The Accuracy of Individualized Three Dimensional-printing Template Assisted I125 Radioactive Seed Implantation for Recurrent head and Neck Cancer

Background:Individualized threedimensional-printing template(3D-PT) is developed to facilitate I 125 radioactive seed implantation (RSI) for recurrent head and neck cancer, while most of the previous studies were focused on the ecacy and safety proles, study on the accuracy of I 125 RSI is lacking. Therefore, the aim of this study is to evaluate the accuracy involvingneedle puncture and dosimetry of individualized 3D-printing template (3D-PT) assisted I 125 radioactive seed implantation (RSI) for head and neck cancer. Materials and Methods: From February 2017 to January 2020, clinical data of 41 patients (mean age, 58.5 ± 16.1 years; 28 males) with recurrent primary (48.8%)/secondary (51.2%) head and neck cancer underwent individualized 3D-PT assisted I 125 RSI under our institute were retrospectively reviewed. Results:A total of 430 [mean, 10.5 (range 3–17) per patient] needles were inserted. Technical success rate was 100% without major complication. The mean entrance deviation was 0.090 cm (95% Condence Interval, 0.081–0.098), whichwas signicantly larger in patients with primary cancer than patients with secondary cancer (0.107 ± 0.012 vs. 0.072 ± 0.012 cm, p (cid:0) 0.001).The mean intraoperative depth and angular of the needle were consistent with planned depth and angular (6.23 ± 0.24 vs. 6.21 ± 0.24 cm, p = 0.903; 83.14 ± 3.64 vs. 83.09 ± 3.66 degrees, p = 0.985, respectively). The mean deviation between planned and intraoperative depth and angular of the needle were 0.168 ± 0.024 cm and 1.56 ± 0.14 degrees, respectively, without signicant difference involving cancer type and implantation site (p = 0.065 and 0.092, respectively). The post-plan dosimetry parameters, including D90, D100, V100, V150, V200, conformity index, external index, and homogeneity index, were


Background
Brachytherapy (BT) is a speci c form of radiotherapy (RT) consisting of the precise placement of radioactive sources directly placed into or next to the tumor [1]. Currently, BT has the potential to deliver an ablative radiation dose ( 100Gy) over a short period of time directly to the altered tissue area with the advantage of a rapid fall-off in dose, and consequently, sparing of adjacent organs [2,3]. BT was recommended by both The Head and Neck Working Group of the European Brachytherapy Group and American Brachytherapy Society as one of the treatment for head and neck cancers, as external beam radiotherapy (EBRT) alone is di cult to spare adjacent normal tissues such as the salivary glands, the mandible, and mastication muscles which sustain undesirable late effects [4,5].
As the standard option for early stage prostate cancer, I 125 radioactive seed implantation (RSI) is also commonly used for recurrent head and neck cancer, providing a high local tumor control and preservation of organ functions [1,6,7]. While the deviation between planned and intraoperative needle pass way may occur even under the image guidance of Ultrasound, Computerized Tomography(CT), and Magnetic Resonance Imaging(MRI), leading to the mis-implantation of radiative seeds and unnecessary radiation or damage of surrounding critical organs and tissues. Owing to the dense critical organs and tissues (e.g. eyes, major vessels and nerve) in head and neck region, the accuracy of I 125 RSI and post-plan dosimetry was extremely critical for patients with head and neck cancer. Individualized three dimensional-printing template (3D-PT) was developed to facilitate I 125 RSI for head and neck cancer in order to improve the accuracy, optimize post-plan dosimetry, and shorten the RSI duration [8][9][10][11].
However, most of the previous studies involving 3D-PT assisted I 125 RSI were focused on the e cacy and safety pro les, mainly including local control, survival and complications [7,[11][12][13]. As the mean entrance deviation, planned and intraoperative deviation of needle depth and angular were deemed to represent the accuracy of needle related interventions, only 2 studies involving these parameters were published to described the accuracy of 3D-PT assisted I 125 RSI for head and neck cancer [9,14]. Indicated by existed evidence, 3D-PT assisted I 125 RSI may provide satis ed accuracy with consistent planed and post-plan dosimetry without additional complications [9,11,14], while cases reported in these studies was limited.
Here, the aim of the study is to evaluate the accuracyinvolving needle puncture and dosimetry of individualized3D-PT assisted I 125 RSI for patients with recurrenthead and neck cancer in our institute.

Study Design
The retrospective study was approved by our institutional review board and the requirement to obtain written informed consent was waived. The electronic database of our institute was searched and reviewed to identify eligible patients. Patients who underwent 3D-PT assisted I 125 RSI under CT guidance for the treatment of recurrent primary/secondary head and neck cancer between February 2017 to January 2020 were included. The indications for 3D-PT assisted I 125  The technical success rate, number of needle and seed, and the mean entrance deviation, the depth and angular of the planed and intraoperative inserted needles, and planed and post-plan dosimetry pro les were recorded.Subgroups by cancer type (primary/secondary) and implantation site (head/neck, bounded by the connecting line of the lower margin of the jaw, the mandibular angle, the tip of the mastoid process, superior nuchal line, and the external occipital carina) were further analyzed. Technical success was de ned as successful needle insertion and implantation of I 125 seed in the targeted tumors per preoperative/intraoperative plan. The mean entrance deviation was de ned as the super cial distance between the planed needle entrance point and the actual intraoperative needle entrance point. The depth and angular of the needle were directlycalculated and extracted from the BT Treatment Planning System (BT-TPS) after fusing the preoperative and intraoperative CT images into the same coordinate axis. The depth of the needle was de ned as the tip of the planed/inserted needle to the template surface when the needle is deemed deep enough/in place during the retrusive implantation. The angular of the needle was de ned as the angle between the needle and the horizontal axis when the needle is deemed in place. The planed and post-plan dosimetry parameters, including the prescription dose (PD), seed number, gross tumor volume (GTV), D90, D100, V100, V150, V200, conformity index (CI), external index (EI), and homogeneity index (HI), were recorded and compared. D90 and D100 refer to the dose delivered to the 90% or 100% of GTV, respectively. V100, V150, and V200 refer to the percentage of GTV receiving 100% or 150% or 200% of prescription dose, respectively.
Preoperative planning Patients all underwent contrast-enhanced CT with 2.5-mm or 5-mm (rarely, for large tumors only)resolution within 2-3 days before RSI. All patients were xed with a bow cap/vacuum pad in selected suitable gesture according to the lesion location and facilitation for RSI and then marked with surfacepositioning line. Then the CT images were transferred into the BT-TPS (Beijing Feitian Industries Inc and Beijing University of Aeronautics and Astronautics, Beijing, China). The preoperative plan was then established by de ning GTV and adjacent organs at risk (OARs), determining PD according to expert consensus on I 125 RSI [15] and clinical experiences gained in our department, commonly 110 to 160Gy, the radioactivity of seeds (usually 0.4-0.7 mCi), and the quantity and distribution of seeds and needles, designing needle puncture pathway, and verifying the dose calculations of the GTV and OARs (e.g., eyes, brainstem, spinal cord, groove for vertebral artery, throat, trachea, and glands). The optimization for D90 of GTV and doses delivered to the OARs was made by the physicists.
The individualizedplanning data in the BT-TPS was transferred into 3D imaging and reverse engineering software (Beijing Feitian Industries Inc and Beijing University of Aeronautics and Astronautics, Beijing, China) for digital modeling of individualized 3D-PT. The modeling data was then nally optimized with postprocessing in Magics 19.01 software (Materialise Company, Belgium) and the individualized3D-PT was produced by 3D light-cured rapid-forming printer RS6000 (Shanghai Liantai 3D Technology Company, Shanghai, China). The 3D-PT with 3 mm thickness contained individualizedinformation such as body-surface characteristics of the treatment area, localization markers, and entrance hole for 18 gauge needle [16], Figure 1.

Intraoperative implantation
All RSI procedures were performed with local anesthesia under the CT guidance. After skin preparation and sterilization, the 3D-PT was aligned to the therapeutic region according to the outline characteristics, reference line on the 3D-PT, previously marked surfacepositioning line, and positioning laser, Figure 2A-B. Then CT was performed to con rm the tting of 3D-PT exactly in position according to thepreoperative planning and design data in BT-TPS. Malposition between the BT-TPS data and the current CT image was adjusted in real-time, and then 2-3 locking needles (18 gauge) followed by the seed implantation needles (18 gauge) were percutaneously inserted via the pre-designed holes on the 3D-PT, Figure 2C-D. After all the needle were deemed in place, the I 125 seeds were implanted and delivered using the Mick applicator in a retrusive manner with 0.5/1.0 cm interval according to the preoperative plan and intraoperative re-plan, which was made and executed if necessary, Figure 2E-H.

Postoperative veri cation
All patients were re-evaluated immediately with CT simulator after I 125 RSI to validate the post-plan distribution of the I 125 seeds and rule out potential complications. Then, the CT images were transferred to BT-TPS to verify post-plan dosimetry, Figure 3. Dosimetry parameters including D90, D100, V100, V150, V200, CI, EI, and HI were evaluated. All RSI procedures were performed in accordance with relevant guidelines and regulations, as also described in published study [7,12].

Statistical analysis
Continuous variables were compared using paired t-test between planed data and intraoperative/postplan data. As 84.2% of the cancer in the head was primary cancer and 81.8% of the cancer in the neck was secondary cancer, subgroups by cancer type and implantation site were further analyzed in multivariate analysisusing linear regression model. A 2-sided p-value < 0.05 was considered as statistically signi cant difference. Statistical analyses were performed using SPSS software (version 26.0; SPSS, Chicago, IL, USA).

Subgroup analysis
In the univariate analysis, the entrance deviation in patients with primary cancer were signi cantly larger than patients with secondary cancer(0.107 ± 0.012 vs. 0.072 ± 0.012 cm, p 0.001) and was comparable in patients with implantation in the region of head and that of neck (0.089 ± 0.011 vs. 0.090 ± 0.013 cm, p = 0.938). The mean deviation between planned and intraoperative depth of the needle had no signi cant difference between patients with primary and secondary cancers (0.169 ± 0.041 vs. 0.167 ± 0.026 cm, p = 0.951) or between patients with implantation in the region of head and that of neck (0.152 ± 0.043 vs. 0.182 ± 0.025 cm, p = 0.224). While the mean deviation between planned and intraoperative angular of the needle was smaller in patients with primary than secondary cancers (1.18 ± 0.19 vs. 1.94 ± 0.19 degrees, p 0.001) and also smaller inpatients with implantation in the region of head than that of neck (1.25 ± 0.19 vs. 1.84 ± 0.19 degrees, p 0.001). Table 3.
In the multivariate analysis usinglinear regression modelinvolving both cancer type and implantation site. The linear regression involving entrance deviation had statistical signi cance (F = 17.064, p 0.001) with adjusted R 2 = 0.07. The entrance deviation in patients with primary cancer were signi cantly different in patients with secondary cancer(p 0.001, standardized B = -0.378) and were also signi cantly different between patients with implantation in the region of head and that of neck (p 0.001, standardized B = 0.266). The linear regression involving deviation between planned and intraoperative depth and angular of the needle both had no statistical signi cance (F = 2.748, p = 0.065; F = 2.398, p = 0.092, respectively) with adjusted R 2 = 0.008 and 0.006, respectively.

Discussion
The present study evaluated the accuracy involving needle puncture and dosimetry o ndividualized 3D-PT assisted I 125 RSI for recurrent primary/secondary head and neck cancer. As a result, the mean entrance deviation was 0.1 cm. The mean intraoperative depth and angular of the needle were well consistent with planned depth and angular without signi cant deference. The post-plan dosimetry parameters, including D90, D100, V100, V150, V200, CI, EI, and HI, were also well coordinate with planned dosimetry pro les without signi cant deference. Therefore, this study indicated that the accuracy of needle puncture was satis ed with consistent planed and postoperative dosimetry pro les of individualized3D-PT assisted I 125 RSI for patients with recurrent head and neck cancer.
Since the introduction of 3D-PT in the clinical practice, few studies investigate the accuracy of 3D-PT assisted needle related interventions [14,17]. As revealed by a non-inferiority randomized clinical trial that enrolled 200 patients for localizing small pulmonary nodules [17], localizer deviation did not signi cantly differ between the 3D-PT group and CT-guided group (mean, 8.7 vs. 9.6 mm; p = 0.36). The mean procedural durations were 7.4 minutes for the template-guided group and 9.5 minutes for the CT-guided group (P < 0.001). The mean CT related radiation dose was 229 mGy × cm in the template-guided group and 313 mGy × cm in CT-guided group (p < .001) [17]. Indicating that the use of the 3D-PT for localization showed e cacy and safety that were not substantially worse than those with the CT-guided alone, while signi cantly simplifying the procedure and decreasing patient CT related radiation exposure.
For patients with head and neck cancer, the relative stable craniocerebral structure may fascinate the usage of individualized 3D-PT and the deviation of needle puncture during RSI maybe prone to be smaller than that of localizing pulmonary nodules. Ming-Wei Huang et al [14] reported 25 patients with head and neck tumors implanted with I 125 radioactive seeds under the assistance of 3D-PT. The mean entrance deviation for all inserted needles was 1.18 ± 0.81 mm varying from 0.857 ± 0.545 to 1.930 ±0.843 mm at different sites and was signi cantly smaller in the parotid and maxillary regions (belong to head region), which is signi cant smaller than that of localizing pulmonary nodules mentioned above and seems similar to that of reported here (0.81-0.98 mm). In the present study, entrance deviation was also signi cantly different in patients with implantation in head and neck region and in patients with primary cancer and secondary cancer in multivariate analysis, but was only larger in patients with primary cancer in univariate analysis. Meanwhile, in the study by Ming-Wei Huang et al [14], the mean angular deviation was 2.08 ± 1.07 degrees varying from 1.85 ± 0.93 to 2.73 ± 1.18 degrees at different sites and was signi cantly larger (indicating less accurate placement) in the sub-mandibular and upper neck area (neck region), than in the other regions (head region), which also seems similar to that reported here (1.56 ± 0.14 degrees). Interestingly, in the current study, angular deviation was lager in patients with cancer in the neck region than in the head region, and also lager in patients with secondary cancer than primary cancers in univariate analysis. However, in multivariate analysis, both planned and intraoperative deviation of depth and angular have no statistical signi cance involving both cancer type and implantation site. This may be owing to the in uence of other potential factors as R 2 of the linear regression turns out to be 0.1. Therefore, whether the accuracy of 3D-PT assisted RSI varies by cancer type or implantation site, further high-quality study is needed before conclusion is drawn.
As for dosimetry pro le, in the above study of Ming-Wei Huang et al [14], the D90 was larger than the planed PD and ranged from 122Gy to 198Gy (mean 163.8 ± 22.6Gy), which seems higher than that reported here (range, 90-170; mean, 136.1 ± 7.7 Gy). The V100 was larger than 95% and the V150 was less than 50% in all patients and other planed and post-plan dosimetric data was not reported and compared in their study. In a study by Ji Z et al [16] comparing the dose distributions of post-plan data with preoperative plan for 3D-PT assisted RSI, a total of 14 patients with malignant tumors (majority located in pelvic cavity) were enrolled. The average post-plan D90, V100, and V150 were smaller than the planned ones, and average post-plan V200 and minimum peripheral dose of GTV were larger than the planned ones, however, there was no statistical difference in any these parameters between the two groups except for V100 (p=0.027). Sun et al [18] compared the dosimetric data between preoperative planand post-plan veri cation in 3D-PT assisted CT-guided RSI for thorax movement tumors. All of the included dosimetry parameters changed slightly, while the difference was also not statistically signi cant (all p > 0.05). Yansong Liang et al [13] reported the dosimetric accuracy of 3D-PT assisted I 125 RSI for the treatment of cervical lymph node metastasis in 15 patients. There was also no signi cant difference for all the parameters (D90, V90, V100, and V150) between preoperative plan and post-plan veri cation (all p > 0.05). Similarly, as also revealed in the current study, the post-plan dosimetry was completely meet the requirements of the preoperative plan involving 3D-PT assisted RSI without signi cant deference.
The present study has several limitations. First, this was a retrospective study and was therefore prone to selection bias. Second, the absence of a control group limits evaluation of the superiority of3D-PT assisted CT-guided RSI over bare-handed CT-guided RSI. Third, the depth and angular of the needle was calculated after fusing the preoperative and intraoperative CT images into the same coordinate axis in the current study, instead of same CT image, suffered from potential fusion error. However, this is the only way to compare preoperative plan with intraoperative data. Fourth, in subgroup analysis for implantation site, further re ned subregion classi cation instead of head and neck, e.g. the parotid and masseter region, maxillary and paranasal region, the retromandibular region, and submandibular and upper neck region, was not applied in the present study, limited by the power of statistics in such small group of patients and with 53.7% of the cancers located in the neck. Finally, there was dilemma in grouping the tumors located at the boundary of head and neck region (submandibular and upper neck area), which may stretch over both head and neck region and were mainly classi ed according to location of the lesion center, leading to potential distraction of the results.

Conclusion
Within the limitation of this study, individualized 3D-PT assisted I 125  Plus-minus data = mean ± standard deviation; number in parentheses = percentage of patients.   Figure 1 A-D:Digital modeling of individualized 3D-printing template (3D-PT) in patient with recurrent head and neck cancer and simulated needle pass way; E-F: The 3D-PT of head and neck with 3 mm thickness contained information such as body-surface characteristics of the treatment area, localization markers, and entrance hole for 18-gauge needle.