1

Investigation of overriding the HU of bowel gas in the PTV to create a more robust
treatment plan for pancreatic radiotherapy.
Authors: Kelli Braith R.T.(T)., Alexander Kehren R.T.(R)(T)., Ally Roberts R.T.(T)., Nishele
Lenards Ph.D., CMD, R.T.(R)(T), FAAMD, Ashley Hunzeker, M.S., CMD, Matt Tobler, CMD
Medical Dosimetry Program at the University of Wisconsin - La Crosse

Abstract:
Introduction:
Continuous advances to the treatment planning process in external beam radiation
therapy (EBRT) are necessary to ensure high quality and accurate treatments are delivered to
patients. Accurate dose calculation algorithms are pivotal in ensuring the efficacy and safety of
treatment plans in radiotherapy, and are evolving to become more accurate, especially in
inhomogeneous mediums. Principle-based algorithms such as the Monte-Carlo (MC) are
considered the most sophisticated calculations.1 This is because MC algorithms include all
possible beam particle processes and interactions as the radiation travels through various
mediums, creating an accurate but time -consuming calculation.1 Model-based algorithms such
as Collapsed Cone Convolution (CCC), and Analytical Anisotropic Algorithm (AAA) are the
more commonly used methods and are often preferred due to their high-level accuracy and fast
computation times which is imperative for busy and fast-paced clinics. These model-based
algorithms have been extensively used in clinics for dose calculations and produce acceptable
treatment plans; however, these models have some limited dose accuracy in areas with large
volumes of air.2 The high level of dose calculation accuracy is one of the reasons that MC
models are superior and why a new photon dose calculation algorithm has been developed to
mimic principle-based algorithms.

The emergence of Acuros XB Advanced Dose Calculation (AXB) has improved the
accuracy of plans, and highlights the differences in dosimetric reporting compared to methods
like AAA.3 The AAA model overestimates dose in the presence of air, while numerous studies
have shown that AXB more accurately demonstrates the air/tissue interface compared to AAA,
particularly when representing the lack of backscatter dose upon entering air at the interface. 4,5
During the planning of head and neck cancers where there are significant air cavities, Asher et al 6
 2

studied the effects of removing the air from the laryngeal PTV to overcome the buildup dose at
the air/tissue interface. The researchers created 2 PTVs, the first encompassed the air cavity in
the larynx. For the second PTV, the researchers contoured the volume around the larynx so that
the air cavity was excluded. Their results concluded that when the optimizer created a plan that
excluded air from PTV; the coverage of the PTV was not compromised, the maximum dose was
reduced, and a more homogenous plan was created.6 Transitioning from AAA to AXB presents
challenges in treatment planning, notably observed when the Treatment Planning System (TPS)
optimizer strives to distribute adequate dose to air structures within the Planning Target Volume
(PTV).7
The existence of air-filled regions in PTVs, whether due to cavities such as sinuses or gas
in the small bowel, creates a challenge for radiation dose distribution. When a beam passes
through a patient, the lower density air cavities can cause a reduction in the number of laterally
scattered and back scattered electrons.8 This disruption of electronic equilibrium results in an
underdose to the proximal region of the air/tissue interface and an overdose in the distal region of
the interface.9,10 The robustness of a treatment plan can be reduced by the presence of air and can
lead to an underdosing of the PTV, which has the potential to cause a recurrence of the disease,
and may create unnecessary hotspots with the PTV or organs at risk (OAR).
Treatment plan robustness is also diminished when there is a lack of reproducibility
during the treatment delivery process. The reproducibility of the treatment plan is a critical
aspect of EBRT and can be compromised when the size and location of the planning volumes
can vary daily. In clinical practice, issues persist with reproducing air in the bowel during
treatment planning. This has been demonstrated by the significant and undesirable dose
variations (up to 28%) in small bowel doses among rectal cancer patients during weekly CT
scans.11 Rarely is the small bowel's volume and form the same day-to- day, which leads to
variability with dose value calculations. The small bowel loops are always in motion due to
peristalsis, where they undergo both fluctuating displacements of the wall, and significant
variations in volume resulting from content changes.11 This issue of reproducibility becomes an
issue for pancreatic patients, where the location of bowel gas near the PTV can change daily,
thereby potentially changing the dose distribution in the region and causing excessive hotspots.
When creating a treatment plan using the AAA model algorithm, the optimizer
misrepresents the dose distribution at the air/tissue interface caused by air or gas within the PTV.
 3

Although there has been an increase in AXB usage, dosimetrists are still finding that at the
air/tissue interface, there is insufficient dosage in the proximal region and excessive dose in the
distal region. This can cause unnecessary hot spots, and because of the daily anatomic tissue
variations within the abdominal cavity, can also cause dose changes to the PTV and OAR. The
problem is that when creating an Acuros XB plan with air in the treatment volume, the optimizer
overcompensates fluence through the air cavity leading to target and OAR dose changes with
anatomic variations. The purpose of this study was to determine if adjusting the Hounsfield Unit
(HU) values for bowel gas in the PTV will decrease hotspots and improve the desired dose
distribution for pancreatic radiotherapy. The researchers hypothesized that by adjusting the HU
values for bowel gas in the PTV, the maximum dose to the small bowel, large bowel, stomach,
and duodenum (H1A) and PTV (H2A) would all decrease. The researchers hypothesized that by
making this change to the HU value of the bowel gas, the volume of PTV receiving 105% of the
dose (V105% ) (H3A) will also decrease.
Methods and Materials
Patient Selection
Fifteen patients that had previously received pancreatic radiation therapy treatments from
a single clinic were chosen for this retrospective study. The study included patients who were
treated using both a breath-hold method and a free-breathing method. All patients met the
criterion of having at least 0.1 cc of air within the PTV during their simulation scan. Patients
were only excluded from the study if there was less than 0.1 cc of air in their pancreatic PTV.
The 15 patients were treated on a Varian Ethos machine and received a daily CT imaging scan
prior to each treatment.
Contouring
The original treatment plan, created by a certified medical dosimetrist (CMD) and
approved by a radiation oncologist, was copied, and transferred to a new research course in the
TPS Eclipse. Every structure was duplicated, and the contours remained unchanged from the
original plan. The OAR contours included the small bowel, large bowel, stomach, liver, and
duodenum. The researchers expanded the target volume, previously drawn by the radiation
oncologis, by 2 mm. This structure became the PTV+2mm. The purpose of this step was to
encompass any air that was immediately outside of the PTV. Without the .2mm margin, any air
near the PTV would contribute to creating a dose build-up region.
 4

Treatment Planning
To remove as many variables as possible, one researcher was responsible for creating two
new plans: one uncorrected for air in the PTV and one corrected for air in the PTV. Also, only
one researcher was responsible for creating an HU_Air structure within the plan to maintain
consistency with contouring throughout the study. The researcher utilized the image thresholding
function of the TPS contouring workspace to identify HU values ranging from -1000 to -800.
These identified values were then designated as the HU_Air contours. After the HU_Air
structure was created, it was cropped from extending outside of the PTV+2mm structure. This
process was repeated for each of the 15 pancreatic cases.
The first step in creating the new plans was to identify the priority 1 dose parameters
from the original plan, as indicated by the radiation oncologist. The priority 1 dose parameters
are listed in Table 1. Once the dose parameters had been identified and established as planning
guidelines, the researchers employed the AXB algorithm and initiated the optimizer from the
initial stage of the four-step optimization procedure. The aim was to meet all priority one dose
parameters, and the optimization process was reiterated until those objectives were fulfilled.
Some plans were unable to achieve the priority one metrics. These cases were optimized to attain
similar objectives to the CMD's original plan. The process was repeated for all 15 patients and
was designated as Plan 1.
After completing Plan 1, all the structures, including the PTV+2mm and HU_Air, were
copied and transferred to a new research course in Eclipse. The OAR contours and targets were
also unchanged from the original plan. These were designated as Plan 2. The HU_Air structures
were then overridden and given the HU value of zero. The researchers chose the HU value of
zero to represent the average HU of soft tissue in the treatment area. The dose parameters from
the original plan were utilized as a reference, and once again, the optimization procedure was
repeated until all the priority one dose parameters were fulfilled. Several plans failed to meet the
priority one metrics; thus, they were optimized until comparable objectives were reached.
Verification Process
To track changes to the OAR and PTV dose distributions between plans 1 and 2, the
researchers chose to use one daily CT scan per every 5 treatments for the verification process.
The verifications came from the daily cone-beam computed tomography (CBCT) scans on the
Varian Ethos machine. When registering the verification scan to both plans, the researchers
 5

aligned the treatment area including the stent, soft tissue, and spine. The images were only
aligned in the axial, sagittal, and coronal planes since the Ethos machine does not have 6 degrees
of freedom (6DOF) capabilities. After the images were aligned, the target and OAR volumes
were rigidly transferred onto the verification CT scan. The OAR in the vicinity of the HU_Air
structure was adjusted to accurately represent the anatomical features observed on the day the
pre-treatment CBCT was acquired. To perform the verifications, the researchers used the exact
monitor units (MU) from the original plan to recalculate the dose on plans 1 and 2. This was
done to maintain control variables from the original plan, and the revised dose distributions were
assessed to evaluate the significance of the HU override.
Plan Comparison
The observed metrics were the maximum doses (Dmax) of four OAR (small bowel, large
bowel, stomach, and duodenum), the Dmax dose within the PTV, and the V105% of the PTV. Four
verification scans were performed on the 15 patients and each metric was averaged to give one
observational value. Those values were compared between the plan with no air override (Plan 1)
and the plan with the air override (Plan 2). If a patient’s OAR was not encompassed by the
HU_Air structure and/or a 1.0 cm margin around the structure, the Dmax was not included, and a
corrected data set was used in the analysis process. The researchers determined that Dmax doses
for OAR that were > 1.0 cm from the HU_Air structure were not relevant to the study. As noted
in the treatment planning process, there were 3 patients whose clinical target volume (CTV) and
PTV priority 1 goals were scarified to meet the OAR priority 1 goals. The researchers were able
to achieve a PTV of D95 > 98% for 2 of the 3 patients. The OAR priority 1 metrics were
achieved for the 3rd patient, however, the D95 > 98% was not met.

Statistical Analysis
After all the data from the 15 patients was collected, the Shapiro-Wilk test of normality
was conducted to assess whether the observed values were normally distributed and to validate
the use of the t-test. P-values greater than .05 suggested that the data followed a normal
distribution pattern. Three p-values below .05 were observed, indicating non-normality. A
Wilcoxon Signed-Rank (WSR) Test was conducted for non-normal values. Table 2 gives the
mean and standard deviation of the difference (SD) data for each metric, along with the t-test
statistics and p-values. A significance criterion of P < 0.05 was utilized to evaluate if the 3 null
hypotheses could be rejected.
 6

For small sample sizes (n < 50), the Shapiro-Wilk test is recommended for normality
testing due to its higher sensitivity in detecting nonnormality. This test is widely recognized and
commonly employed.12 A t-test is a statistical tool that measures the significance of the
difference between the means of two groups, while considering their variation or distribution. 13
The WSR test is the nonparametric counterpart of the paired t-test. The method does not make
any assumptions about the distributions of the original population, but it does assume that the
distributions of the differences are symmetrical.14
The Dmax doses of the small bowel were compared for each plan, however, only 6 of the
15 patients were included in the corrected data set (Figure 1). For the other 9 patients, the small
bowel volume was greater than 1.0 cm from the HU_Air structure and deemed irrelevant for the
purpose of this study. The Shapiro-Wilk p-value was .3463, because this value is normal, the
additional t-test and WSR test was not performed. The t test p-value, measuring the significance
of the difference between the means of both plans, was .2814. There was very little difference in
average values between the plans for the small bowel Dmax metric, and with P-values > 0.05, the
null hypothesis (H10) failed to be rejected (Figure 1).
Only 6 of the 15 patients were included in the corrected data set for the large bowel D max
dose metric. The large bowel volume was > 1.0 cm from the HU_Air structure for the other 9
patients in the study. The Shapiro-Wilk p-value was .1743 (normal), therefore an additional t-
test and the WSR test was not necessary for this metric. The p-value for the t-test was .4173.
With P-values > 0.05 there is no significant difference between the plans for the large bowel
metric; the null hypothesis (H10) failed to be rejected (Figure 2).
The corrected data set for the Dmax dose of the stomach included 9 of the 15 patients. For
the other 6 patients in the study, the stomach volume was not within 1.0 cm of the HU_Air
override structure. The Dmax dose of the stomach was 1 of 3 metrics that showed non-normality,
due to a Shapiro-Wilk p-value of .0000. Because of the non-normality of the p-value, an
additional t-test and a non-parametric WSR test were performed. The t-test p-value was .2409
and the WSR p-value was .5937. Although the Shapiro-Wilk test showed non-normality for the
Dmax dose of the stomach, the 2 other tests resulted in P-values > 0.05, which indicates that the
null hypothesis (H10) failed to be rejected (Figure 3).
The corrected data set for the Dmax dose of the duodenum included the greatest number of
patients when compared to the other OAR metrics. Fourteen of the 15 patients were included in
 7

final calculations, with only 1 patient having the duodenum volume > 1.0 cm from the HU_Air
structure. The p-value for the Shapiro-Wilk test was .6302 (normal), and therefore did not
require further testing. The t-test p-value was .4260. With P-values > 0.05 for the duodenum
metric, the null hypothesis (H10) failed to be rejected (Figure 4).
The data sets for the 2 PTV metrics were not corrected and included all 15 patients since
those values were not dependent on the PTV’s proximity to the HU_Air structure. For the Dmax
dose within the PTV, the Shapiro-Wilk test revealed non-normality because of a p-value
of .0000. An additional t-test and the WSR tests were performed, and the ensuing values were a
t-test p-value of .1176 and a WSR p-value of .3666. The P values of the Dmax dose within the
PTV were > 0.05; therefore, the null hypothesis (H20) failed to be rejected (Figure 5).
For the final metric, the V105% of the PTV (measured in cc’s) was compared between
Plans 1 and 2. This was another metric that was determined to be non-normal because the
Shapiro-Wilk p-value was .0010. Additional testing was performed; t-test and WSR, and the
resulting p-values were .5023 and .7729, respectively. Though the Shapiro-Wilks test showed
non-normality with this metric, the p-values of the additional tests were > 0.05; therefore, the
final null hypothesis (H30) failed to be rejected (Figure 6).
 8

References
1. Kim DW, Park K, Kim H, Kim J. History of the photon beam dose calculation algorithm in
radiation treatment planning system. Prog Med Phys. 2020; 31:54-62.
https://doi.org/10.14316/pmp.2020.31.3.54
2. Fogliata A, Nicolini G, Clivio A, Vanetti E, Cozzi L. Dosimetric evaluation of acuros xb
advanced dose calculation algorithm in heterogeneous media. Radiat Oncol. 2011;6:82.
https://doi.org/10.1186/1748-717X-6-82
3. Cakir A, Akgun Z. Dosimetric Comparison of Integral Radiation Dose: Anisotropic Analytical
Algorithm and Acuros XB in Breast Radiotherapy. Int J Med Phys Clin Eng Radiat Oncol.
2019;08(02):57-67. https://doi.org/10.4236/ijmpcero.2019.82006
4. Bassi S, Tyner E. 6x Acuros algorithm validation in the presence of inhomogeneities for VMAT
treatment planning. Rep Pract Oncol Radiother. 2020; 25(4):539-547.
https://doi.org/10.1016/j.rpor.2020.03.018
5. Yan C, Combine AG, Bednarz G, et al. Clinical implementation and evaluation of the acuros dose
calculation algorithm. J Appl Clin Med Phys. 2017;18(5):195-209.
https://doi.org/10.1002/acm2.12149
6. Asher D, Amestoy W, Studenski MT, et al. Dosimetric comparison of intensity-modulated
radiation therapy for early-stage glottic cancers with and without the air cavity in the planning
target volume. Med Dosim. 2019;44(4):405-408. https://doi.org/10.1016/j.meddos.2019.02.007
7. Ito T, Monzen H, Kubo K, et al. Dose difference between anisotropic analytical algorithm (AAA)
and acuros xb(AXB) caused by target’s air content for volumetric modulated arc therapy of head
and neck cancer. Rep Pract Oncol Radiother. 2023; 28(3):399-406.
https://doi.org/10.5603/RPOR.a2023.0032
8. Gibbons JP, Khan FM. Khan’s the Physics of Radiation Therapy. Wolters Kluwer; 2014.
9. Joshi CP, Darko J, Vidyasagar PB, Schreiner LJ. Dosimetry of interface region near closed air
cavities for Co-60, 6 MV and 15 MV photon beams using Monte Carlo simulations. J Med Phys.
2010;35(2):73-80. http://dx.doi.org/10.4103/0971-6203.62197
10. Seif F, Bayatiani MR, Hamidi S, Kargaran M. Investigating the effect of air cavities of sinuses on
the radiotherapy dose distribution using monte carlo method. J Biomed Phys Eng. 2019;9(1):121-
126. http://dx.doi.org/10.31661/jbpe.v9i1Feb.1046
 9

11. Li S, Gong Y, Yang Y, Guo Q, Jianjun Q, Tian Y. Evaluation of small bowel motion and
feasibility of using the peritoneal space to replace bowel loops for dose constraints during
intensity-modulated radiotherapy for rectal cancer. Radiat Oncol. 2020; 15(211).
https://doi.org/10.1186/s13014-020-01650-z
12. Mishra P, Pandey CM, Singh U, Gupta A, Sahu C, Keshri A. Descriptive statistics and normality
tests for statistical data. Ann Card Anaesth. 2019;22(1):67-72.
https://doi.org/10.4103/aca.ACA_157_18
13. Wainer H, Braun HI, Educational Testing Service. Test Validity. L. Erlbaum Associates; 1988.
14. Holmes D. Statistical methods in laboratory medicine. In: Clarke W, Marzinke M. Contemporary
practice in clinical chemistry. 4th ed. Cambridge, MA: Academic Press; 2020:15-35
15. De Martino F, Clemente S, Graeff C, Palma G, Cella L. Dose calculation algorithms for external
radiation therapy: an overview for practitioners. Appl Sci. 2021; 11(15):6806.
https://doi.org/10.3390/app11156806
16. Tajaldeen A, Ramachandran P, Alghamdi S, Geso M. On the use of AAA and acuros xb
algorithms for three different stereotactic ablative body radiotherapy (SABR) techniques:
volumetric modulated arc therapy (VMAT), intensity modulated radiation therapy (IMRT) and
3D conformal radiotherapy (3D-CRT). Rep Pract Oncol Radiother. 2019;24(4):399-408.
https://doi.org/10.1016/j.rpor.2019.02.008
17. Muñoz-Montplet C, Fuentes-Raspall R, Jurado-Bruggeman D, Agramunt-Chaler S, Onsès-
Segarra A, Buxó M. Dosimetric impact of acuros xb dose-to-water and dose-to-medium reporting
modes on lung stereotactic body radiation therapy and its dependency on structure composition.
Adv Radiat Oncol. 2021;6(4). https://doi.org/10.1016/j.adro.2021.100722
18. Kumar L, Bhushan M, Kishore V, Yadav G, Gurjar OP. Dosimetric validation of Acuros® XB
algorithm for RapidArc™ treatment technique: a post software upgrade analysis. J Cancer Res
Ther. 2021;17(6):1491-1498. https://doi.org/10.1016/j.adro.2021.100722
19. Mishra A, Pathak R, Raj Verma T, et al. Evaluation of radiation treatment planning algorithms in
IMRT and VMAT: a comparative dosimetric study in lung equivalent heterogeneous medium. J
Biomed Phys Eng. 2023;13(6):503-514. https://doi.org/10.31661/jbpe.v0i0.2206-1508
20. Chang X, Yuan Y, Wang G, et al. The effects of intestinal air cavity on dose distribution of
volume modulated arc therapy for cervical cancer. Ann Palliat Med. 2022;11(4):1326-1335.
http://dx.doi.org/10.21037/apm-22-66
 1
0

21. Bush K, Gagne IM, Zavgorodni S, Ansbacher W, Beckham W. Dosimetric validation of acuros xb
with monte carlo methods for photon dose calculations. Med Phys. 2011; 38(4):2208-2221.
https://doi.org/10.1118/1.3567146
 1
1

Tables

Table 1. Priority 1 dose parameters goals for PTV and OAR that researchers attempted to
achieve when creating Plans 1 and 2.
Structures Priority 1 Goals
PTV D95% ≥ 100%,
PTV V95% ≥ 99%
CTV V100% > 95%
Stomach V45 ≥ 0.03cc
Duodenum V48 ≥ 0.03cc
Duodenum V45 ≥ 5cc
Small Bowel V45 ≥ 0.03cc
Large Bowel V48 ≥ 0.03cc

Table 2. The mean of differences, SD, and p-values for: Shapiro Wilk, t-test, and WSR tests, for
all metrics.
Standard Wilcoxon
Mean of Shapiro-Wilk t-test
Metric Deviation of Signed-Rank
Differences test p-value p-value
Differences p-value
Duodenum Dmax dose 0.0407 0.800 .6302 0.4260
Stomach Dmax dose 0.6711 2.7295 .0000 0.2409 0.5937
Small Bowel Dmax dose 0.1267 0.5010 .3463 0.2814
Large Bowel Dmax dose 0.5333 5.9360 .1743 0.4173
Dmax Dose in PTV 1.0853 3.3885 .0000 0.1176 0.3666
V105% in PTV -.005 3.3160 .0010 0.5023 0.7729
 1
2

Figures

Figure 1: Comparison between the average small bowel maximum dose (%) by patient. Plan 1
(no air override) in green, and Plan 2 (air override) in purple.
 1
3

Figure 2: Comparison between the average large bowel maximum dose (%) by patient. Plan 1
(no air override) in green, and Plan 2 (air override) in purple.

Figure 3: Comparison between the average stomach maximum dose (%) by patient. Plan 1 (no
air override) in green, and Plan 2 (air override) in purple.
 1
4

Figure 4: Comparison between the average duodenum maximum dose (%) by patient. Plan 1 (no
air override) in green, and Plan 2 (air override) in purple.

Figure 5: Comparison between the average maximum dose (%) within the PTV by patient. Plan
1 (no air override) in green, and Plan 2 (air override) in purple.
 1
5

Figure 6: Comparison between the V105% of the PTV (cc’s) by patient. Plan 1 (no air override) in
green, and Plan 2 (air override) in purple.