Average motion range
This study provides new information on correlated PT and associated LN respiratory motion during radiotherapy. The magnitude of motion averaged over the treatment course was comparable for PT and LN with large interpatient variations. As observed in other reports, the major trajectory of motion was in craniocaudal direction both for PT and LN [
4,
7,
9,
12,
13,
15,
17,
19]. The amount of respiratory motion with on average 4 mm in z-direction for PT and LN in our study appeared smaller than in other reports with average motion ranges up to 11 mm for PT and up to 7 mm for LN [
8,
9,
12]. Larger PT motion ranges have been reported for small peripheral tumors [
4,
5] and lower lobe tumors [
3,
15,
24], whereas in our population of stage IIIA lung cancers no extreme motion ranges were observed, likely due to adherence or invasion of the mediastinum. It is well known that the range of LN motion depends on the LN location [
9,
13]. In the present analysis, LN location was, with one exception, supracarinal. The observed 3D motion of 5 mm agrees well with reports of about 5 mm motion for region 4 LNs [
9,
11,
19]. The average motion for carina was the same as for LN and comparable to 5 mm reported by van der Weide [
18].
Relative respiration-related displacement has so far not been investigated except for a report by Piet et al. [
10] who identified relative motion between C and LN as a potential cause for low yield rates with transbronchial biopsy. The displacement between LN-C of 5 mm in z-direction was similar to the 4 mm observed in our study. Investigation of relative motion is of particular interest for the development of tumor tracking techniques for LA-NSCLC [
25,
26]. Tracking of LA-NSCLC is challenging as it requires simultaneous tracking of both PT and LN, with LN usually difficult to identify on planar x-ray or CBCT images. While LN motion was positively correlated with C motion indicating that C might be a good surrogate for LN respiratory motion, as C is readily visible on standard set up imaging, such as kV x-rays and CBCT, relative respiration-related displacements between LN and C were comparable to LN and PT. Therefore, PT and C appear to be comparable surrogates for respiratory motion of LN. Investigating the motion properties of PT, LN and surrogates relative to each other is of interest for the development of motion models in the complex geometries of lung cancer. In addition, information on relative respiration-related displacements is clinically relevant for the selection of appropriate breathing phases for gating techniques where typically end expiration phases with little motion are selected.
Temporal variations in motion range
High rates of ITV misses during radiotherapy have been described by Mohammed [
27] due to position, motion, shape and volume changes. The present study focuses on motion variations, showing that all but one patient had change in motion range >1 mm. Increased ranges of motion >3 mm relative to the planning scan, however, were rare despite the observed volume changes during treatment. The observed variations of relative displacements were also small. PTV margins of 5 mm covered 98 % of PT, 95 % of C and 100 % of LN and 95 % of PT-LN, 98 % of PT-C and 97 % of LN-C relative motion variations assuming use of an ITV to cover initial motion ranges and free breathing situations. In situations with larger motion range on the planning scan, margins might need to be adjusted to cover variations of respiratory motion during therapy. Variations of absolute and relative motion >5 mm were significantly more frequent in patients with larger initial motion range suggesting that patients with large initial motion range might benefit more from reevaluation. So far, few studies have performed repeated motion analysis of PTs over the treatment course. Britton et al. [
15] found increased PT mobility on weekly 4D CTs and suggested repeated 4D CT for reassessment of ITVs. Michalski et al. [
3] repeated 4D CTs after an average of 34 days and observed reproducible target motion in 87 %. Redmond et al. [
17] analyzed PT motion on 2 repeat 4D CTs at 30 and 50 Gy and found no significant motion variation. Clearly, differences in the time periods of reassessment, use or avoidance of biofeedback strategies and the overall observation period have resulted in these seemingly contradictory findings.
Time trends in lymph node motion have rarely been analyzed. While Thomas et al. [
13] evaluated whole LN regions, Bosmans et al. [
19] analyzed individual LN motion over the first 2 weeks of treatment and found only minor decrease in average motion from 5.6 mm to 5.3 mm. Using implanted markers and daily imaging, Schaake et al. [
11] also found minimal average motion changes of <1 mm which is in agreement with our findings. While population-based analyses might reveal small variations, for consideration of re-planning and adaptation, individual patient variations are important. As demonstrated in this study, patients with large initial motion have also larger variations during treatment and therefore might benefit from reassessment of their ITVs. Given the week-to-week variations in motion range, no optimum time point for reassessment can be defined. The need for reassessment is influenced by the scenario selected for differential (PT and LN or C) motion management in LA-NSCLC, which depends on the combination of image guidance strategy (e.g., repeated x-ray imaging, 4D CBCT), respiration management (e.g., tracking, breath hold, free breathing) and patient-specific factors (e.g., location and number of involved lymph nodes, availability of implanted markers in PT and/or LN).
Assuming a scenario of “real-time” PT tracking for LA-NSCLC during free breathing, day-to-day and intrafraction variations of the respiratory motion range would be accounted for during the tracking process. Only a small margin would be required to cover the time lag between the assessment of the target position and adjustments of the treatment field. To cover the LN motion in this scenario, an ITV based on appropriate volumetric scans, e.g., all breathing phases of a 4D CT planning scan, should be generated. Based on our findings, both increases in LN motion range and relative displacements of LN relative to PT or C > 3 mm are rare and are usually covered by a 5 mm PTV margin with image guidance of either PT or C. As an alternative to tracking the PT, tracking C or even LN (provided they are made visible by implanted markers) would be an option if large ITVs in the mediastinum due to large LN motion are prohibitive with regards to normal tissue toxicity. Ideally, all involved targets, PT and LNs, should be tracked independently for optimum target coverage and normal tissue sparing. In scenarios without tracking that use (4D) CBCTs for motion assessment, ITVs of PT and LNs on the initial scan cover both absolute and relative motion ranges. As shown in our study, absolute and relative motion increases >5 mm were observed for PT in one scan (2 % of all scans) and for PT-LN in 3 scans (5 % of all scans). As absolute motion and relative displacement are related, margins of 5 mm should be sufficient to cover both absolute and relative variations in respiratory motion. This is, however, an estimate which ignores other important sources of uncertainty such as delineation error and the quadratic nature of error summation for margin generation.
In the present study, motion range was measured on 4D CTs that cover only few respiratory cycles, potentially underestimating actual intrafraction motion variations. It has been shown, however, that motion ranges in general remain stable during one fraction [
14,
16]. Both 4D CT imaging artifacts and contouring variability might have influenced the present data. Several measures as described above were applied to improve image quality and contouring consistency. Most importantly, only one physician performed all contouring to avoid interobserver variation.