Intra‐ and inter‐fractional variations of tumors with fiducial markers measured using respiratory‐correlated computed tomography images for respiratory gated lung stereotactic body radiation therapy

Abstract Purpose This study evaluated the intra‐ and inter‐fractional variation of tumors with fiducial markers (FMs), relative to the tumor‐FM distance, to establish how close an FM should be inserted for respiratory‐gated stereotactic body radiation therapy (RG‐SBRT). Methods Forty‐five lung tumors treated with RG‐SBRT were enrolled. End‐expiratory computed tomography (CT) (CTplan) and four‐dimensional‐CT (4D‐CT) scans were obtained for planning. End‐expiratory CT (CTfr) scanning was performed before each fraction. The FMs were divided into two groups based on the median tumor‐FM distance in the CTplan (Dp). For the intra‐fractional variation, the correlations between the corresponding tumor and FM intra‐fractional motions, defined as the centroid coordinates of those in each 0–90% phase, with the 50% phase of 4D‐CT as the origin, were calculated in the left‐right, anterior‐posterior, and superior‐inferior directions. Furthermore, the maximum difference in the tumor‐FM distance in each phase of 4D‐CT scan, based on those in the 50% phase of 4D‐CT scan (Dmax), was obtained. Inter‐fractional variation was defined as the maximum distance between the tumors in CTplan and CTfr, when the CT scans were fused based on each FM or vertebra. Results The median Dp was 26.1 mm. While FM intra‐fractional motions were significantly and strongly correlated with the tumor intra‐fractional motions in only anterior‐posterior and superior‐inferior directions for the Dp > 26 mm group, they were significantly and strongly correlated in all directions for the Dp ≤ 26 mm group. In all directions, Dmax values of the Dp ≤ 26 mm group were lower than those of the Dp > 26 mm group. The inter‐fractional variations based on the Dp ≤ 26 mm were smaller than those on the Dp > 26 mm and on the vertebra in all directions. Conclusions Regarding intra‐ and inter‐fractional variation, FMs for Dp ≤ 26 mm can increase the accuracy for RG‐SBRT.


INTRODUCTION
Stereotactic body radiation therapy (SBRT) delivers a precise high dose of radiation to a limited target volume.][3][4][5][6] Lung tumors can move by more than 30 mm. 7 In lung SBRT, the tumor motion during breathing results in significant geometric uncertainty during high-dose delivery to the target.This geometric uncertainty can be addressed by the introduction of an expanded internal target volume (ITV).][10] To reduce geometric uncertainty and adverse event risk, SBRT for lung tumors with severe respiratory motion requires appropriate motion management. 113][14] In these respiratory motion management methods, fiducial markers (FMs), inserted either in the tumor itself or in close proximity, are often used as internal surrogates to localize the tumor. 15,16t our institution, respiratory-gated SBRT (RG-SBRT) for lung tumors is performed using the real-time tumor monitoring system. 17,18This method can be used to monitor FM during irradiation using two fluoroscopic images.The treatment beam to the target was turned on only when the selected FM was located within a few millimeters of the planned three-dimensional position of the FM.RG-SBRT is effective in reducing ITV.To achieve highly precise RG-SBRT using a real-time tumor monitoring system, the positional accuracy of the FM as an internal surrogate for locating the tumor is important.
Several authors have reported the intra-fractional variation between the FM used as an internal surrogate and the lung tumor during respiration. 19,20Inter-fractional variations of lung tumors with respect to FM due to various factors, such as tumor distortion by daily radiation therapy and FM migration, have also been reported. 20,21n lung RG-SBRT, the inter-fractional variation as well as the intra-fractional variation of the lung tumor can result in geometric uncertainty in dose delivery to the tumor, increasing the target volume and the risk of adverse events if an inappropriate FM is used as an internal surrogate.
An FM is often implanted using an endobronchial approach because of its minimally invasive nature and short intervention time. 16,22In this approach, the locations where the FMs can be placed are limited because the FMs are implanted along the small bronchi.Physicians try to place the FM transbronchially close to the lung tumor; however, it is difficult to place the FMs within a few millimeters of the tumor.A few studies have shown that the intra-or inter-fractional variation of lung tumors correlates with the distance between the lung tumor and FM. 23,24However, little is known about how closely the FM needs to be inserted into lung tumors during SBRT.
Herein, we aimed to quantitatively evaluate the intraand inter-fractional variations in lung tumors with FMs and compared the intra-and inter-fractional variations based on the distance between the lung tumor and FM.

Patients
Forty-three patients treated for one-two lung tumors with RG-SBRT in the end-expiratory (EE) phase during free-breathing using a real-time tumor monitoring system between April 2017 and August 2021 were enrolled in this study.Patients requiring home oxygen therapy at the time of treatment were excluded.Patient characteristics are presented in Table 1.
Prior to computed tomography (CT) simulation, 2−6 FMs were inserted near each tumor using bronchoscopy.The gold marker with diameter of 1.5 mm (Disposable Gold Marker; Olympus Medical Systems, Tokyo, Japan) was used as the FM.The FMs that dropped out before the CT simulation were excluded from the evaluation.Residual FMs were evaluated at the time of the CT simulation.
The Institutional Review Board approved this study and the requirement for written informed consent was waived because of its retrospective design.

CT data acquisition and contouring
A CT scanner (SOMATOM Definition AS; Siemens Healthcare, Erlangen, Germany) was used for CT imaging.One to two weeks after FM placement (median,7 days), CT simulations were performed for treatment planning, and each patient was immobilized in the supine position using a Vac-Lok system (CIVCO Medical Solutions, Coralville, Iowa, USA).Four-dimensional-CT (4D-CT) under free-breathing and breath-hold CT in the EE phase (CT plan ) scans were obtained with a 2 mm slice thickness for planning.The CT scans were acquired while the respiratory phase was monitored using a real-time position-management system (Varian Medical Systems, Palo Alto, California, USA).Breath-hold CT in the EE phase (CT fr ) was also obtained with a 2-mm slice thickness before each fraction to confirm FM migration.If cone-beam CT (CBCT) scanning was substituted to confirm FM migration, it was not included in the CT fr .
All CT image sets were imported into Eclipse treatment planning system (Varian Medical Systems, Palo Abbreviations: CT fr , computed tomography scans obtained at the end-expiratory phase before each fraction to confirm FM migration; FM, fiducial marker. Alto, California, USA).All FMs and lung tumors in 10 phases of 4D-CT and CT plan and lung tumors in each CT fr sets were contoured with the treatment planning system by a radiation oncologist with 8 years of experience.
The centroid coordinates of the contoured FMs and lung tumors were acquired from each CT image.The three-dimensional distance between the centroid of the tumor and that of the FM in CT plan (D p ) was calculated for each FM.The FMs were divided into two groups based on the median D p .When D p was greater than 95 th percentile of D p , the FMs were excluded from the D p groups.

Evaluation of the intra-fractional variation using 4D-CT
In this study, the intra-fractional variation was defined as the intra-fractional motion of the tumor relative to the FM during respiration.4D-CT scans were used to assess the intra-fractional variation.Two items were evaluated for intra-fractional variation: the correlation between tumor and FM intrafractional motions and the difference between tumor and FM distance in each phase of 4D-CT scanning (Figure 1).
First,the centroid coordinates of the tumor and the FM in each 0−90% phase with the 50% phase of the 4D-CT scan as the origin were calculated as the tumor and the FM intra-fractional motion.The correlation between the corresponding tumor and FM intra-fractional motion was determined in the left-right (LR), anterior-posterior (AP), and superior-inferior (SI) directions for each FM group based on the median D p .
Next, the distance (D n% ; 0 ≤ n ≤ 90) between centroids of the tumor and FM was acquired in each 0−90% phase of 4D-CT scan in the LR, AP, and SI directions.For each FM, the maximum difference (D max ) between D n% and D 50% was calculated in each direction.The D max was compared between the two FM groups based on the median D p .

Evaluation of the inter-fractional variation using breath-hold CT at EE
CT plan and CT fr scans were used to assess the interfractional variation.First, CT fr scans were registered in CT plan based on the vertebra.The distances between the centroids of the tumors on CT plan and CT fr scans were calculated in the LR, AP, and SI directions.In each tumor, the maximum distances among all CT fr -based datasets were calculated as inter-fractional variations based on the vertebra.
Next, CT fr scans were registered in CT plan based on each FM.The distances between the centroids of the tumors in CT plan and CT fr scans registered based on the FM were calculated in each direction.In each FM, the maximum distances among all CT fr -based datasets were calculated as the inter-fractional variations based on the FM (Figure 2).
The inter-fractional variation values in each direction were compared between the three groups based on the two D p groups and the vertebra group.If the FM had migrated during the treatment period, it was evaluated up to that point.
F I G U R E 1 Definition of the tumor and fiducial marker (FM) intra-fractional motions (a), and D max for the evaluation of the intra-fractional variation (b).Two items were considered in the evaluation of intra-fractional variation using 4D-CT scans: the correlation between tumor and FM intra-fractional motions, and the difference between tumor and FM distance in each phase.(a) Tumor and FM intra-fractional motions were defined as the centroid coordinates of the tumor and FM in each 0−90% phase, with the 50% phase of the 4D-CT scanning used as the origin.(b) D max is defined as the maximum difference in the distance between the tumor and FM in each phase of 4D-CT scanning based on the distance between the tumor and FM in the 50% phase of 4D-CT scanning.
F I G U R E 2 Definition of the inter-fractional variation.Inter-fractional variation was defined as the maximum distance between the centroid of the tumors captured on CT images at the end-expiratory phase for treatment planning (CT plan ) and before each fraction (CT fr ) among CT fr -based datasets when those CT scans were registered based on each fiducial marker (FM) or vertebra.

Analysis
All statistical analyses were performed using JMP ver16.1.0(SAS Institute Inc., Cary, North Carolina, USA).Pearson's correlation coefficient was used to evaluate the correlation between the tumor and the FM intrafractional motion.A strong correlation was defined as a Pearson's |R| > 0.70.The Wilcoxon rank-sum test was used to compare D max values between the two D p groups. 25 To compare inter-fractional variation between the three groups, based on the two D p groups and the vertebral group, the Kruskal-Wallis test with Steel-Dwass test was used for analysis.Statistical significance was set at p-values of < 0.05.

Inserting and grouping the FMs
Overall, 167 FMs were inserted near lung tumors using the endobronchial approach.One patient experienced pneumothorax as a side effect of FM insertion, but recovered on their own.Further, 16 FMs dropped out before the CT simulation.

Intra-fractional variation
D max values and correlation between tumor and FM intra-fractional motions were evaluated.Figure 3 shows the scatter plot and Pearson's correlation coefficients between the corresponding tumor and FM intrafractional motions.For the D p ≤ 26 mm group, the FM intra-fractional motions were significantly and strongly correlated with the tumor intra-fractional motions in the LR (R = 0.77, p < 0.001), AP (R = 0.88, p < 0.001), and SI (R = 0.96, p < 0.001) directions.The FM intra-fractional motions were significantly and strongly correlated with the tumor intra-fractional motions only in the AP (R = 0.83, p < 0.001) and SI (R = 0.92, p < 0.001) directions in the D p > 26 mm group.In the LR direction, FM intra-fractional motion was significantly and moderately correlated with tumor intra-fractional motion in the D p > 26 mm group (R = 0.53, p < 0.001).
Figure 4 shows the D max values in the two D p groups.The median (quartile) D max values in the LR, AP, and SI directions were 1.0 mm (0.6−1.3), 1.2 mm (0.8−2.0), and 2.0 mm (1.4−3.1) in the D p ≤ 26 mm group, and 1.2 mm (0.8−1.9), 1.6 mm (1.0−2.2), and 2.7 mm (1.5−3.7) in the D p > 26 mm group, respectively.The D max values in the D p ≤ 26 mm group were smaller than those in the D p > 26 mm group in the LR (p = 0.031), AP (p = 0.027), and SI (p = 0.048) directions.
The inter-fractional variation values in the D p ≤ 26 mm group were significantly smaller than those in the D p > 26 mm group in the LR (p < 0.001), AP (p < 0.001), and SI (p = 0.044), respectively.The interfractional variation values in the D p ≤ 26 mm group were also significantly smaller than those in the vertebral group in the LR (p = 0.003), AP (p < 0.001), and SI (p < 0.001) directions.The inter-fractional variations in the D p > 26 mm group were significantly smaller than those in the vertebral group in the AP (P = 0.017) and SI (p < 0.001) directions.In the LR direction, the inter-fractional variation was not significantly different between the D p > 26 mm and vertebra groups (p = 0.966) (Figure 5).
Among the residual FMs at the time of CT plan imaging, six FMs migrated or dropped out of the CT fr scanning during the treatment period.

DISCUSSION
In this study, the FM was inserted using an endobronchial approach.As this method involves the placement of the FM in the bronchus, it may be difficult to place the FM near the tumor.Another common method of FM insertion is the transcutaneous approach.The transcutaneous approach has the potential to place FMs closer to the tumor compared to the endobronchial approach.However, the transcutaneous approach has a higher risk of pneumothorax as a complication of chest tube treatment. 26,27Meanwhile, the endobronchial approach has a lower risk of pneumothorax. 16,23,27,28n the present study, only one case of pneumothorax improved with follow-up.The endobronchial approach is preferable for the minimally invasive placement of FMs.
Willmann et al. 29 reported that tumor and FM motions correlated well in the AP and SI directions in their study using 4D-CT scanning.Our results are comparable to these results.Furthermore, in our study, a correlation between tumor and FM intra-fractional motion was observed in the LR direction.A correlation between tumor intra-fractional motion and FM intra-fractional motion was also observed in the D p > 26 mm group, but the correlation was found to decrease, especially in the LR direction, in our study.Only a few studies have indicated the specific distances between tumors and FMs. 25,30,31Akasaka et al. 25 analyzed the three-dimensional distance of the coiled FM to the lung tumor in CT scans for planning at the expiratory and inspiratory phases (D diff ) and the ratio of ITV to gross tumor volume (GTV) (ITV/V GTV ) created by 4D-CT and CT scans at the inspiratory phase for treatment planning.They showed that D diff and ITV/V GTV were smaller when the distance between the edges of the tumor and the FM was less than 10 mm.Our study defined the distance between the tumor and FM as the distance between both centroids, while they defined the distance between both margins (i.e., closest distance).Therefore, regarding the difference between the 26 and 10 mm distances, it is necessary to consider the size of the tumor and the FM.In a preliminary study of FMs using 4D-CT scans in eight patients, Yamasaki et al. 30 reported that misalignment in the expiratory phase was significantly greater when the distance between the FM and the lung tumor was 30 mm or greater than when it was less than 30 mm.They also reported that the misalignment in the expiratory phase was smaller than 2.5 mm when the distance between the FM and lung tumor was less than 25 mm in their study.They concluded that the FM should be inserted within approximately 25 mm from the lung tumor.Smith et al. 31 analyzed the motion correlation between lung tumors and the surrounding tissue in 10 lung cancer patients with deformable registration between the EE and end-inspiratory phase on 4D-CT scans.They suggested that an FM inserted within 30 mm of the lung tumor represented tumor motion with an error within 3 mm in many patients.These reports 25,30,31 on intrafractional variation support our findings: D max values of the FMs for D p ≤ 26 mm were smaller than those of the FMs for D p > 26 mm in all directions.Considering our results, the prior reports on intra-fractional variation, 25,30,31 and the distance from the tumor where the FM can be inserted into the bronchus, 26 mm may be a reasonable distance between the FM and the lung tumor.
Furthermore, the present study showed that the interfractional variations in the D p ≤ 26 mm group were also significantly smaller than those in the D p > 26 mm group in all directions.As the inter-fractional variation as well as the intra-fractional variation of the lung tumor can affect the ITV, the FM for D p ≤ 26 mm can be expected to reduce the ITV in terms of both intra-and inter-fractional variations.Therefore, a tumor-FM distance of 26 mm or less can be one of the distance criteria for the FM placement in terms of both intra-and inter-fractional variation values of the lung tumor.No other studies have examined inter-fractional variation with respect to the specific distance between the lung tumor and the FM.Roman et al. 24 suggested that smaller distances between the lung tumor and FMs resulted in smaller interfractional variation values in seven patients with locally advanced lung cancers.This report supports our study findings.
Image-guided set-up based on the FM is considered to have less positional variation than that based on the vertebra. 32The present study showed similar results.Furthermore, our simulated study using CT images showed that the inter-fractional variation with a set-up based on an FM was smaller than that with a set-up based on the vertebra even when the distance between the tumor and FM was greater than 26 mm (D p > 26 mm).According to our results, even if the FM is not implanted within 26 mm from the tumor, the FM has a certain efficacy in terms of the reduction of the interfractional variation.However, in the present study, there was no significant difference in inter-fractional variation in the LR direction between the set-up based on the vertebra and FM in the D p > 26 mm group.This finding may be due to the lower correlation in the LR direction than in other directions between the FM intra-fractional motion of D p > 26 mm and tumor intra-fractional motion.
Our study has some limitations.Motion artifacts in respiratory-correlated CT images can affect contouring.In addition, we used CT images with 2 mm slice thickness to evaluate the position of 1.5-mm the diameter FM.The CT slice thickness can also affect contouring.Contouring uncertainty may lead to overestimation or underestimation of tumors and FMs.Furthermore, the poor reproducibility of breath-holding with the use of expiratory breath-holding CT scans in the assessment of inter-fractional variation may have affected the results. 20This limitation can be improved using 4D-CT or 4D-CBCT scanning. 23However, a problem of increased radiation exposure exists. 33

CONCLUSION
This study evaluated intra-and inter-fractional variations based on the distance between lung tumors and FMs.A distance of 26 mm or less between the FM and lung tumors improved both intra-and inter-fractional variations.When selecting the appropriate FM for lung SBRT, a distance of 26 mm or less between the lung tumor and the FM can be one of the distance criteria.

AU T H O R C O N T R I B U T I O N S
Yuki Manabe and Takehiro Shiinoki designed this study.All authors contributed to the interpretation of data and revision of the manuscript.Yuki Manabe wrote the manuscript with the support of Takehiro Shiinoki and Hidekazu Tanaka.All authors approved the final report for publication and agreed to be accountable for all aspects of this work.

F I G U R E 3
Scatter plot and Pearson's correlation coefficients between the corresponding tumor and fiducial marker (FM) motions defined as the centroid coordinates of the tumor and FM in each 0−90% phase with the 50% phase of the 4D-CT scan as the origin.Those (a) in the left-right, (b) the anterior-posterior, and (c) the superior-inferior directions for the D p ≤ 26 mm group.(d) Left-right, (e) anterior-posterior, and (f) superior-inferior directions in the D p > 26 mm group.

F I G U R E 4
Maximum differences of the distance between tumor and fiducial marker (FM) (D max s) in the D p ≤ 26 mm and in the D p > 26 mm groups (a) in the left-right, (b) the anterior-posterior, and (c) the superior-inferior directions.D max was defined as the maximum difference in the distance between the tumor and the FM in each phase of 4D-CT scanning, based on the distance between the tumor and the FM in the 50% phase of 4D-CT scans.

F I G U R E 5
Differences of the inter-fractional variations based on the D p ≤ 26 mm group, on the D p > 26 mm group, and on the vertebra (a) in the left-right, (b) the anterior-posterior, and (c) the superior-inferior directions.Inter-fractional variation was defined as the maximum distance between the centroid of the tumors in CT scans at end-expiration for treatment planning (CT plan ) and in CT scans at end-expiration before each fraction (CT fr ) among CT fr -based datasets when the CT scans were fused based on each fiducial marker (FM) or vertebra.

TA B L E 1
Characteristics of patients and tumors.

Table 2
presents the characteristics of patients in the FM group.The median (range and 95 th percentile) value of the D p was 26.1 mm (4.8−67.0 and 58.4 mm).Seven FMs were excluded because those D p values were larger than the 95 th percentile of D p (i.e., larger Characteristics of FM groups. than 58.4 mm).Seventy-four FMs were divided into the D p ≤ 26 mm group and D p > 26 mm group (Table2).