1 Background
Single-fraction
192Ir high-dose-rate brachytherapy (HDR-BT) of the liver is an ablation technique which has shown promising results with respect to safety and efficacy in the treatment of nonresectable primary and secondary liver malignancies [
1‐
3]. HDR-BT provides steep dose gradients at the surface of the target volume due to the low
γ-ray energy of
192Ir and use of a point source, and thus can be used to treat several malignancies in one session or recurrent malignancies sequentially without seriously impairing the functional hepatic reserve [
4]. To prevent recurrence at the tumor margins, catheter placement and dwell positions of the
192 Ir point source have to be carefully planned [
5]. The accuracy of dose application is predominantly dependent on catheter positioning. Computed tomography (CT) was used to monitor catheter implantation, and 3D CT datasets acquired in breath-hold were used for treatment planning. For irradiation patients were transferred from the CT unit to the brachytherapy unit. Dislocation of catheters during patient transfer might be a potential source of error with respect to correct dose application at the target site. Additionally, the liver is an elastic organ and could be deformed between catheter implantation and irradiation.
The treatment of larger tumors with an
192Ir point source requires the implantation of approximately 1 catheter for each 1 - 2 cm of tumor diameter. The contributions of several catheters with numerous dwell positions to the planned dose in a large part of the target volume lead to regional prolongation of irradiation. Several authors describe an increased normal tissue dose tolerance for prolonged radiation therapy or pulsed dose rate (PDR) radiation therapy [
6,
7] even if the total irradiation time is less than one hour [
8].
The present study aims at addressing two methodical aspects of HDR-BT: First, to investigate the limits of catheter positioning accuracy and its clinical importance. Second, to investigate if effects of prolonged irradiation times on the tolerance dose of normal liver parenchyma are important for clinical practice and may have to be taken into account in treatment planning.
3 Results
The validation of image registration accuracy using landmarks yielded a mean deviation of 2.64 mm (25% quartile width (
Q25 ): 0.28 mm, 75% quartile width (
Q75): 4.51 mm). Thus registration accuracy proved to be sufficient for evaluating catheter positioning accuracy. A total of 161 MRI examinations of 62 irradiation effects were performed 6 and 12 weeks after HDR-BT. Table
1 shows the mean volume and threshold dose of hepatocyte dysfunction (T1-w images) and interstitial edema (T2-w images) and corresponding liver tolerance doses as well as the standard deviation between the examinations at 6 and 12 weeks (6W and 12W).
Table 1
Normal liver tissue tolerance dose and volume of irradiation effect
Dose/Gy | 13.7 ± 4.8 | 16.7 ± 5.0 | 14.3 ± 6.2 | 16.6 ± 6.4 |
Volume/cm3 | 190.3 ± 158.6 | 127.2 ± 118.8 | 190.0 ± 166.4 | 157.0 ± 143.5 |
A total of 96 follow-up MRI examinations of 30 patients with 38 irradiation effects were assessed to analyze methodical limitations of catheter positioning accuracy. Only patients with unidirectionally implanted, i.e., nearly parallel, catheters were included in the evaluation. The median number of catheters inserted was 2 (Q25:1, Q75: 3 catheters; range: 1-8 catheters).
Table
2 presents the axial, orthogonal, and total shifts (in mm) between the center coordinates of the irradiation effects and tolerance dose volumes in relation to the direction vectors of catheter implantation. The mean axial shift of hepatocyte dysfunction (T1-w images) was -5. 3 ± 5.4 mm and of interstitial edema (T2-w images) -5. 6 ± 6.0 mm in plane, indicating a shift of the irradiation effect volume against the corresponding tolerance dose volume in the direction of the catheter entry sites. The orthogonal shift as a surrogate for registration inaccuracy due to liver deformation was 4.0 ± 2.5 mm on T1-w images and 4.6 ± 2.6 mm on T2-w images.
Table 2
Shift between irradiation effect and planned dose distribution
Axial shift/mm | -5.3 ± 5.4 | -5.6 ± 6.0 |
Orthogonal shift/mm | 4.0 ± 2.5 | 4.6 ± 2.6 |
Total shift/mm | 7.7 ± 4.4 | 8.4 ± 4.4 |
AC
S
| 1.14 ± 0.43 | 1.04 ± 0.49 |
The orthogonal and axial shifts between the center coordinates of the irradiation effect volume and the tolerance dose volume in relation to the direction vector of catheter implantation were highly correlated in the T1-w and T2-w MRI sequences (p = 0.003 and p < 0.001, respectively), as were the shifts between 6 and 12 weeks examinations (p = 0.001 and p = 0. 004, respectively).
The asymmetry coefficient of the orthogonal and axial shifts of the center coordinates of the irradiation effect and corresponding tolerance dose volume in relation to the direction vector of catheter implantation, AC
S
, was 1.14 ± 0.43 for hepatocyte dysfunction and 1.04 ± 0.49 for interstitial edema, indicating that the axial shift as a surrogate for catheter dislocation within the catheter track was predominant (p < 0.005). The asymmetry coefficient was significantly affected by the MRI sequence used (p = 0.014) but not by the change in the irradiation effect volume between the 6-week and 12-week examinations (p = 0.48).
A total of 129 follow-up MRI examinations of 44 patients with 48 irradiation effects were assessed to analyze the effect of prolonged irradiation time on the tolerance dose of normal liver parenchyma. All irradiation effects were induced by at least 2 brachytherapy catheters. The median number of catheters per irradiation effect was 4 (Q25: 3; Q75: 6 catheters; range: 2-11 catheters). The average time for complete application of the radiation dose was 1865 ± 758 seconds (range: 844 - 4432 seconds).
The volumes of the mismatch areas, „MA+" and „MA-", averaged over the 6-week and 12-week follow-up MRI examinations and T1-w and T2-w acquisitions, was 40.6 ± 28.9 cm
3 (23.5 ± 10.1%). The differences between the mismatch area volumes with regard to 6-week and 12 week follow-up examinations and T1-w and T2-w MRI are small, see Table
3. The average dose in „MA+" is approximately 12Gy 6 weeks and 14Gy 12 weeks after the intervention. The average dose in „MA-", is approximately 22-23Gy 6 weeks and 28Gy 12 weeks post intervention, see Table
3. The difference between the average doses in the mismatch areas is significant (
p < 0.0001). The values for the catheter contribution indices in the mismatch areas,
I
P
(
MA+) and
I
P
(
MA -), as well as the asymmetry coefficients of the catheter contribution indices in the mismatch areas,
AC
I
, with respect to hepatocyte dysfunction and interstitial edema and the corresponding follow-up dates are displayed in Table
3. The mean of
AC
I
is > 0 in each subgroup, indicating that the catheter contribution index in „MA+" is slightly higher than in „MA-".
I
P
(
MA+) and
I
P
(
MA-) are significantly affected by the volume loss of the irradiation effect between the 6-week and 12-week follow-up examinations and consecutive shifts of the mismatch areas towards the high dose regions of the dose plan (
p = 0.0014). There is no significant difference between
I
P
(
MA+) and
I
P
(
MA-) with respect to hepatocyte dysfunction and interstitial edema (
p = 0.9).
Table 3
Mean dose, deviation of mean dose from normal liver tissue tolerance dose, and dose protraction in mismatch areas
D(MA+)/Gy | 12.0 ± 4.3 | 14.1 ± 4.4 | 11.8 ± 5.4 | 14.0 ± 6.3 |
D(MA-)/Gy | 23.2 ± 11.9 | 28.5 ± 11.0 | 22.2 ± 11.6 | 27.7 ± 15.1 |
ΔD(MA+)/Gy | -2.1 ± 2.8 | -3.2 ± 1.9 | -2.1 ± 4.3 | -3.0 ± 3.1 |
ΔD(MA-)/Gy | 9.1 ± 7.5 | 11.2 ± 6.8 | 8.3 ± 6.6 | 10.7 ± 8.8 |
I
P
(MA+) | 1.67 ± 0.33 | 1.69 ± 0.26 | 1.67 ± 0.31 | 1.70 ± 0.27 |
I
P
(MA-) | 1.45 ± 0.39 | 1.35 ± 0.37 | 1.45 ± 0.37 | 1.39 ± 0.36 |
AC
I
| 0.17 ± 0.28 | 0.25 ± 0.27 | 0.16 ± 0.26 | 0.23 ± 0.22 |
V (MA +/MA-)/cm3 | 42.0 ± 26.7 | 38.2 ± 31.2 | 40.8 ± 29.2 | 43.0 ± 33.1 |
V (MA +/MA-)/% | 21.8 ± 11.1 | 23.9 ± 7.8 | 23.1 ± 0.8 | 27.0 ± 9.0 |
4 Discussion
In this study, we sought to assess two methodical aspects of HDR-BT: first, limits of catheter positioning accuracy and, second, effects of prolonged irradiation on the tolerance dose of normal liver parenchyma. The mean shift between the center coordinates of the irradiation effect volume and corresponding tolerance dose volume in relation to the direction vector of catheter implantation is ≈ - 5 mm in plane, indicating a shift of the irradiation effect in the direction of the catheter entry site. The shift is within the slice thickness of 5 mm of the treatment planning CT but larger than could be explained by registration inaccuracy, which is ≈ 3 mm, and inaccuracy due to local liver deformation in the follow-up images, resulting in an overall registration inaccuracy of ≈ 4-5 mm.
Determination of catheter positioning accuracy might be limited by the delineation of the brachytherapy catheters in the treatment planning CT since applicator geometry is entered manually. Partial volume effects in the treatment planning datasets could be a potential source of error in the treatment planning procedure, especially for catheters in oblique direction, since correct placement of the starting point of the catheter is dependent on conspicuity of the catheter tip.
Another limitation is the dislocation of catheters between acquisition of the planning CT and irradiation. Although the angiographic sheaths containing the catheters were secured to the skin by suture, retraction of the brachytherapy catheters within the catheter tracks might potentially occur due to patient movement, e.g., when the patient is transferred from the CT unit to the brachytherapy unit, and liver movement during respiration. However, the extent of the shift between an irradiation effect and the center of the planned dose distribution does not suggest a significant dislocation of the brachytherapy catheters within the catheter tracks.
The systematic shift between the irradiation effect volume and planned dose distribution has to be considered in treatment planning when defining the CTV to avoid underdosage of the tumor periphery. In our institution, the CTV comprises the tumor volume visible on contrast-enhanced CT scans plus a 5-mm safety margin. With regard to treatment planning, we conclude that a slice thickness exceeding 3 mm potentially impairs catheter positioning accuracy. We furthermore propose that it would be beneficial to increase the safety margin of the CTV in the direction of the catheter tips from 5 to 10 mm to avoid underdosage and consecutive recurrence at the tumor margin. The amount of mismatch (Table
3) between planned dose distribution and irradiation effect volume is determined by the registration accuracy or possibly by biological effects but does not allow to assess the reproducibility of the CTV. Two studies evaluated the accuracy of target positioning in extracranial stereotactic radiotherapy (ESRT) using special patient fixation. For mobile soft tissue targets, such as liver metastasis, Wulf et al. [
17] reported mean target deviations of 0.9 ± 4.5 mm, 0. 9 ± 3.0 mm, and 3.4 ± 3.2 mm in the craniocaudal, anteroposterior, and lateral directions, respectively, when breathing control was applied. The mean 3D deviation of the targets was 6.1 ± 4.6 mm.
For single-fraction therapy, Herfarth et al. [
18] reported mean target set-up deviations between treatment planning and treatment of 4. 0 ± 2.5 mm, 2.2 ± 1. 8 mm, and 2.2 ± 1.7 mm in the craniocaudal, anteroposterior, and lateral directions, respectively. The mean 3D deviation of the targets was 5.7 ± 2.5 mm.
The total in-plane deviation of the target location in our study was slightly higher, 4-6 ± 2-6 mm. However, we determined the effective positioning accuracy by comparing the shift between the irradiation effect in follow-up MRI and planned dose distribution. The authors quoted above compared treatment planning images with control CT datasets acquired before treatment [
17,
18] and did not evaluate the treatment effect.
Based on metric analysis of target mobility and set-up inaccuracy in the CT simulation prior to or during treatment, safety margins for defining the planning target volume (PTV) of about 5 mm in axial and 5 - 10 mm in craniocaudal direction are commonly added to the CTV in ESRT of lung and liver tumors [
19]. In contrast to the present study, Wulf et al. evaluated the reproducibility of the CTV of lung and liver tumors within the planning target volume (PTV) over the entire course of hypofractionated treatment in CT simulation prior to application of each fraction [
19]. The mean volume ratio of the PTV to the CTV was 2.2 ± 0.6 in liver targets. The authors showed that especially liver tumors with a CTV exceeding 100 cm
3 were susceptible to target deviation exceeding the standard safety margins for PTV definition. They suggested to increase the PTV by adding a larger safety margin to ensure adequate target dose deposition in these CTVs. In brachytherapy, the applicator moves to a certain extent together with the target and there is no need to increase the safety margin for larger tumors.
Catheter dislocation in brachytherapy was mainly investigated in fractionated HDR brachytherapy of the prostate, which differs from the technique used here in that a much larger number of catheters are implanted for more than one day. Imaging techniques (cone beam CT and CT) were used to assess catheter dislocation between the first and second fraction, i.e., over 24 hours. Foster et al. found a mean catheter displacement of 5. 1 mm, resulting in a significantly (
p < 0.01) decreased mean prostate
V100 (volume receiving 100Gy or more) from 93.8% to 76.2% [
20]. Five patients had maximum catheter displacement exceeding 10 mm. Simnor et al. found a mean movement in caudal direction relative to the prostate base between the first and second fraction of 7. 9 mm (range 0-21 mm). Planning target volume dose
D90% was reduced without movement correction by a mean of 27.8% [
21]. Kim et al. found an average (range) magnitude of craniocaudal catheter displacement of 2.7 mm (- 6.0 to 13.5 mm) using bone markers and 5.4 mm (-3.75 to 18.0 mm) using the center of two gold markers [
22]. Catheter dislocation in fractionated HDR brachytherapy of the prostate is in the same range as in the present study but, because of the much more complex irradiation geometry, the impact on dose coverage is much larger.
We assessed the effect of prolonged irradiation times on the tolerance dose of normal liver tissue to determine its relevance for treatment planning. A catheter contribution index served as a surrogate for prolonged pulsed dose administration in nonoverlapping areas of the irradiation effect volume and the corresponding tolerance dose volume. The catheter contribution index was slightly but significantly higher in „MA+" than in „MA-", indicating a prolongation of dose application in „MA+" compared to „MA-". Based on published data, we would have expected to find an increased tolerance dose of the liver parenchyma in areas irradiated for a longer time, i.e., by several catheters [
6,
7], even if the overall irradiation time is less than one hour [
8]. However, we found a decreased tolerance dose of the liver parenchyma in areas where the radiation dose was applied by several catheters for a prolonged period of time.
We hypothesize that the effects of prolonged irradiation on the tolerance dose of normal liver tissue might have been obscured by other factors. For instance, biological effects such as reactive inflammatory changes may mimic irradiation effects, or scarring of the liver tissue induced by catheter insertion may cause retraction of the irradiation effect towards the catheter entry site. Furthermore, we propose that inaccuracies in the positioning of the brachytherapy catheters are more pronounced in areas where several catheters contribute to the total irradiation dose and that the total applied effective dose in „MA+" was higher than would have been expected from the treatment plan. Since steep dose gradients are an inherent quality of interstitial HDR-BT, the shift of active dwell positions of one or several catheters towards the tumor periphery would be sufficient to significantly increase the applied dose outside the CTV. As the number of catheters increases, the probability of a dose shift due to slight inaccuracy in catheter positioning likely increases as well.
We conclude that the effects of prolonged irradiation time are of minor importance for interstitial HDR-BT compared to other factors such as positioning accuracy of brachytherapy catheters and do not have to be taken into account in treatment planning in HDR-BT if the total irradiation time does not significantly exceed one hour.
The study has several limitations. Obviously one key issue of the study is the registration accuracy. The validation of registration accuracy was based on corresponding vessel bifurcations identified in the planning CT and follow-up MR images by an experienced radiologist [
23,
24]. We applied affine registration, allowing 12 degrees of freedom, which compensates for whole organ deformation and yielded an accuracy of ≈ 3 mm with respect to vessel bifurcations within the central parts of the liver, comparable to other studies [
25,
26]. Affine registration has been proven to be precise and robust for liver registration [
25‐
27]. However, local liver deformation resulting from compression by adjacent organs (such as the stomach), different respiration levels, or the implanted catheters in the treatment planning CT data might not be sufficiently compensated for. To adequately compensate for these effects a finite element model-based deformable image registration would have been superior [
23,
24]. We tried to compensate for the limitations of affine registration by restricting the registration to the liver [
25]. Using this procedure, we achieved a registration accuracy with a mean deviation of 2.64 mm, which was smaller than that of the nonrigid registration used by Elhawary et al. [
28], for which the authors reported a mean target registration error of. 4.1 mm and a mean 95
th
-percentile Hausdorff distance of 3. 3 mm.
Second, the catheter contribution index has to be considered a rough simplification, merely providing a first estimate of the effect of prolonged dose administration. Dose administration was considered highly prolonged if the index was 2 (meaning that each catheter of the brachytherapy implant contributed < 50% of the irradiation dose in the mismatch area). It was considered fairly prolonged if the value was between 1 and 2 (indicating that more than 25% of the total irradiation dose in the mismatch area was applied by more than 1 catheter), and nonprolonged if the value was ≤ 1 (meaning that 75% or more of the total irradiation dose in the mismatch area was applied by 1 catheter only). Nevertheless, the tool is sufficient to rule out practically relevant effects of prolonged dose administration in HDR-BT in vivo.
6 Competing interests
The authors declare that they have no competing interests.
7 Authors' contributions
LL, CW: data analysis, manuscript preparation.
PW, JR: study coordination, study design.
MS, KM: data acquisition.
SK: data analysis
All authors read and approved the final manuscript.