3D‐printed headrest for frameless Gamma Knife radiosurgery: Design and validation

Abstract Purpose Frameless Gamma Knife stereotactic radiosurgery (SRS) uses a moldable headrest with a thermoplastic mask for patient immobilization. An efficacious headrest is time consuming and difficult to fabricate due to the expertise required to mold the headrest within machine geometrical limitations. The purpose of this study was to design and validate a three‐dimensional (3D)‐printed headrest for frameless Gamma Knife SRS that can overcome these difficulties. Materials and methods A headrest 3D model designed to fit within the frameless adapter was 3D printed. Dosimetric properties of the 3D‐printed headrest and a standard‐of‐care moldable headrest were compared by delivering a Gamma Knife treatment to an anthropomorphic head phantom fitted with an ionization chamber and radiochromic film. Ionization measurements were compared to assess headrest attenuation and a gamma index was calculated to compare the film dose distributions. A volunteer study was conducted to assess the immobilization efficacy of the 3D‐printed headrest compared to the moldable headrest. Five volunteers had their head motion tracked by a surface tracking system while immobilized in each headrest for 20 min. The recorded motion data were used to calculate the average volunteer movement and a paired t‐test was performed. Results The ionization chamber readings were within 0.55% for the 3D‐printed and moldable headrests, and the calculated gamma index showed 98.6% of points within dose difference of 2% and 2 mm distance to agreement for the film measurement. These results demonstrate that the headrests were dosimetrically equivalent within the experimental uncertainties. Average motion (±standard deviation) of the volunteers while immobilized was 1.41 ± 0.43 mm and 1.36 ± 0.51 mm for the 3D‐printed and moldable headrests, respectively. The average observed volunteer motion between headrests was not statistically different, based on a P‐value of 0.466. Conclusions We designed and validated a 3D‐printed headrest for immobilizing patients undergoing frameless Gamma Knife SRS.


| INTRODUCTION
The Gamma Knife Icon (Elekta, Stockholm, Sweden) system equipped with an integrated cone-beam computed tomography (CBCT) allows for frameless stereotactic radiosurgery (SRS) treatments to be performed. Whereas traditional Gamma Knife treatments are delivered in one fraction with the patient immobilized using a stereotactic frame affixed to the patient's skull, frameless treatments are delivered with the patient immobilized in a moldable patient-specific headrest and a thermoplastic mask. This system enables fractionated Gamma Knife treatments. A CBCT of the daily setup, which is used to correct for daily patient positioning, is acquired and registered to the planning CT or MRI. Intrafraction motion is monitored using the high definition motion management (HDMM) system, which uses an infrared (IR) camera to monitor an IR reflective marker on the patient's nose as a surrogate for head motion. 1 The combination of CBCT positioning and HDMM has been demonstrated to provide sufficient localization and motion management for Gamma Knife SRS treatments. [2][3][4] The patient-specific mask and headrest used for immobilization are created during the patient's treatment simulation on the Gamma Knife unit. Creation of the headrest during simulation can be challenging due to a combination of factors, especially for centers with limited experience in creating radiotherapy immobilization, such as Gamma Knife units in neurosurgery departments, or for centers that frequently rotate staff who may have limited previous experience with Gamma Knife simulations. The primary challenge is being able to mold the patient headrest within the geometrical limitations of the Gamma Knife imaging and treatment system. The CBCT system has a limited field of view, making it critical that the patient is indexed in a location within the frameless adapter such that the entire skull can be imaged for accurate registration. Treatment collisions also represent a challenge, as target locations very anterior or inferior can potentially be untreatable due to possible collision of the patient and the source assembly. Commercial moldable headrests currently used are activated by heat or water and have a limited amount of time to be shaped, which can make it challenging to position the patient in an optimal treatment position considering the aforementioned geometrical limitations. This problem has been previously described in the literature by Li   The headrest was 3D printed in polylactic acid (PLA) material using a GigaBot 3 3D printer (re:3D Inc., Houston, TX). The printer settings used were a nozzle temperature of 240°C and a bed temperature of 60°C. The model was printed with a layer height of 200 microns, two shell layers, and 2% infill percentage with a rectilinear pattern, leaving the headrest mostly hollow yet strong enough to support a patient's head.

2.B | Dosimetric study
Treatment planning for Gamma Knife is done in Leksell GammaPlan software (Elekta Instrument AB, Stockholm, Sweden). In typical clinical use, dose calculations are performed using the TMR algorithm, which calculates dose based on homogeneous water within the shape of a patient's skull as defined by the planning image data set. 8 As no heterogeneity corrections are performed, the headrest is not modeled when generating the treatment plan. This meant it was necessary to conduct a dosimetry study to determine if the 3D-printed BALTZ ET AL. The planned treatment was delivered to the phantom and the ionization chamber integrated charge was recorded. Two repeat ionization measurements were acquired for this setup. The procedure was repeated with the head phantom setup in a Moldcare Cushion that had been formed to the phantom, which represents the current standard-of-care headrest. 3D-printed film as comparison using 2% dose difference, 2 mm distance to agreement, global normalization, and 20% dose threshold as the gamma index criteria.

2.C | Immobilization study
Minimizing patient motion during treatment is the primary purpose of immobilization, and is particularly critical for SRS treatments that utilize smaller treatment margins than non-SRS treatments. In order to be used clinically, a standardized 3D-printed headrest must be able to immobilize the patient comparably to the current patientspecific headrest used. A study was conducted to compare the immobilization efficacy of the standardized 3D-printed headrest and the current patient-specific standard-of-care headrest.
Intrafraction patient motion during frameless Gamma Knife treatments is monitored with the HDMM system. If the system measures patient movement in excess of a predefined threshold, typically on the order of 1.5 mm, the system will pause the treatment until the patient positioning is back within tolerance. This system could be used to provide an accurate in vivo measurement of the immobilization efficacy of the two headrests. However, leakage radiation is always present in the vicinity of the Gamma Knife, even when the shield doors are closed. In the interest of complying with ALARA principles, an immobilization study could not be conducted using the Gamma Knife built-in HDMM system. Instead, AlignRT (VisionRT, London, UK), a commercial system used to perform surface-guided radiation treatment was used to mimic the HDMM system. This system has been previously used to characterize new immobilization devices [11][12][13] and provides the capability to monitor patient movement with submillimeter accuracy similar to the HDMM system on the Gamma Knife.
An immobilization study was conducted with 5 volunteers to compare movement while being immobilized in the 3D-printed headrest vs while immobilized in a moldable standard-of-care headrest.
The experimental setup used for the immobilization study is presented in Fig. 4. A jig was made to hold the Gamma Knife MR frameless adapter on a linear accelerator treatment couch. The MR adapter has the same geometry as the adapter used for treatment.  The frame rate and accuracy of the motion tracking of the AlignRT system is dependent on the size of the surface ROI tracked.
The uncertainty for motion tracking data acquired in this study was estimated by placing a stationary anthropomorphic head phantom in the headrest and tracking the ROI shown in Fig. 5

3.A | 3D-printed headrest
The headrest took 8 h to print and had a material cost of approximately $15. The final headrest fit tightly in the Gamma Knife frameless adapter, shown in Fig. 6.

3.B | Dosimetric study
Images from the CT scans acquired for the HU comparison of the headrests are shown in Fig. 7. The standard-of-care headrest had an average HU of −914 AE 31 corresponding to an effective density of 0.07 g/cm 3 , while the 3D-printed headrest had an average HU of −848 AE 244 and effective density of 0.14 g/cm 3 . The inside of the 3D-printed headrest is mostly air, which has a lower HU than the material in the standard-of-care headrest. However, the PLA material of the 3D-printed headrest has a density of 1.10 g/cc, which leads to the 3D-Printed headrest having a larger average HU and standard deviation compared to the moldable headrest. Overall, the HUderived effective density of the standard-of-care and 3D-printed headrest were comparable.

3.B.1 | Ion chamber measurements
The results of the ion chamber measurements are presented in

3.B.2 | Film measurements
The gamma index map and isodose comparison for the films are presented in Fig. 8. The calculated gamma index was 98.6% of points passing. Considering the inherent uncertainty of EBT3 film is at least 3.2%, 15 the gamma index suggests very good agreement of the two films. These results demonstrate that the delivered dose distribution for each headrest was clinically equivalent and provides additional evidence of dosimetric equivalence of the two headrests.

3.C | Immobilization study
Plots of the vector distance vs time for each volunteer are presented in Fig. 9. Data on the mean displacement and standard deviation for each volunteer and immobilization setup are presented in Table 2.
The uncertainty in the motion tracking data was estimated as 0.2 mm, which was the average vector distance recorded when tracking the ROI shown in Fig. 5 on the stationary anthropomorphic head phantom.
Qualitatively, it can be seen in the plots in Fig. 9 that each volunteer had very similar motion trends while immobilized in the 3Dprinted headrest and the standard-of-care headrest. This was reflected in the recorded data, with the difference in the average motion of the volunteers for each headrest being only 0.05 mm. The paired t-test comparing the average movement between the 3D-printed and standard-of-care had a P-value of 0.466, meaning there was not a statistical difference in the volunteer movement while immobilized in the two headrests. The difference in standard deviations was within 0.08 mm, with the standard deviation for the 3D-Printed headrest being slightly smaller. These results demonstrate that the 3D-printed headrest has comparable immobilization efficacy as a typical patientspecific standard-of-care cushion. While the tape immobilization used in this study would be less effective than a full thermoplastic mask used in frameless Gamma Knife treatments, even with less restrictive head immobilization, the volunteer motion between headrests in this study was comparable and demonstrates the standardized 3D-printed headrest would not be a limiting factor in immobilization. In addition to improving simulation, the 3D-printed headrest could potentially enable patients to undergo frameless treatment who otherwise would not have been able to be treated due to collision. Targets located very anterior or inferior in the skull can be challenging to treat due to collision of the patient in the machine when positioning the target at isocenter. As shown in Fig. 7, the 3Dprinted headrest developed in this study is significantly thinneron the order of 1 cmthan a typical moldable headrest currently used. The thinner headrest provides additional clearance, which could make the difference in being able to treat a target in a collision susceptible location. Another technique used to work around collision issues in conventional Gamma Knife treatments is to change the gamma angle, which is the angle the frame locks into the fixation mount. Typically a neutral head position, which is achieved with the 90°gamma angle, is the default angle used for frame-based One of the primary strengths of 3D-printing technology is the ability to rapidly manufacture unique devices. Although this study primarily focused on validation of a standardized design, the headrest design developed in this study could be used as a foundational shape to generate patient-specific headrests. Prior studies have demonstrated the ability to use a patient's pre-treatment diagnostic imaging data to 3D-print patient-specific treatment devices. [16][17][18][19] The majority of Gamma Knife patients have a diagnostic MRI, which could be used to generate a patient-specific headrest model that could be 3D-printed prior to a simulation. This technique could offer similar time savings during simulation as using a premade standardized headrest, while potentially offering better immobilization because it would be molded to the patient's specific head shape.

| DISCUSSION AND CONCLUSION
In conclusion, we have designed and validated a 3D-printed headrest that can be used for patient immobilization in frameless Gamma Knife treatments. Use of the 3D-printed headrest can alleviate the challenges associated with using the current standard-of-care moldable headrests. The design developed in this study can be utilized as a foundation for future research in 3D-printed head position assisting and patient-specific headrests.
T A B L E 2 Summary of VisionRT movement data recorded for volunteer immobilization study.

Mean vector distance (mm)
Standard deviation