Omar Niazi

Mrs.Sasser

Intern Mentor P-6

8 February 2019

Proton Therapy Uncertainty and Various Applicable Methods to

Improve Treatment Effectiveness

Introduction

The overall topic of this paper, cancer, is the second leading cause of death globally with

1 in 6 deaths resulting from any form of cancer. In the United States, 1 in 3 men and women will

develop cancer at some point in their lifetime, while the mortality rate of cancer is 1 in 5 people

in the United States (Lifetime Risk of Developing or Dying From Cancer, 2018). Continuous

developments in cancer treatment have worked to reduce this unfortunate statistic, however a

majority of treatments such as radiotherapy (RT), chemotherapy (CT), and immunotherapy (IT)

can lead to long lasting impairments and extraneous damage to the human body. The focus of

this paper, proton therapy (PT), is a recent development that was discovered in 1946 by Robert

R.Wilson (History of Proton Therapy, n.d.). It functions by accelerating a proton to a high energy

state, where it is administered in the body and releases the radiation over time. It is unique due to

a phenomenon known as the Bragg’s peak, where the proton gradually releases energy as it

travels and then releases all of its energy in one dose at the end of its travel (Mohan et.al,2016).

This limits the damage to normal healthy tissue in the body, while maximizing the destruction of
tumoral cells. Over time, this type of therapy has been enhanced by gradual advancement in

technology and precision in the mechanical aspects of treatment. However, it is not perfect.

While PT has a reputation for its accuracy, safety, and effectiveness, it also has its drawbacks. It

is very sensitive to natural errors in treatment, resulting in great uncertainty in treatment that

result from many factors. These factors that influence proton therapy accuracy are range

uncertainty and setup variation which can significantly delay and reduce the effectiveness of

treatment planning. However, they can be significantly nullified through applicable methods

such as robust optimization, kilovoltage scanning, and Monte Carlo simulations.

Proton Therapy Uncertainties

Many modern cancer treatments have difficulty adapting to modern developments and

mutations in cancer cells. These cells have become more radioresistant, metastatic, and even

malignant. Relapse of cancer is at an all-time high, however, one new method of cancer

treatment has shown to work to combat these issues: proton therapy. Proton therapy is known for

its accuracy and effectiveness at eliminating cancer cells, especially with its unique property, the

Bragg peak. However, no cancer monotherapy is perfect, and proton treatment is no exception to

this. Proton particles are very sensitive to external forces and collisions with localized tissues,

therefore leading to uncertainty in proton treatment. These uncertainties can derive from various

factors; some are uncontrollable, while some are easily neutralized. However, the two main

drivers of uncertainty in proton treatment are each individual's unique anatomical structure, and

differences in proton particle properties.

First, it is necessary to discuss the various types of uncertainty in proton treatment: range

uncertainty and set-up variation. Range uncertainty is defined as the deviation of a proton
particle from its target point in the distal tumor volume. This can be attributed to various

phenomena, such as the natural properties of a proton particle including sensitivity and energy

release, and tissue density which results in protons releasing different energies at different points

in the body. Protons are very different in comparison to conventional photon (radio) therapy

through various contrasting properties such as the Bragg peak, exit dose, speed of travel, and

sensitivity. However, the most critical of these contrasting properties is the speed of travel. Since

protons travel at a very slow pace in comparison with photons, not all of the protons accelerated

to the same energy will travel the same distance (Paganetti, 2012). This results in protons

executing different Bragg peaks at different depths over time, which could potentiate risk to

close proximity organs, make uncertainty harder to calculate for proton physicians and increase

uncertainty in treatment. On the other hand, tissue densities can result in increased range of

uncertainty in therapy. This is because much of the organ tissues in the human body vary based

on density (Figure 1 shows the density values of tissues in various organ systems). These tissue

structures serve as “linear stopping powers” (LSPs) which cause the proton particles to lose

energy as they travel and approach the Bragg peak which initiates the stopping of the proton

(Palta, n.d). As protons travel through these tissues, they release energy during their movement

through the tissue. However, the proton releases different amounts of energy based on the

density of the tissue, and this density determines how much energy a proton loses over the

duration of its travel. The denser a tissue, the more energy a proton releases, and therefore the

shorter the distance it will travel. On the contrary, a less dense issue will induce less energy

emission from a proton, and therefore its course will not be hindered to the same degree. Human

body anatomy naturally does not have the same tissue over a certain distance in the body, as

various organs are in close proximity to one another and various tissue structures are utilized in
an organ structure. A clear indication of this is in the small intestine, which has epithelial tissue,

skeletal muscle tissue, and smooth muscle tissue. Moreover, the amount of each respective tissue

in each individual varies from person to person, thus leading to increased uncertainty during

treatment. Overall, it has been shown that these tissue densities and variation can cause

uncertainty between 1-3 mm away from the target tumoral volume, which can result in critical

damage to organs in close proximity to the tumor (Paganetti, 2012).

Figure 1: This table develop by Uri.M at Harvard University lists the organ identification
(ID) number, medium, density, and mass of each organ/tissue.
 Set-up variation, on the other hand, is also another factor that induces range uncertainty

during treatment due to anatomical variation. This variation is caused by the daily repositioning

of organs, muscles, bones, and other structures possibly due to epigenetics, adaptations that serve

as a stimulus response to the direct environment. This means that a patient that obtains a tumoral

CT scan one day, will not produce that same CT scan another. Therefore, these daily

repositionings increase uncertainty in proton treatment, and it remains imperative that repeated

CT scans of the tumoral area be conducted to compensate for the daily reposition of anatomical

structures (Zhang, 2007). However, these CT scans cannot be repeated within a short time

interval, otherwise, the photon radiation would induce an increased growth of cancer within the

patient's body. Due to this, it is highly unlikely that a patient will be placed in the same position

during the previous treatment fraction as the muscles and bones will not rest in the same manner

(Trofimov, 2001). This, in turn, increases range uncertainty in treatment, and ultimately makes it

more difficult to treat cancer cells.

Overall, the range uncertainties induced by set-up variation and tissue density variation

have a drastic effect on the accuracy, safety, and effectiveness of treatment. As many organs and

bodily systems are in close proximity to one another, tumors often prove difficult to treat due to

their adjacency to critical organs. This is why proton therapy physicians express persisting

concern about treatment uncertainties in the proton beams as these uncertainties may inflict

damage or unnecessary radiation dosage on other critical organs (Trofimov, 2001). This is

especially concerning, considering how the body is 75% water-based, and proton uncertainty in

an aquatic environment is inevitable. The range of margin of error for this uncertainty is ±3.5%

on the distance traveled in a water environment in addition to naturally being 1 mm off course of

the target due to natural uncertainty (Schuemann, 2014). As somatic and tumoral cells are 10-100
um in diameter, this means that uncertainty in treatment can be off by over 1,000,000 cells,

thereby killing normal body cells and allow the remaining tumoral cells to develop resistance and

to reproduce creating a more malignant tumor.

These range uncertainties cause disparities in the accuracy of treatment, ultimately

reducing its effectiveness and lengthening the treatment period. However, they can be reduced

through various optimization methods. One such method that is very prominent in the radiation

oncology world, is robust optimization.

Robust Optimization

In any form of radiation treatment, it is very difficult to confine beamlets to a certain area

of space due to the inconsistency of proton properties. For many years, proton physicians have

struggled in finding the answer to this problem. However, the answer may not lie in the scientific

aspect of proton treatment.

This potential solution is known as robust optimization, an alternative calculus-based

calculation that deals with reducing uncertainty of trajectory and beam-like compounds by

confining it to a small area in space. What makes this method unique, is that when applied to

proton therapy, it influences the objective and constraint functions that proton physicians

calculate. According to Amir Ahmadi, a professor at Princeton University and mathematical

optimization expert, robust optimization assumes that the objective and constraint functions only

belong to certain areas in a given volume, and this function area is defined as the tumoral volume

between the distal and proximal ends of the tumor (Ahmadi 2016). The proximal and distal ends

of a tumor refer to the positions of which the tumor is closest to the radiation beam and furthest

from the radiation beam respectively. This means that the proton beams in treatment are
manipulated to be confined to the normal tumoral volume, ensuring that the proton particles

engage in direct contact with the tumoral cells, ultimately killing them. However, robust

optimization is a very difficult method to apply, especially to proton therapy. As proton particles

are all positively charged, the subatomic particles tend to repel one another and ultimately split

into microbeams. This tendency makes it very difficult for proton physicians to make accurate

calculations, as thousands of beamlets in each individual dose voxel (tumoral volume) must be

computed to minimize radiation delivered to critical organs (Liu et al., 2011). Despite robust

optimization’s benefits and drawbacks, it undoubtedly reduces range uncertainty to a significant

degree.

When robust optimization is applied to the microbeams of protons, it confines the space

of the beam to reduce range uncertainty. As mentioned before, the calculation confines the beam

space into a beam that emulates the nominal dose volume, the optimal where the protons are

unaffected by uncertainties, thus resulting in a more accurate delivery and more effective

treatment (Liu et al., 2012). Ironically, robust optimization can reduce range uncertainty by using

range uncertainty as a primary source of treatment improvement. Through the use of repeated CT

scanning and recurring uncertainty calculations, range uncertainty can be preempted due to

obtained information about patient anatomy, tissue density, and many other factors which help

proton physicians create a hypothetical defined space where the beam should land on the tumor

through worst and best case optimization, ultimately reducing range uncertainty during the

treatment course (Pflugfelder et al., 2008). Robust optimization seems like the ideal solution to

reducing range uncertainty, and its overall effect on treatment is very positive.

Due to robust optimization’s confinement of proton microbeams to a confined space, the

robustness or strength of proton treatment increases along with the increased sparing of normal
tissues and critical organs from radiation in comparison to conventional methods (Liu et al.,

2011). Finally, the restriction and confinement of beamlets through robust optimization results in

each proton beamlet covering the tumoral target uniformly over the nominal dose volume,

leading to a balanced and localized single-field uniform dose distribution (LSFUD) (Liu et al.,

2012). This uniform dose distribution allows for proton particles to enter the tumor from all

angles, allowing for a stronger delivery of treatment and increased killing of malignant tumor

cells. Figure 2 illustrates the difference between conventional method PTV optimization and the

uprising robust optimization. It can easily be identified that robust optimization provides a much

Figure 2: This image shows the difference in accuracy between robust optimization and
PTV optimization.

smaller and confined space to the proton beam, even when a 3.5% uncertainty control was

applied to both plans. Moreover, robust optimization increased regularity in the shape of beam

confinement, allowing for a uniform dose distribution, which increases the effectiveness of

treatment. In conclusion, robust optimization is an objective way of reducing uncertainty,

however there are also statistical and preemptive methods to reduce uncertainties aside from

preemptive uncertainty calculation.

Alternative Methods to Reducing Range Uncertainty

 Works Cited

Density and mass of each organ-tissue.pdf. (n.d.). Retrieved from

https://bionumbers.hms.harvard.edu/files/Density%20and%20mass%20of%20each%20

organ-tissue.pdf

Paganetti, H. (2012). Range uncertainties in proton therapy and the role of Monte Carlo

simulations. Physics in Medicine and Biology, 57(11), R99–R117.

https://doi.org/10.1088/0031-9155/57/11/R99

Palta, J. R. (n.d.). Understanding the Uncertainties in Proton Therapy, 42.

Schuemann, J., Dowdell, S., Grassberger, C., Min, C. H., & Paganetti, H. (2014). Site-

specific range uncertainties caused by dose calculation algorithms for proton therapy.

Physics in Medicine and Biology, 59(15), 4007–4031. https://doi.org/10.1088/0031-

9155/59/15/4007

Trofimov, A., Nguyen, P. L., Efstathiou, J. A., Wang, Y., Lu, H.-M., Engelsman, M., …

Zietman, A. L. (2011). Interfractional variations in the set-up of pelvic bony anatomy

and soft tissue, and their implications on the delivery of proton therapy for localized

prostate cancer. International Journal of Radiation Oncology, Biology, Physics, 80(3),

928–937. https://doi.org/10.1016/j.ijrobp.2010.08.006