Comparison of IGRT techniques
Mean differences between TPUS and CBCT determined shifts of 2 to 3 mm for each dimension were observed. The resulting Euclidean distance was 6 mm on average. However, for individual setups differences up to 20 mm occurred. The difference between TPUS and CBCT was larger than 5 mm and 10 mm in 58 % and 11 % of all monitored treatment sessions.
Robinson et al. evaluated the differences between CBCT and transabdominal US imaging. They reported a mean discrepancy of 10 mm while our evaluation resulted in an average distance of 6 mm [
12]. Other studies compared the accuracy of setup verification for implanted fiducial markers and US in each cartesian direction [
11,
13]. The mean differences in setup errors were up to 2–3 mm which is comparable to our results with a median deviation of 2.5–2.6 mm deviation.
In our study, larger differences were observed in longitudinal direction (3.2 ± 2.7) mm while smaller deviations were seen in lateral direction (2.7 ± 2.3) mm. This is in agreement to the results published by other groups [
11,
13]. The Bland-Altmann analysis resulted in limits of agreement of 8 to 10 mm covering 95 % of the measurements. The largest limit of agreement was calculated in longitudinal direction (6.5 mm/−9.4 mm). This is in agreement to the results published by van der Meer et al. who obtained a limit of agreement of 10 mm in longitudinal direction [
13]. In contrast, Robinson et al. observed the greatest disparity between CT and transabdominal US in the vertical direction [
12].
Discrepancies between TPUS and CBCT of 5 mm and more occurred in 58 % of our monitored treatment sessions. Discrepancies of 10 mm were exceeded in 11 %. Again, Robinson et al. observed larger deviations: 5 mm were exceeded in 89 % and 10 mm were exceeded in 42 % of the measurements [
12]. Van der Meer et al. showed that in 56 % of all patients the difference in at least one direction was larger than 5 mm (>10 mm for 11 %) [
13]. In the present study, the distribution of setup errors showed a larger range for TPUS (0.8 - 16.9 mm) than for CBCT (0.6 - 13.3 mm). The mean and standard deviation were slightly smaller for CBCT imaging (TPUS: 6.0 ± 2.9 mm and CBCT: 5.4 ± 2.7 mm). The absolute difference in 3D between both imaging modalities was (6.0 ± 3.1) mm with a median deviation of 5.6 mm. Subsequently, setup differences between TPUS and CBCT are 6 mm on average. These preliminary results will be examined in a larger patient population in our further research. Some groups report that the accuracy of US is comparable to fiducial marker based localization [
11,
13] while another group concluded that the US fails to deliver an acceptable level of geometric accuracy with regard to prostate localization [
12].
In comparison to US, external localization systems like MRI-guided IGRT and electromagnetic tracking systems achieve a geometric accuracy of 1–2 mm [
9,
18]. Mayyas et al. compared the interfraction setup errors between different localization techniques (kV, CBCT and US imaging and electromagnet transponders) [
19]. They found that that the kV and CBCT shifts are comparable to that of US.
Image acquisition for CBCT is a user independent technique, which is widely used for setup verification during radiation therapy for prostate cancer [
20]. However, CBCT leads to additional patient dose. That is admittedly low in comparison to the treatment related dose, but is not fully negligible in the patient cohort of prostate EBRT with a relatively favorable prognosis. The CBCT is reasonably not used for setup verification in every treatment session. Whereas, a daily use of the US imaging is very well possible and can potentially detect a disadvantageous bladder and rectum filling earlier. This information can be incorporated in the treatment workflow by asking the patient to adapt bladder or rectum filling.
Besides setup verification, the transperineal setup of the US device additionally offers the possibility of intrafraction target monitoring and possible motion management in the future. A first proof of principle experiment of US based motion tracking was performed in a phantom study by Schwaab et al. with promising results but the need of further research [
21]. Colvill et al. studied the dosimetric impact of tracking and gating to account for intrafraction motion during prostate radiotherapy [
22]. O’Shea et al. reviewed the applicability of intrafraction monitoring with US for prostate and other treatment sites [
23]. Until further results on the practicability and accuracy of TPUS are available, certainly the combination of setup verification by CBCT and intrafraction target organ and OAR motion monitoring by TPUS imaging can use the benefits of both imaging modalities.
Limitations
One limitation is that no rotational errors were determined for prostate positioning. The US imaging and registration offers translational errors only. For setup verification of patients with prostate cancer, we correct for translational errors according to our protocol due to practicability and patient stability. Effects of rotations and translations of the prostate should be taken into account during target delineation and margin definition. The correction of rotational errors becomes more important for the hypofractionated radiotherapy treatment of prostate cancer and treatment of large, complex target volumes or simultaneous, single isocenter irradiation of multiple targets [
24]. The correction of rotational errors can induce substantial secondary patient displacements if the patient is not immobilized sufficiently [
25].
Another limitation of our study is the small number of patients. The results will be examined in a larger patient population in further research.
The applicability of US for prostate localization is discussed controversially in the literature due to its uncertainties like user variability (scanning variability and registration variability). Different probe pressures during image acquisition can lead to unreproducible target organ movements [
4,
11‐
13]. The influence of the probe pressure on the reproducibility is still an issue of concern [
4,
10,
13,
16]. The advantage of the Clarity system is the Autoscan probe decreasing the scanning variability. As we found out in a recently published study, the transperineal probe setup shifted the penile bulb in cranial direction and thereby increased the dose to the penile bulb [
16]. However, in that study, no relevant changes in the prostate or PTV volumes were observed and organ motion of the prostate was only seen to a minor extent mainly in the superior direction. [
26]. This seems to substantiate the assumption that probe pressure induced organ motion is reduced to a minimum by the transperineal scanning approach compared to transabdominal US. The variability of probe pressure is further reduced within the TPUS Clarity® system by probe fixation and a height scale as well as the display of the current and planned probe position in the software. Nonetheless, as low pressure of the TPUS probe as feasible to still achieve appropriate image quality is recommended for limitation of structure shifts and further reduction of placement variability.
User variability is influenced not only by scanning, but also by image registration and the manual placement of the Reference Positioning Volume. For the transabdominal US approach, van der Meer et al. reported a user variability of 2–3 mm – including scanning and matching variability [
13]. Van der Meer determined the intra- and interoperator variability for transabdominal US. In their study, the intraoperator match variability was in the range of 1 mm and the interoperator variability was in the range of 1.3 - 1.8 mm. Because the accuracy of US is highly influenced by the user more effort is required to standardize US imaging compared to CBCT or portal imaging, as van der Meer already stated [
13]. This certainly includes a comprehensive employee training but also improvements from the technical side are warranted like automated image registration and segmentation.