imager. The values of u-v coordinates were scaled to the isocenter plane.
To locate the radiation isocenter using the WL method, the BB phantom was imaged with a 10 × 10 cm 2 square MV field shaped by the multi-leaf collimator (MLC). The images were acquired at eight gantry angles (225°, 270°, 315°, 0°, 45°, 90°, 135°, 180°) and two opposing collimator angles (90°and 270°). A total of 16 MV images were obtained to compute the location of the radiation isocenter. In previous studies, only 4 MV images (four cardinal gantry angles and one collimator angle 0°) were employed for simplicity. 6,10 The use of opposing collimator angles in this study was intended to improve the accuracy of radiation isocenter localization.
The MV images were processed with an in-house MATLAB (MathWorks Inc., Natick, MA) program. The details of the algorithm were reported previously. 16 Briefly, the radiation field center in each MV image was located relative to the BB center. Then the 16 radiation CAX were reconstructed in the 3D x-y-z space. Finally the radiation isocenter was determined as the point that had the minimal average distance from all CAX. With eight gantry angles and two opposing collimator angles, the uncertainty in the resulting radiation isocenter was estimated to be less than 0.2 mm. 16 For each gantry angle and collimator angle, we also computed the distance of the radiation CAX to the WL radiation isocenter. The maximal and mean distances for the 16 radiation CAX were used to characterize the size of the isocenter sphere. The gantry sag was computed as the longitudinal (v or z direction) shifts of the CAX during the gantry rotation. The couch rotation is not considered in this study because (a) the IsoCal procedure does not include the couch rotation, (b) the mechanical walkout during couch rotation is patient-dependent, that is, varying with the patient weight and how the weight is distributed on the couch, (c) the couch walkout can be corrected by realigning the couch with the calibrated ceiling lasers.
The digital WL tests were performed twice on each of the two Varian TrueBeams and two Varian Clinac 21EX linacs.

2.D | IsoCal corrections to image centers
where θ t was the couch angle in degrees when the CBCT scan was acquired. If θ t = 0°, AcqIso′ = AcqIso. Third, the IsoCal correction was computed as where FOV was the field-of-view, and ST was the slice thickness of the CBCT images.
The final IsoCal-corrected image center position was the sum of the default image center position and the IsoCal correction described above. After the BBs were detected in the MV, kV, or CBCT images, the image centers were localized relative to the BB center. Given the WL radiation isocenter position as determined from Section 2.C above, the image center offsets relative to the WL radiation isocenter could be easily computed.
A useful by-product of doing the WL test is to quantitate the wobble of radiation fields. Figure 7 shows the wobble of radiation linacs (Fig. 2). Thus, IsoCal has the capability to significantly improve IGRT accuracy and potentially enable the linacs for stereotactic types of radiation therapy.
The IsoCal calibration consists of vendor-provided hardware, software, and methodology. Many technical details and much of the interpretation of the IsoCal results appear as a "black box" to clinical medical physicists. Thus, it is imperative to have an independent, more transparent method to assess the effectiveness of IsoCal. The WL test has been established for decades as a precise way of localizing the radiation isocenter of a linac. The concepts of using a simple BB phantom and circular or square radiation fields in the WL test are well understood. In this study, we used the WL test to locate the radiation isocenter, which was independent of the IsoCal methodology. Unlike in some previous studies, 12 we did not use Varian toolkits (eg, IsoLock  F I G . 5 . Offsets of CBCT image center from the WL isocenter on Clinac1 (a) and Clinac2 (b). x, y, z were the lateral, vertical, and longitudinal dimensions of the room coordinate system. IsoCal was turned on (green) or off (red). The measurements were made twice (circles and crosses).
F I G . 6 . Offsets of CBCT image center from the WL isocenter on TrueBeam1 (a) and TrueBeam2 (b). The measurements were made twice (circles and crosses). | 489 compared to the previous study (collimator = 0°, MLC leaves subject to gravitational pull). 10 Furthermore, we estimated that the IsoCal corrected image centers were within 0.5 mm of the "true" radiation isocenter. This estimation was based on ≤0.4 mm coincidence between the WL isocenter and the image centers, 0.1 mm systematic errors in the WL isocenter in this study, and 0.1 mm random errors.
Finally, the ways that Varian implements the IsoCal corrections are worthy of discussion. On the Clinac series, the IsoCal correction is imbedded in the DICOM header of the acquired images. Interpretation of this information is not a straightforward task for the end users, even with the guidance of the vendor's DICOM Conformance Statement or system manuals. On the TrueBeam machines, the Iso-Cal correction is applied to adjust the imager position. Once the image is taken, no extra shift is needed to correct the image center.
This latter practice eliminates the need of interpretation or guesses by the third party and thus is more desirable in terms of preventing potential errors in patient positioning.

| CONCLUSIONS
We investigated the effectiveness of IsoCal on Varian's Clinac and TrueBeam machines. The 2D MV, 2D kV, and CBCT image centers were found to coincide with the WL radiation isocenter within 0.4 mm on both types of Varian machines. If not corrected by Iso-Cal, the image centers on the Clinac machines could be more than 1 mm off from the WL isocenter. Our independent study indicates that IsoCal is essential in improving the image center accuracy on the Varian linacs to within 0.5 mm accuracy. The method developed in this study can be used as a vendor-independent QA tool to assess the isocentricity of the image centers and radiation central axes.

ACKNOWLEDGMENTS
We thank Michael Worley and the Department of Scientific Publications for editorial review of the manuscript. We are also grateful for the anonymous reviewers' scrutiny of the manuscript and insightful critique.

CONF LICT OF I NTEREST
The authors do not have any conflicts of interest to declare. This work was not sponsored by Varian Medical Systems.