"Wednesday 12 June"

The RTsafe avatar phantom is a novel water based humanoid phantom, modelled on a patient’s CT scan, with a highly detailed skull anatomy created by virtue of its 3D printed construction. During the commissioning of three Gamma Knife Icon units, a variety of tests were devised using these phantoms:

Point dose measurements with ionisation chambers (PTW Semiflex 31010 and Pinpoint 3D 31014) were performed for single (16mm) shot plans and compared with the Leksell GammaPlan (LGP) calculated dose using the TMR10 algorithm. 2D measurements using EBT3 film and 3D measurements using gel dosimetry allowed End-to-End testing for complex multi-isocentric plans.
It was possible to mount the Leksell G frame on the rigid outer surface of the phantom. This allowed measurement of the degradation of stereotactic accuracy due to frame distortion. The four mounting pins were torqued to values between 20 and 90cNm in 10cNm steps. For each torque setting, separate CT and CBCT scans were acquired. Correction of the location of a reference isocentre was investigated when a treatment plan based on the stereotactic (fiducialised) CT was co-registered with a pre-treatment CBCT.

Results
Point dose measurements for the single shot plans demonstrated a mean error of -8.0% and -6.1% for Semiflex and Pinpoint respectively when compared to the TMR10 algorithm, but reduced to -3.4% when planned in the heterogenous phantom with Convolution. Local gamma pass rates for 5%/1mm were >95% for 2D film studies. 3D gel dosimetry yielded acceptable correlation between LGP and measured dose when assessed via DVH comparison and gamma index analysis.
Comparison between targeting using conventional stereotactic CT and CBCT showed a mean agreement of 0.25mm over a wide range of torque settings.

Conclusion
The RTsafe phantom has demonstrated its versatility for use in commissioning the Gamma Knife Icon

A novel full-system test (FST) phantom and method have been developed to demonstrate and quality assure the geometric accuracy of image co-registration and overall shot delivery in the context of SRS using Gamma Knife® Icon™.  The method uses eight Vernier scale bars to achieve sub-voxel precision co-registration accuracy measurements and pin-located radiochromic films to determine overall shot delivery precision.  A Procrustes superimposition analysis method was used to assess residual registration errors and decouple these from focal precision errors which also contribute to the observed shot position full-system test error.

Validation tests demonstrated that artificially applied randomly generated synthetic registration errors of < 0.15 mm could be accurately detected and quantified.  Cross-validation of full-system test results with the manufacturer standard focal precision test demonstrated that both approaches measure similar focal precision errors, to within < 0.1 mm, and that registration and focal precision components of the full-system geometric error can be successfully decoupled using our Vernier registration analysis approach. 

CBCT co-registration errors were shown to be of comparable magnitude to the focal precision errors, demonstrating that CBCT registration based in-mask treatments can achieve sub-voxel inter-fraction geometric accuracy, rivalling traditional frame-based immobilisation.  Whilst real patient treatments also exhibit intra-fraction motion, the use of IFMM monitoring has been shown to restrict this error to the same order as the inter-fraction motion errors reported here. This novel full-system geometric test method and phantom design concept is in principle applicable in principle to any SRS technique involving high (sub-voxel) image co-registration performance, enhancing confidence in rigid registration based positional correction for these critical applications.

Introduction:

The Gamma Knife Icon allows adaptive, fractionated radiotherapy (a-gkFSRT) of cerebral lesions in a stereotactic environment using cone-beam computer tomography (CBCT) (re)positioning and thermoplastic mask fixation. Interfractional patient motion is countered with translational table movement and rotation of the treatment plan/shots, resulting in an updated dose calculation. Here, we analyzed interfractional patient motions and the corresponding plan adaptions.

Material and Methods:

We recorded a total of 439 fractions for 36 patients (15 male and 21 female) that underwent a-gkFSRT for intracranial lesions (meningioma, brain metastasis resection cavities, primary metastases, vestibular schwannoma and pituitary adenoma). For each fraction, we analyzed the mean interfractional patient motion and compared the resulting deviation after adaptive planinng. Furthermore, a subset of 198 fractions were analyzed in terms of plan quality of the daily plan adaption. Finally we analyzed the largest patient motions and the resulting deviations after plan adaption.

Results:

For all 439 fractions, the interfractional translation shifts were 0.05±0.55mm, -0.39±0.59mm and -0.08±1.37 mm in x-, y- and z-direction, respectively. The interfractional rotational differences were -0.15±0.98°, -0.09±0.62° and -0.15±0.93° around the x-, y- and z-axis. When analyzing 198 selected fractions, we found a deviation between planned and delivered fraction doses of -0.05±0.15% for the D_min to the target, 0.08±0.40% for D_max to the target, 0.00±0.06% for target volume coverage, 0.00±0.00% for PCI and 0.24±0.37% for gradient. Of note, even the largest interfractional patient shift (>2mm or >2°) did not result in clinically relevant deviations of dose distribution after plan adaption with only minimal deviations in gradient (<0.72%) and D_min to an organ-at-risk (‑11.55%). 

Conclusion:

Interfractional patient shifts in a-gkFSRT are in submillimeter ranges and do not require patient repositioning. Daily plan adaption results in plans that are almost identical to the reference treatment plan, even in case of major interfractional positioning shifts.

The Icon-model Gamma Knife (GK) introduced on-board CBCT for GK SRS.  Intended to facilitate mask-based SRS, Icon also provides an opportunity for QA of conventional frame-based patients via pre-treatment image verification.

Stereotactic definition of planning MR images for our frame patients is based upon conventional fiducial marker localisation.  Pre-treatment CBCT is performed routinely for these cases.  Within Leksell GammaPlan (LGP) pre-treatment CBCT is co-registered to planning MRI.  Since both should share a common stereotactic space, LGP-reported co-registration translations and rotations are ideally zero.  In practice non-zero values result from definition and co-registration uncertainties, but excessive non-zero values could indicate frame slippage or fiducial box displacement.

Analysis of 501 co-registrations from 470 patients fitted with one of four frames (n = 126, 121, 126 and 128, respectively) between Dec-15 and Jan-19 is presented.  CBCT was co-registered against a 120-slice 0.8x0.8x1.5mm voxel MPRAGE (n=465) or a 52-slice 0.4x0.4x1.0mm voxel CISS (n=33) acquired on a Siemens Avanto (n=483) or Aera (n=18).

Overall mean(S.D.) X/Y/Z translations and rotations were 0.16(0.39)/0.11(0.26)/0.67(0.34)mm and 0.15(0.31)/-0.05(0.22)/0.26(0.16)deg, respectively.  Mean differences of 0.00/0.21/0.38mm and 0.42/0.01/0.19deg MPRAGE-vs-CISS were significant (p<0.01) for Y,Z rotations and X,Z translations, as were Avanto-vs-Aera mean differences of -0.10/0.00/0.06mm and -0.20/0.03/0.38deg for X,Z rotations.   Single factor ANOVA indicated significant (p<0.01) mean differences in translations and Z rotation between frames.  No correlation to LGP-reported definition errors was found.

Pre-treatment CBCT offers valuable verification of MR fiducial-based stereotactic definition integrity beyond the LGP-reported definition errors.  Three cases of frame slippage were identified by this process, all characterised by excessive (>2 S.D.) values for at least two of the translation/rotation values as compared to the overall data.  More detailed analysis of our data has indicated dependence upon sequence, scanner and frame and these factors should be considered when interpreting pre-treatment CBCT verifications on Icon.

Aim

The recent upgrade of the BrainLab Exactrac system (v6.2) removes the need for the use of a Frameless Localizer and Target Positioner (TarPo) at CT, allowing for treatment prepositioning based on a fusion of an internally stored CT of the frameless SRS mask base. The redundancy of the TarPo gives way for the use of the Frameless Radiosurgery Positioning Array at CT, allowing it to be contoured and its attenuation accounted for in iPlan dosimetry. We report on the initial setup accuracy of the mask base fusion (MBF) feature compared to the TarPo localisation (TL) method using ExacTrac <v.6.2 for patients receiving stereotactic radiosurgery (SRS) or fractionated SRS (fSRS) for intracranial tumours.

Methods

94 SRS (39 TL, 55 MBF) and 72 fSRS (40 TL, 34 MBF) patients were retrospectively analysed. The median initial image corrections in 6-degrees of freedom (DOF) were compared between TL (n=79) and MBF (n=79) for both SRS and fSRS (first fraction only) patients with a Wilcoxon-Rank test (p<0.05). Systematic and random error was calculated for all 6DOF using the daily initial corrections of all fractions of the fSRS patients.

Results

The median initial corrections for the TL method and MBF method respectively were: lateral shift -0.52mm vs -0.97mm (p=0.06), longitudinal shift 0.18mm vs 1.11mm (p=0.001), vertical shift 0.80mm vs -1.56mm (p <0.001), lateral angle -0.03deg vs -0.21deg (p=0.18), longitudinal angle 0.43deg vs 0.05deg (p=0.03) and vertical angle -0.15deg vs 0.07deg (p=0.37). The systematic error for positioning accuracy of fSRS was 1.5mm or less for all directions regardless of setup method, except the vertical shift (6.1mm using TL and 2.2mm using MBF). The random error for both methods was 0.9mm or less for all directions.

Conclusion

The initial setup accuracy is comparable between the TL and MBF methods. The longitudinal shift, vertical shift and longitudinal angle corrections are greater with the MBF method, but are less than 2mm or 0.5deg making them clinically inconsequential to the image verification process. An advantage of the MBF method is the redundancy of the TarPo allows for the Frameless Radiosurgery Positioning Array to be placed at CT and contoured, with attenuation then accounted for in the planning dosimetry, reducing the dose variation produced by the array during SRS delivery.

Leksell Gamma Knife Icon (Elekta AB) includes a cone-beam CT (CBCT) to define the stereotactic space without the need for an invasive frame. The aim of this study was to analyze the differences between the stereotactic frame - based coordinates and the CBCT - determined coordinates.

We performed CBCT before frame-based stereotactic Gamma Knife radiosurgery for 212 patients as an additional quality assurance (QA) procedure within radiosurgery treatment. The rotational and translational shifts, maximum shot displacement,delivered maximum doses for critical structures and coverage of the tumors were recorded. The factors investigated were z-coordinates of the right and left posterior pins, tumor localization, Leksell stereotactic coordinates of the tumor, tumor volume,  mean and maximum definition errors for the MR study. The statistical analysis was performed by the R statistical package.

The maximum shot displacement in anatomy was more than 1 mm for 32 patients. Planned tumor coverage was no less than 99% but the delivered one was less 95% in 12 cases. The z-coordinate of the tumor (p=0.036), volume of the tumor (p=0.045) were associated with differences of  the coverage. The x-coordinate of the tumor (p=0.029) and the mean definition error (p=0.024) on MR images were associated with maximum shot displacement.

The understanding of the differences between the stereotactic frame - based and the CBCT 3D stereotactic space is an important aspect of Gamma Knife Icon radiosurgery.

The differences between radiological and mechanical isocenters in case of frame-based radiosurgery are often assumed to be 0.2 - 0.5 mm. But the uncertainty of target and the structure localization could lead to more than 5% reduction of the prescribed dose coverage of the tumor.  The method of stereotactic space definition (frame or CBCT) that obtains more accurate results should be determined as well as the clinical significance of the demonstrated shifts are supposed to be defined

Object: Leksell Gamma Knife Icon enables us to apply new methods of immobilization using mask fixation and the option of fractionated treatment. This provides exceptional accuracy and precision of radiosurgery, making it a possibility for many more disease types and many more patients to be treated. We have consistently selected mask fixation, except 3 AVM patients, who needed digital angiography after frame fixation for dose planning.

Methods: We retrospectively analyzed 566 patients (702 times) who underwent Gamma Knife Icon using mask fixation between September 25th, 2017 and December 31th, 2018 at Rakusai Shimizu Hospital. The most common disease was brain metastases (384 patients), followed by meningioma (78), vestibular schwannoma (24), AVM (17), trigeminal neuralgia (15) and others (48). Patients with small, few, newly diagnosed, and non-eloquent area tumors were treated in a single session. If the tumor volume was larger than 5.0 ml, recurrence, or the location was in an eloquent area, we applied a fractionated schedule. If the tumor number was large, we selected a multisession schedule. Therefore, 209 patients were treated in a 　 single session, 377 with fractionation, and 116 with multiple sessions. For higher accuracy, we changed the upper limit of the HDMM system from 1.5mm to 1.0mm for malignant tumors and 0.5mm for benign tumors.

Results: We selected fractionated schedules as follows; 7.0 Gy x 5Fr (5-10 ml), 4.2Gy x 10Fr (10-20ml), 3.7Gy x 10Fr (20-30ml), 3.2Gy x 10Fr (30ml-) for malignant tumors, and 2.7Gy x 10Fr for benign tumors. Compared with frame fixation, almost all of patients (97%) who had previously experienced the frame fixation felt more comfortable.

Conclusions: Although these results are limited to short periods, survival rated, local control rates and qualitative survival rated in patients unsuitable for SRS, such as those with large, recurrent, and eloquent site lesions, were within the acceptable ranges. Further examination is needed for comparison with staged Gamma Knife radiotherapy, Cyber-Knife and Novalis radiotherapies