Fast intra-fractional image-guidance with 6D positioning correction reduces delivery uncertainty for stereotactic radiosurgery and radiotherapy

Abstract

The purpose of this study is to analyze intra-fractional positioning uncertainty for stereotactic radiotherapy and radiosurgery of cranial tumors. Specifically, we wish to determine the use of intra-fractional image guided patient positioning verification is necessary during delivery of “frameless” stereotactic radiotherapy and radiosurgery (SRT/SRS) and non-coplanar radiation beams, and if positioning uncertainty is associated with overall treatment time. Orthogonal radiographic treatment verification data was extracted for 288 patients and 1344 fractions, and were analyzed with respect to 3D translational and angular corrections once during treatment delivery of SRT/SRS. We find that positioning corrections greater than 2 mm are required for approximately 6% of beams, and that the magnitude of the translational corrections was significantly associated with the delay time between beams (p=0.003). Further, we find that the maximum angular and translational deviations were associated (p<0.001). We conclude that a subgroup of SRT/SRS patients may have considerable positioning error unless this is monitored and corrected during treatment, and that keeping the imaging and delivery times below approximately 5 min is beneficial towards clinically relevant geographical errors. In case longer time-delays than 5 min occurs, the treatment staff should consider acquiring a new set of radiographs in order to verify the patient’s position, assuming this technically feasible to be performed quickly.

1. Introduction

Brain metastases comprise a considerable fraction of patients at our center, and may occur in as much as 20-40% of patients with cancer1. The incidence of brain metastases has been increased in recent years primarily due to the overall improved outcomes for several cancers, but also as a result of the increase availability of diagnostic methods following the introduction of magnetic resonance imaging (MRI)2. The treatment option for brain metastases of stereotactic radio-surgery (SRS) has been proven to be effective in the treatment of brain metastases in numerous reports3. A fundamental aspect of the safe application of SRS lays in the high geometrical precision of treatment delivery, owing to the very sharp dose fall off from the target volume, as well as particularly in the case of SRS: the substantial radiation dose delivery in a single treatment. Accuracy of SRS are generally considered to be approximately 1-2 mm with modern radiation therapy equipment, following appropriate quality assurance and Image-Guided Radiation Therapy (IGRT) treatment protocols4 (and the earlier AAPM Report 54 Schell et al). This level of accuracy has been seen to be possible to achieve using both frame-based and frameless, i.e. (for instance) aqua plastic mask, fixation devices5,6. However, the level of accuracy for the whole population may be acceptable on average, but there might still be a fraction of patients who are inadequately treated due to positional errors that might occur unless positioning is closely monitored. This is due to the fact that treating with one or a few fractions leaves little room for positional errors to average out in the way of conventional multi-fractionated treatment course. Multi-fractionated treatment schedules has enabled radiation oncologists to address positional uncertainty by assigning appropriate Planning Target Volume (PTV) margins, by use of statistical considerations of radiation dose target coverage7. With SRS, however, positioning errors attributable to an entire treatment will likely have a more profound impact on the radiation dose delivered and the resulting biological effect stemming from that than for conventional radiation therapy delivered in multiple fractions. This fact makes it crucial to control the target position prior to or even during treatment, using technology for monitoring the target position8,9 or using e.g. target tracking or beam gating technology10,11, though these techniques might be more crucial for extra-cranial targets.

At our institution, intra-fraction IGRT has been in long use for SRS and cranial SRT and following the introduction of frameless fixation devices. We therefore wished to assess the usefulness of intra-fraction image guided (IG) corrections for SRS and cranial SRT. We hypothesized that a fraction of patients might benefit from intra-fractional IGRT positioning corrections and we wished to quantify how large this fraction might be. Secondarily, we wished to explore if the delay time between beams was associated with large or even catastrophic positioning errors that might occur unless intra-fractional IG is used.

2. Materials and methods

A retrospective analysis of patients treated in our clinic from June 2009 to September 2012 with intra-fraction radiographic image guidance was performed. Treatment data for 288 patients treated for brain metastases using 1344 fractions were analyzed in this study. Patients were scanned using 1 mm slice thickness using MR and CT. The images were subsequently fused and used for target delineation. The patients were normally treated with a single fraction of 18 Gy prescribed to the 90% isodose line. Two different types of patient fixation mask types have been used in the study (“CIP”, Civco Posicast head mask, and “TMB”, Thermoplastic Mask by BrainLab AG). Stereoscopic imaging (ExacTrac®, BrainLab AG, Germany) and 6D couch positioning correction (Robotic couch®, BrainLab, Germany) was performed prior to each treatment beam, which allows precise patient setup. The tolerance limit of 1 mm and 1 degree rotation was used for the intra-fractional positioning verification. If the position was verified to be more than 1 mm translational error in all cardinal axis, and/or if the rotation was verified to be than 1 degree the patient would be repositioned. The new and corrected position would be verified by a second set of verification radiographic images. The patient would be repositioned until using the above procedure until found to be within tolerance.

Three different treatment machines were used for treatment during the period; all were dedicated high-precision stereotactic radiation therapy machines (Novalis Tx, BrainLab, Germany and Varian Medical Systems, USA). All machines were subjected quality assurance procedures following international standards, with pre-treatment verification of plans as well as daily, monthly and yearly service and control of performance.

We evaluated the standard deviation as well as the frequency of positioning errors larger than 1, 2, 3, 4 and 5 mm, which we considered likely to be detrimental to local control of the disease if left uncorrected. Maximum angular in any of 3 rotation axes were derived. The number of maximum angular corrections larger than 1, 2, 3, 4, and 5 degrees were derived. Three-dimensional positioning deviation data and the maximum angular deviations were sorted into three groups based on the time delay between beams. The groups were selected for delay times less than 5 min, 5-10 min and longer than 10 min. Then, we derived the frequency of 3D positioning deviations larger than 2 mm and maximum angular deviations larger than 2 degrees, respectively, in each of the time-delay groups. Statistical calculations and were performed using the R statistical package (version 3.0.1, 2013-05-16).

3. Results

The frequency of the intra-fractional positioning errors is depicted in Figure 1. We observe errors larger than 2 and 3 mm in 6% and 2% of all fractions, respectively, representing a considerable geometrical error without use of intra-fraction image guidance. We note that the remaining error include a registration error, any geometrical distortion of the reconstructed radiographs as well as the verification image radiographs. In addition, any systematic offset of the imaging vs. the treatment isocenter will result in uncorrected positioning errors. Using the pre-treatment and intra-fraction image guidance at our institution, we estimate our overall accuracy to be better than 1 mm for the complete treatment. A total of 840 and 466 treatment beams were evaluable for the CIP and TMB systems, respectively. The average 3D geometrical deviation was 1.04 and 0.95 mm and with standard deviation of 0.71 and 0.74 mm for the CIP and TMB systems, respectively. Similarly, there was no difference in geometrical uncertainty for the three linear accelerators used. The data for all linacs as well as mask fixation devices (i.e. CIP and TMB) were pooled in the subsequent analysis.

Figure 1.

Frequency of (right, A) intra-fraction 3D positioning and (left, B) maximum angular deviations, respectively. The fraction of beams are printed on top of each bar, i.e. 42% and 9%, and 6% and 1%, of the 1344 beams had 3D positioning and angular deviations greater than 1 mm and 1 degree, and 2 mm and 2 degrees, respectively.

A scatter plot between the treatment time and the 3D translational deviation and maximum angular deviation is shown in Figure 2. Our statistical calculations (Table 1) show that no significant correlation between the different projections of the intra-fractional translational positioning 3D deviation. However, the delay time between beams for the patient and the 3D intra-fractional positioning deviation is strongly correlated. Also, as shown in Figure 3, the risk of having a large positioning deviation of 2 mm in 3D or angular deviations larger than 2 degrees is higher for longer delay times (Fisher’s exact test p<0.001). Further, the geometrical 3D translational deviation and is correlated with the maximum angular deviation.

Figure 2.

Three-dimensional translational deviation (left, A) and maximum angular deviation (right, B) vs. delay time between fields. Linear regression yields slope coefficients 0.042 (0.030-0.053, 95% CI) and 0.019 (0.009-0.030, 95% CI) for 3D deviation and Max. angular deviation, respectively.

Table 1.

Spearman correlation coefficients for the treatment beams investigated.

Parameter	Association	Rho	Significance (2-tailed)	N
Lateral dev.	Long. dev.	-0.007	0.785	1344
Vertical dev.	Long. dev.	-0.039	0.157	1344
Lateral dev.	Vertical dev.	-0.012	0.672	1344
Max angular dev.	Delay time	0.092	0.006	910
3D deviation	Delay time	0.376	0.000	910
3D deviation	Max angular dev.	0.174	0.000	1344

Figure 3.

The fraction of three-dimensional translational deviations larger than 2 mm (left, A) and the fraction of maximum angular deviations larger than 2 degrees (right, B), respectively, for delay time between fields grouped for 0-5, 5-10 and 10+ min. The error bars show 95% confidence intervals for the fractions. Fisher’s exact test reveals significance (p<0.001) between the time delay groups for the risk of both large 3D positioning deviations (>2 mm) as well as large angular deviations (>2 degrees).

4. Discussion

In this work, we studied the intra-fraction motion using stereoscopic x-ray imaging during the delivery of SRS and cranial SRT. Our data show that for the vast majority of patients the positioning uncertainty is very small using frameless SRS and cranial SRT. However, by making use of our comparatively very large patient cohort we find that for a limited fraction of patients substantial positioning errors greater than 2-3 mm can be avoided by using intra-fractional position monitoring. Interestingly, we find that a prolonged treatment time was associated with greater risk for treatment positioning errors, both in terms of 3D deviations as well as angular deviations. Based on this finding, we sorted the positioning deviations into groups of 0-5, 5-10 and longer than 10 min time-delay between treatment beams in order to investigate if also the risk was significantly greater for larger positioning deviations to occur between these groups. Here, we considered 2 mm and 2 degrees to be relevant metrics and were referred to as large – which is indeed the case for stereotactic treatments. This analysis could help clinical staff to make informed decisions about when to acquire a new set of verification radiographs for SRS and SRT patients should time-delays occur. Time-delays may occur due to for instance technical problems with the image verification system, treatment delivery system, analysis or control software, patient compliance, etc. We find that time-delays should preferably be kept lower than 5 min in order to maintain a reasonably low risk for positioning deviations, which was significantly lower if the time-delay was less than 5 min. We note, however, that a prolonged treatment time was probably to a certain extent brought about due to the image-guided procedures of the treatment delivery. Unfortunately, our data base does not contain any detailed information of the exact reason or reasons for the time delay introduced.

Our data are reasonably close to those reported by Spadea et al⁹. The authors found somewhat smaller intra-fractional errors of about 0.5 mm using IR-optical localization and similar errors of about 1 mm using x-ray verification. However, Spadea et al estimated the positioning errors before and after treatment, while we performed several verifications during treatment after couch rotation so our collected data are not directly comparable.

5. Conclusions

The risk of a major intra-fraction positioning error (>3 mm) occurring appeared to be small (approximately 2% of the fractions) but non-negligible without intra-fractional image guidance for our SRT and SRS patients utilizing a couch rotation. The positioning errors were similar for both mask types with an average geometrical 3D deviation of 1.0 mm and standard deviation of 0.7 mm. This implies the necessity of using image guided patient positioning correction in case the couch-top was moved between the treatment fields. Statistical analysis shows that patient delay time between beams is strongly correlated to the three-dimensional intra-fractional translational deviation. This suggests that the total patient imaging, image processing as well as the image registration and including the treatment time should be minimized as much as possible in order to bring patient positioning uncertainty to its minimum. We find that treatment staff should obtain a new set of verification radiographs should a time-delay longer than 5 min occur, assuming this verification is technically feasible to be performed quickly.