Efficient and accurate stereotactic radiotherapy using flattening filter free beams and HexaPOD robotic tables

Abstract

Flattening filter free (FFF) high dose rate beam technique was introduced for brain stereotactic radiosurgery (SRS) and lung Stereotactic Body Radiotherapy (SBRT). Furthermore, a HexaPOD treatment table was introduced for the brain SRS to enable correction of rotational setup errors.

19 filter flattened (FF) and 57 FFF brain SRS treatments, as well as 451 FF and 294 FFF lung SBRT treatments were evaluated to assess differences in intrafractional motion (IFM) between FF and FFF treatments. For brain SRS the accuracy of rotational corrections was assessed as well.

For SRS the treatment time was reduced by 21%, and for SBRT the treatment time was reduced by 25%. However, only for SBRT the IFM was significantly reduced, from 1.9 mm to 1.6 mm. For brain SRS, table correction in 6D greatly improves patient setup precision observed as a reduction in mean residual rotational setup error from 0.3° (SD1.2°) to 0.06° (SD 0.3°).

1. Introduction

The central issue in radiotherapy is the balance between the desired effect, shrinkage or eradication of tumour manifestations, and normal tissue complications arising from irradiation of healthy tissue. Serious normal tissue complications typically arise either when large proportion of the volume of a parallel organ, e.g. the lung, is irradiated or when any part of a serial organ, e.g. the optical nerve, receives a high dose [1].

For both intra- and extracranial stereotactic radiotherapy, precise and accurate delineation, planning, image guidance, and treatment techniques are used to allow planning and delivery of very conformal dose distributions. This minimises the irradiated volume of healthy tissue surrounding the tumour, and allows a separation of tumour and critical structure to ensure that even critical structures close to the intended target receives doses well below the treatment dose. In this way, stereotactic radiotherapy can deliver very high doses to small treatment volumes while unwanted side effects are minimised.

Even in this precise and accurate stereotactic regime, a variety of causes lead to spatial uncertainties. Image guided treatment (IGRT) on a linear accelerator is typically preceded by cone beam CT (CBCT) imaging which is compared with a reference CT scan allowing for discrepancies to be corrected by moving the treatment table [2]. However, the table motion is not perfect [3] and is traditionally restricted to translations in the 3 orthogonal directions. Rotational components are typically not corrected; instead the rotational localisation error is sought reduced by careful immobilisation of the patient [4]. Immobilisation also minimises intrafractional motion (IFM) of the patient. However, it is impossible to completely avoid this uncertainty in stereotactic body radiotherapy (SBRT) and frame­less stereotactic radiosurgery (SRS), in part due to long treatment times caused by large fractional doses.

In the presence of spatial uncertainties, either one has to accept the resulting risk of a geographical miss or one has to introduce margins in the treatment planning [5]. In the former approach, the risk increases with larger uncertainties, while in the latter the margins increase with larger uncertainties, especially in treatments with very few fractions. Larger margins invariably increases the volume of healthy tissue exposed to high doses and impedes the dosimetric separation of target and critical structures. In both cases immediate benefits come from minimising the setup uncertainties, as these may inhibit successful stereotactic radiotherapy.

A linear accelerator equipped with a HexaPOD table allows the correction of rotational errors found by CBCT and thus reduces the residual positioning error prior to treatment. Introduction of the Flattening Filter Free (FFF) treatment technique reduces the beam-on time by allowing a high dose rate [6] and may in this way reduce the risk of large IFM. The purpose of this study was to evaluate the effect of rotational table corrections on treatment accuracy and to evaluate the effect on intrafractional motion by the introduction of FFF in patients treated with frameless brain SRS and lung SBRT.

2. Materials and Methods

All patients from our institution receiving stereotactic radiotherapy as described below, treated with Volumetric Modulated Arc Therapy (VMAT) and with IGRT co-registrations available were included in this study. They were all planned in the Pinnacle treatment planning system and treated on Elekta linear accelerators with Precise table. The SRS treatments were performed using a HexaPOD tabletop.

2.1 Brain SRS treatment

This study included 19 filter-flattened (FF) and 57 FFF treatments, in which small brain tumours/metastases were treated to 20 Gy in 1 fraction between March 2013 and August 2014. The typical treatment consisted of three VMAT arcs: One full 360° arc and two non-coplanar arcs of 220°, with table rotation of ± 45°. These patients were immobilised with the Orfit HP system with 3-point Efficast thermoplastic masks [4]. Patient position was determined from a pre-correction CBCT in the treatment isocenter, and no patient required table motion to avoid collision during CBCT. Co-registration to the reference CT was performed using an automatic bone match of a region covering the entire skull. All CBCT registrations used the same settings. Patient position was corrected in 6 dimensions (6D) using the HexaPOD table, which have specified precision as better than 0.5mm/0.5° (95 % CI) [7]. Before treatment the patient position was verified with a post-correction CBCT. Halfway through each fraction a mid-fraction CBCT was carried out. This IGRT treatment sequence is illustrated in Figure 1. The pre-treatment effect of rotational patient correction was estimated by the difference in the rotational displacements (pitch, roll and yaw) between pre-correction and the post-correction scan. Translational and rotational IFM was estimated by the difference between the post-correction and the mid-fraction scan.

Figure 1.

IGRT workflow diagram for SRS.

2.2 Lung SBRT treatment

Small peripheral lung tumours were treated in 3 fractions, with the GTV receiving 66 Gy and the PTV receiving 45 Gy according to the Scandinavian SBRT phase II trial SPACE protocol [8]. 451 FF and 294 FFF treatment fractions between April 2010 and August 2014 were analysed. The typical treatment consisted of two VMAT semi-arcs of 200° and 180°, both avoiding the contralateral lung. All treatment planning and IGRT was performed on the mid-ventilation phase of a CT scan performed on a freely breathing patient [9]. Patients were immobilised with a VacFix bag with arms positioned above the head. Patient position was determined from a 4D-CBCT [10, 11] in the treatment isocenter, and no patient required replanning or table motion to avoid collision during CBCT. Co-registration to the reference CT was performed using a dual match, where the bulk patient positioning is validated using an automatic grey value match of a region covering the thoracal spine. Subsequently, an automatic grey value match of the tumour region was performed and the Precise table was translated to align the tumour region. A clinically unacceptable deviation of the bulk patient would lead to a re-immobilization of the patient, but this did not occur in any of these cases. All CBCT registrations used the same settings. Halfway through each fraction, between the two semi-arcs, another 4D-CBCT was performed. This sequence is illustrated in Figure 2. Translational IFM was estimated by considering the difference between the pre-treatment scan and the mid-fraction scan.

Figure 2.

IGRT workflow diagram for Lung SBRT

2.3 Statistics

Normality of the translational and rotational coordinate distributions was evaluated in each main direction by use of Q-Q plots. The pooled group of all translational coordinates, as well as the pooled group of all rotational coordinates were evaluated likewise. Levene’s test was used to test for differences in variances, and a t-test was used to test for differences in means. Differences in vector length of the translational IFM for FF and FFF was analysed by a Mann-Whitney U test. Statistical analysis was performed in SPSS v.21.

3. Results

The Q-Q plots did not indicate any deviations from normality.

3.1 Brain SRS treatment

The pre-correction rotational displacements of brain SRS patients were distributed with a 0.3° mean and 1.2° standard deviation (SD). Post-correction the mean was 0.06° and the SD 0.3°. The distributions are shown in the histogram in Figure 3. Both the difference of the mean values and SD between pre-correction results and post-correction were statistically significant (p<0.001). There was no statistical difference between the distributions of the post-correction scan and the mid-fraction scan.

Figure 3.

Relative histogram of angular discrepancies for SRS. The three different angles as given by the XVI system are pooled, and both FF and FFF are included.

Introduction of the FFF modality for brain SRS reduced the time from the check scan to the mid-fraction scan from 6:01 min (mean for FF treatments) to 4:30 min (mean for FFF treatments) due to the reduced beam-on time with FFF. The median translational IFM changed from 0.4 mm to 0.5 mm while the 90 percentile changed from 0.7 mm to 1.0 mm. The mean rotational IFM was 0.08° for both FF and FFF, while the SD changed from 0.35° to 0.29°. The distributions of translational and rotational IFM are shown in Figure 4 and 5, and a histogram of the measured vector length of the translational IFM is shown in Figure 6. The differences between FF and FFF in IFM were not statistically significant.

Figure 4.

Relative histogram comparing translational IFM for FF and FFF for SRS. The three orthogonal directions are pooled.

Figure 6.

Relative histogram of the measured vector length of the translational IFM for SRS for FF and FFF.

3.2 Lung SBRT treatment

Introduction of FFF on lung SBRT reduced the time from the pre-correction scan to the halfway scan from 9:46 min (estimated mean for FF treatments) to 7:44 min (estimated mean for FFF treatments). The median vector length changed from 1.9 mm to 1.6 mm, while the 90 percentile changed from 4.2 mm to 3.7 mm. A histogram of all IFM coordinates and of the IFM vector length is shown in figures 7 and 8. The difference in IFM vector length was statistically significant (p=0.011).

4. Discussion

We have found that for brain SRS, table correction in 6D greatly improved patient setup precision, whereas introduction of the Flattening Filter Free treatment modality did not reduce Intra-Fractional Motion. For lung SBRT, introduction of FFF lead to a small but statistically significant reduction of the IFM.

Figure 5.

Relative histogram comparing rotational IFM for FF and FFF for SRS. The three different angles as given by the XVI system are pooled.

Figure 7.

Relative histogram comparing translational IFM for FF and FFF for lung SBRT. The three orthogonal directions are pooled.

4.1 Brain SRS treatment

A recent meta-analysis indicated that at least for certain groups of patients, SRS without whole brain irradiation should be the preferred treatment [12, 13], and it has been recommended by ASTRO not to routinely add adjuvant whole brain radiation therapy to stereotactic radiosurgery for limited brain metastases as a part of the ‘Choosing wisely’ campaign [14]. It must be expected that SRS without whole brain irradiation will be used increasingly henceforth. To reduce the workload in the presence of multiple brain metastases, it is desirable to treat more than one brain metastasis simultaneously. When multiple loci are treated simultaneously with a single isocenter, target coverage will be compromised by even small uncorrected rotational displacements, e.g. an uncorrected rotational deviation of 2° leads to a 1 mm misalignment 30 mm from the isocenter.

Rotational displacements can be minimised by good immobilisation, and frame based solutions have previously been the gold standard for SRS. However, frame based SRS requires invasive procedures, is time consuming, and uncomfortable for the patient. It is therefore of interest to test frameless approaches. Kataria et al. showed that good frameless procedures can be on par with frame based solution [15]. Our frameless immobilisation in combination with CBCT and the HexaPOD Table 6D corrections obtained setup accuracies and measured IFM better than or similar to published results [15, 16, 17, 18].

Introduction of FFF for the SRS treatments more than doubled the dose rate [6], but only reduced the time from the post-correction CBCT to the mid-fraction CBCT by approximately 25% in our clinic. This is less than reported by Stieler et al. [19], but that study only considered beam-on time which was approximately halved. The time required to obtain the CBCT and compare to the reference is unchanged by FFF, and therefore the overall impact of FFF on treatment time is less than indicated by the reduction in beam-on time. No change occurred in the IFM due to this reduction in treatment time.

The IFM for the SRS treatments was small, and the mid-fraction displacement distribution was indistinguishable from the displacement distribution in the post-correction scan immediately prior to treatment. This suggests that the magnitude of the intrafractional motion for these patients is not larger than the setup uncertainty for this treatment. In fact, the measured IFM for FF treatments in the present study of 0.4 mm/0.35° (SD) is a combination of intrafractional motion and uncertainties in the CBCT reconstruction, the co-registration and in the HexaPOD table motion, and when this is compared to the accuracy results for phantom studies with CBCT and HexaPOD of 0.13 mm/0.4° (SD) [20 (Table 5)], where IFM was vanishing, it can be argued that we are close to the limit of detection of the intrafractional motion.Consequently, it is unlikely that the faster FFF treatment will lead to a reduced IFM.

4.2 Lung SBRT treatment

Introduction of FFF on lung SBRT reduced the time from the pre-treatment 4D-CBCT to the mid-fraction 4D-CBCT by approximately 20%. This reduction is far smaller than reported by Prendergast et al. [21], but the large differences in absolute treatment time suggests that they used very different treatment and imaging techniques , and therefore a direct comparison with the present study may be misleading.

There was a small, but statistically significant, reduction in IFM by the reduction in treatment time. The magnitude of the IFM in this study was slightly smaller than results published by Peguret et al. [22]. Peguret et al. matched a 3D CBCT scan is matched to an average intensity 4D CT dataset, whereas the present study matched the mid-ventilation phase of a 4D CBCT to mid-ventilation phase of a 4D CT dataset. This complicates a direct comparison of the two results.

In line with the current study, both Purdie et al. [23] and Shah et al. [24] concluded that shorter treatment times lead to smaller IFM for SBRT in lung. Purdie et al. reported that treatments with a duration of 14-34 minutes had a mean IFM of 2.2 mm, while a treatment duration of 34-58 minutes gave a mean IFM of 5.3 mm. Shah et al. found that for treatments with an IFM of less than 2 mm, the mean treatment duration was 20.3 minutes, while treatments with an IFM of more than 2 mm had a mean duration of 21.5 minutes. The time between the post-correction CBCT and the mid-fraction CBCT in the present study (corresponding to half a treatment) was shorter than the treatment times given by these two studies, for FF (9:46 min) as well as for FFF (7:44 min) treatments, and the IFM was smaller as would be expected (1.9 mm and 1.6 mm, resp.)

Although the clinical effect of the median IFM is probably small, there may still be a need to determine outliers. As recognised on Figure 8, there were a small proportion of treatments, where the IFM exceeded 1 cm. This likely stems patient motion, e.g. from coughing, that is not observed by the treatment staff and is a valid argument to continue to perform the mid-fraction CBCT.

Figure 8.

Relative histogram of the measured vector length of the translational IFM for Lung SBRT for FF and FFF.

4.3 Concluding remarks

For both lung and brain the beam-on time was reduced significantly by the introduction of FFF. Further reductions in overall treatment times are likely to come from reductions in the time used for imaging, co-registration, and validation, and it becomes desirable to reduce the time spend on imaging if feasible without reducing the quality of the immobilisation. As an alternative to the mid-fraction CBCT, IFM might be evaluated by simultaneous CBCT imaging as discussed by Poulsen et al. [25]. Such an approach may lower the overall treatment time and still be able to detect gross IFM.

Finally, it must be pointed out that IFM as measured in this study does not fully describe all uncertainties involved in stereotactic treatment. Potential sources of errors in the acquisition and transfer of the reference CT, delineation and treatment planning as well as actual dose delivery is not considered in this study. Delineation studies and end to end tests are necessary in order to quantify the total uncertainties involved in the process from CT to dose delivery.

5. Acknowledgments

One of the authors (CB) acknowledges support from AgeCare (Academy of Geriatric Cancer Research), an international research collaboration based at Odense University Hospital, Denmark.