Proton therapy delivery: what is needed in the next ten years?

Abstract

Proton radiation therapy has been used clinically since 1952, and major advancements in the last 10 years have helped establish protons as a major clinical modality in the cancer-fighting arsenal. Technologies will always evolve, but enough major breakthroughs have been accomplished over the past 10 years to allow for a major revolution in proton therapy. This paper summarizes the major technology advancements with respect to beam delivery that are now ready for mass implementation in the proton therapy space and encourages vendors to bring these to market to benefit the cancer population worldwide. We state why these technologies are essential and ready for implementation, and we discuss how future systems should be designed to accommodate their required features.

Introduction

Proton radiation therapy has been used clinically since 1952 and major advancements in the last 10 years have helped establish protons as a major clinical modality in the cancer fighting arsenal. It is hard to argue against the use of protons for treating many solid tumors when looking only at the physical dose distributions obtained from proton beams. However, considering the means to get the calculated doses delivered to the patient in a reliable and cost-effective manner, many scholars in the field start to question the promise and validity of proton therapy. As we look at the road ahead of us from a proton beam delivery perspective, the authors believe we now have the means to make proton therapy highly effective through efficient beam delivery and patient positioning technologies. Technologies will (and should) always evolve but we believe enough major breakthroughs have been accomplished over the last 10 years to allow for a major revolution in proton therapy beam delivery. While other publications have touched on recent technical¹ and medical physics² advancements related to proton therapy, we present here beam delivery technologies and methodologies that we feel are ready for mass implementation in the proton therapy space and will encourage vendors to bring these to market to benefit the cancer population worldwide.

Delivery modalities

While the focus of this review centers on proton beam delivery, this cannot be properly addressed without also touching on patient positioning aspects. We will discuss patient positioning only in the context of achieving better and more efficient beam delivery. We will also discuss how these technologies impact the design aspects of future proton beam delivery systems.

The following is a list of “must-haves” (listed in no particular order) from a beam-delivery perspective for a proton therapy system circa 2030.

Prerequisites for a proton system in 10 years

1. Proton arc therapy.

2. Rapid beam delivery enabling FLASH therapy

3. Proton beam-based image guidance—in vivo range verification and proton beam imaging.

4. Online adaptive therapy—“on the fly”. Bring the beam to the patient.

5. Beam control based on the patient’s respiration cycle—full four-dimensional beam delivery.

6. Fast trimming apertures.

7. Beam focusing conditions—variable spot size.

8. System self-diagnostics and self-calibration—automatic warm-up and checking overnight.

The majority of the prerequisites listed above have been described extensively in recent years by many experts in the field. We will only cite a representative sample of these reviews and not discuss the technologies per se. The reader is encouraged to read these papers to obtain further details.

Arc therapy

The use of arc therapy was first proposed for photon beams in 1965 by Takahashi et al and further developed by pioneers like Brahme and Yu.^3–6 In 1997, Deasy et al proposed the use of arc therapy in protons using the distal edge tracking principle.⁷ Although the clinical value was very well appreciated, this was initially dismissed from both a technology perspective and a range uncertainty point of view.⁸ It was not until the first TomoTherapy^® units, proposed in 1993 by Mackie, were used clinically that the value of arc therapy, specifically from a dose conformity index, was realized for the photon-space.^9–12 Technological developments sparked new interest in proton arc therapy since 2013 and several groups reported on the benefits of such beam delivery techniques.^13–16 Ding et al recently showed the feasibility of delivering continuous spot-scanning proton arc (SPArc) therapy to a phantom at the William Beaumont Hospital proton therapy center using an IBA Proteus^®ONE system (Ion Beam Applications S.A., Louvain-La-Neuve, Belgium).^17,18 The dosimetric benefits of proton arc therapy are shown in Figure 1. A significant reduction in beam delivery time when using SPArc has also been demonstrated by Ding et al.¹⁸

Figure 1.

A robust optimized IMPT treatment plan for a prostate (A) compared to a SPArc plan (B). The dose distribution advantages are illustrated in the DVH (C) and dose difference (D) panels. (Used with permission from Ding et al.¹⁶ DVH, dose–volume histogram; IMPT, intensity-modulated proton therapy; SPArc, spot-scanning proton arc.

The increased benefit from protons is immediately evident when comparing intensity modulated (X-ray) radiation therapy (IMRT), in which the beam intensity can only be modulated transversely, with a proton beam where it can be controlled in all three dimensions. The “range uncertainty problem” (i.e. where the proton beam actually stops) may potentially be reduced to a clinically insignificant problem or even eliminated when proton radiography and proton tomography are used to map out the real tissue stopping powers within the patient (see next section on in vivo measurements).¹⁹ With the stopping point known to a high degree of accuracy, the treatment planning system (TPS) can optimize the dose per useable spot for each beam in the arc. An important aspect of proton arc therapy is that the high-dose spots can be placed anywhere along the beam path, i.e. not at the distal edge only. This allows for increased flexibility with respect to the optimization parameters of choice, e.g. robustness, linear energy transfer (LET),²⁰ target coverage or organ at risk (OAR) sparing.

Future proton therapy systems must be designed to provide enough free space around the patient to facilitate a collision-free rotation while maintaining the desired beam characteristics. The beam delivery systems must also be able to deliver continuous beams at variable intensities while rotating. Gating or synchronizing the beam delivery with the patient’s respiration cycle for thoracic and abdominal targets is also essential.

Fast beam delivery—FLASH therapy

FLASH radiation therapy is defined as delivering the radiation at dose rates in excess of 40 Gy/sec with the rationale that healthy tissue is spared at such high dose rates.^21,22 The underlying mechanism behind FLASH radiation is not yet fully understood but scientists believe that possible mechanisms might be oxygen deprivation of the irradiated medium and rapid recombination reducing the amount of free radicals. Both mechanisms are purely physical phenomena and do not rely on complex biological processes. This has been shown in several small animal experiments.²³ The principle of FLASH radiation is that the healthy tissues must still receive less radiation than the target but at very high dose rates.²⁴ FLASH radiation may allow for a significant reduction in the number of treatment fractions since the healthy tissue toxicity is reduced. Several groups are now actively working on techniques on delivering FLASH radiation and Phase I clinical trials are being planned to start within the next 3 years (Dee Khuntia MD. Chief Medical Officer, Varian Medical Systems. Personal communication. 27 March 2019). The authors believe that a form of arc proton therapy might be the easiest way to deliver FLASH radiation as the healthy tissue is only traversed once. Ordinary pencil beam scanning (PBS) likely cannot be used for this purpose, since the healthy tissue gets exposed each time during dose delivery to a given layer. Newly developed, ultrafast energy changes will help to deliver FLASH radiation as it allows for fast scanning in the depth dimension, delivering a cylinder of high dose rather than a single spot, i.e. “tube scanning.” The length of the “tube of dose” is tailored to the extent of the target in depth.^25,26

In order to be able to deliver FLASH proton therapy future systems must be able to measure the delivered dose accurately at the required high dose rates. The need for highly accurate patient positioning and immobilization is obvious as the number of treatment fractions will be reduced dramatically.

Proton beam-based image guidance—in vivo range verification and proton beam imaging

Modern treatment planning dose calculation algorithms can calculate the absolute dose delivered with high accurately but still struggle to calculate the range primarily due to the uncertainties in the Hounsfield unit to stopping-power conversion.^27,28 One of the properties of proton therapy is the presence of nuclear interactions that generate secondary radiation. This secondary radiation can in turn be measured to obtain valuable information related to the proton interactions inside the patient’s body. This is referred to as in vivo range verification and techniques required to facilitate this have been reviewed extensively by Knopf et al and Parodi et al.^29,30 One of the most documented and researched options involves measuring the prompt γ radiation emitted as the beam penetrates the tissue. The most attractive feature of these techniques is to detect acute changes in the patient’s anatomy, e.g. changes in the content of sinus cavities that can cause drastic changes in the depth of penetration. Analyzing the energy spectra of the prompt γ radiation allows for identifying the elemental composition of the radioactive tissue. This knowledge yields additional advantages. This work is ongoing and has been demonstrated by Hueso-González et al.³¹ Implementing prompt γ imaging in the clinical setting appears to be easier than positron emission tomography (PET) methods mainly due to problems with biological washout and the fact that the cross-section for PET isotope production decreases rapidly with decreasing proton energy, particularly in the region where the protons are stopping, which substantially decreases the accuracy of proton range estimation.^27,28

A complementary technique to reduce range uncertainty involves using the proton beam itself rather than secondary radiation. The use of proton radiography (pRad) was proposed in the 1960s with the first proton radiographs being published in 1968 by Koehler using the proton beam at the Harvard Cyclotron facility.³² Cormack, who shared the Nobel Prize with Hounsfield for inventing the CT scanner, worked with Koehler and reported on proton CT (pCT) in 1976.³³ Similar to X-ray CT, pCT images can be reconstructed from a series of adequately spaced pRad images. The use of pRad to improve the dose calculation accuracy in proton beams was demonstrated by Schneider in 1996 at the Paul Scherrer Institute in Switzerland.³⁴ The development of pRad was stalled for many years as it was believed that pCT is only possible at proton energies above 250 MeV to traverse the patient’s body for all anatomical sites. However, pRad images can play a very important role for a large percentage of cases treated with 230 MeV beams, especially in the cases where setup and range accuracy are more important. In these cases, the beams must stop before reaching critical structures (e.g. the optic chiasm or brainstem), but anatomical changes or complex anatomy, often further impacted by pretreatment surgical procedures, make it difficult to properly account for range uncertainties which can change even during the course of the treatment. A team of researchers from the Northwestern Medicine Chicago Proton Center recently reported on a study showing that 96.8 and 94.7% of proton beams used to treat head and neck and thoracic cases at that facility would have been able to traverse the patients with at least 2 cm of residual range left to facilitate a pRad image.³⁵

A pCT image data set is composed of the measured proton stopping powers in the patient’s body and therefore solves one longstanding issue in proton therapy, i.e. the uncertainties in stopping powers derived from electron densities obtained via X-ray imaging.³⁶ A pRad image is a direct measurement of cumulative stopping powers along the ray lines from the proton source to the respective point in the pRad imaging plane. Several pRad and pCT imaging systems are currently under development and several groups have already reported on actual pRad and pCT images on animals and anthropomorphic phantoms.¹⁹ The range accuracy beyond the imaged object is on the order of 0.2 mm. This is illustrated in Figure 2 which shows a proton radiograph of a CATPHAN line pair phantom obtained with the ProtonVDA system at our institute.³⁷ A 0.2 mm-thick plastic tape that was used to hold the phantom in the beam is clearly visible in the image. It is obvious that a single pRad image, like an X-ray image, will provide enough information to align the patient along the beam axis. However, as an out-of-plane rotation will change the proton range through the patient’s body, it is expected that a single pRad image may also contain enough information to reasonably correct for out-of-plane rotations thanks to the high accuracy of excess range detection. Another benefit of pRad is the significant reduction in imaging dose estimated to be 10–100 times lower than standard X-ray imaging methods.^34,35,38

Figure 2.

A proton radiograph of a CATHPHAN line pair phantom obtained with the ProtonVDA system at our facility. A 0.2 mm-thick plastic tape used to hold the phantom in place is clearly visible in the image: see the horizontal streak across the image.

The spatial resolution of pRad images will probably always be weaker than that for X-ray images due to Coulomb scattering in the patient.³⁷ However, Pankuch et al also reported on a study to investigate whether the spatial resolution of pRad images obtained with current state-of-the-art systems is adequate for patient positioning. Their study employed the use of pRad images calculated with known shifts and rotations with respect to a reference pRad image. A team of independent experts then used two-dimensional image registration techniques, equivalent to what is used in daily patient setups, to derive the applied shifts and rotations. The same process was repeated using reference and shifted X-ray images. The study revealed that the experts could obtain the corrections within clinically acceptable tolerances, i.e. less than 1 mm translations and 0.5° rotations. The differences in corrections obtained using X-ray and pRad images was statistically insignificant.³³

Future proton therapy systems must be designed to enable the deployment of in-vivo dosimetry equipment about the patient prior to (proton imaging) or during beam delivery (prompt γ or PET) without impacting the patient’s position or comfort (e.g. claustrophobia). Proton imaging also requires extremely low beam intensities which means that the beam delivery systems must be able to delivery low beam currents in the order of 3 × 10⁶ protons/sec over a 20 × 20 cm² area accurately and safely.³⁸ Achieving these milestones will then pave the way to bring online adaptive therapy into the clinical proton therapy space.

Online adaptive therapy—on the fly. Bring the beam to the patient

The standard practice in radiation therapy is to use x-ray-based imaging techniques only to determine where the patient is with respect to the radiation beam and then send a six-dimensional correction vector to the patient positioner system (PPS) to move the patient into the desired position for treatment. This is referred to as “bringing the patient to the beam.” Proton therapy provides an exciting opportunity to reverse this process. Using pRad or pCT images provides adequate information to recalculate the proton beams (either static beams or arcs) according to the patients existing position and anatomy. The former might only require a spot pattern shift while the latter will require a reoptimization. There is pushback within the community against recalculating an adaptive plan without formal verification (i.e. patient-specific qualityassurance). However, the new beams can be easily verified with a low dose pRad image prior to delivering the beam³⁵ and there is a point that is often overlooked: an adapted plan is not a new plan.³⁹ The previous beams must serve as the baseline for the adapted plan. Several parameters such as the total imparted energy (total monitor units), the maximum beam energy, and the total number of layers should be used as a baseline to ensure that the adapted beam is similar to the original beam. We tested this proposal with a two-field mediastinal plan. The isocenter was shifted by 10 mm in all three directions and a roll of 1° was introduced. The beams were reoptimized to the same dose constraints and objectives. The total number of monitor units (MU) increased by 0.8 and 0.34% for two beams respectively. The change in the number of MU per energy layer for one of the beams is illustrated in Figure 3. This shows that after adapting the beams for a relatively large displacement from isocenter the beam parameters for both beams did not change significantly. This could serve as a safety benchmark for real-time beam adaptations.

Figure 3.

A two-field mediastinal plan reoptimized for two different isocenters using the same optimization constraints and objectives. The isocenter for the second plan was shifted by +10 mm in X, Y, and Z, and a 1° roll was applied relatively to the first plan. The total number of monitor units per energy layer for the two plans for the second beam is compared in the bottom right panel. The DVH curves are compared in the upper right panel. DVH, dose–volume histogram

In order to facilitate fast online adaptive therapy in an efficient manner, the image guidance in future proton therapy systems must be done using the facility’s TPS. All modern TPS systems can perform rigid and deformable image fusions between images from multiple modalities. If this is done at the treatment console, the clinician can get an immediate evaluation of the dose on the real-time image set and can trigger an adaptation of the beam that is about to be delivered if needed.⁴⁰

Beam control based on the patient’s respiration cycle—full four-dimensional beam delivery

Cancer incidence rates published by the WHO indicate that approximately 55% of potentially fatal cancers requiring treatment are impacted by breathing motion.^41,42 The significance of breathing motion on clinical outcomes varies by anatomical location and proximity to the diaphragm.⁴¹ It is therefore crucial that future proton therapy systems are able to automatically sense this motion and coordinate the beam delivery accordingly. This can be in the form of active gating, synchronization, or active target tracking.⁴³ Audiovisual guidance is essential in this regard. Keall et al recently reviewed 27 papers addressing motion management and found that 21 of these papers indicated an advantage for audio or visual guidance. The best results were observed when both audio and visual cues were offered to the patient to help with stabilizing or controlling their breath rates.⁴⁴ Several methods can be used to track the breathing motion, but one that holds significant promise in this regard employs an infrared camera to measure the temperature of the air entering and exiting the patient’s nostrils.^45,46 That information can be used to quantify the tidal volume of air exchange that can be accurately correlated to the breathing phase and in turn synchronize the beam delivery.⁴⁵ This should be supplemented with surface rendering systems that can monitor for unwanted or involuntary patient motion during treatment.

There are third-party systems that perform motion and respiration monitoring while other systems help with reducing (High Frequency Percussive Ventilation)⁴⁷ or even eliminating respiratory motion (Breath-hold). Future systems should focus on integrating these into the control system architecture to enhance workflow efficiencies making respiration monitoring readily available per the physician’s prescription or the clinician’s judgement.⁴⁸

Fast trimming apertures

PBS revolutionized proton therapy from a usability and dose conformation perspective but the penumbra for shallower beams delivered via PBS is significantly larger than when the beam is shaped with an aperture.⁴⁹ Monte Carlo calculations now allows for accurate dose calculations of PBS with edge-sharpening apertures. The benefits of a sharper penumbra are well documented in many clinical cases, especially where the edge of the beam is in close proximity of a critical structure (e.g. parotid gland and the brainstem). Static beam-shaping apertures are a workflow impediment and can only shape the edge of the beam in the outermost extent of the target. Hyer et al proposed the use of fast-moving leaves that can trim the spots on the beam edge for every layer being delivered.^50,51 Mevion recently introduced a very innovative adaptive aperture into clinical practice.⁵² The improvement in penumbra obtained with using an aperture for classical PBS beams with in air spot sizes in the order of 3 mm σ at the highest energy, is shown in Figure 4.⁵³ As can be seen from Figure 4, the aperture or trimmer blades only improves the penumbra for ranges less than 17.5 cm. A more recent publication, using spot sizes closer to those provided by modern PBS machines established this cross-over point at a much reduced depth of 12 cm.⁵⁴ The thickness of the trimmer blades used for these beams can therefore be reduced providing the use of the trimmer is interlocked for higher energies. This will reduce the weight and improve the dynamics of such devices. Winterhalter et al⁵⁵ demonstrated that by adding a thin energy specific multileaf collimator to conventional PBS beam delivery nozzles will combine the benefits of small PBS with a sharp penumbra and will improve the conformity index for shallow targets.

Figure 4.

Penumbra (P) as a function of proton range (R) for a spot-scanning beam with MLC collimation and uniform spot weights (square) and for spot scanning beam without MLC and variable spot weight (circle). Data are provided for 3 Bragg peak depths (4, 10 and 20 cm) corresponding to proton energies of 72, 118 and 174 MeV, respectively. Analytical fits to the data are provided in order to estimate the crossing point at R. 17.5 cm, corresponding to a proton energy of 159 MeV.⁵² (Used with permission from Bues et al.⁵⁰ MLC, multi leaf collimator.

Beam focusing conditions—variable spot size

As explained above, the proton beam penumbra for shallow targets are not optimal due to the size of the beam spot.^49,53 One method to overcome this is to use a smaller spot size on the edge of the beam and larger spots in the center of the target by changing the beam focus as required. This will allow for a sharper beam edge while reducing the risk of dose inhomogeneities in the target due to potential spot position inaccuracies. It will also reduce the number of spots required to fill the target which will reduce the beam delivery time. Vendors are reluctant to implement this as it is technically challenging to install additional focusing elements, specifically in gantry systems where space is restricted. Verifying the beam focus condition with the desired accuracy is also challenging. One method that was recently proposed by Meier et al is to first deliver the spots on the edge of the beam with a small spot and then defocus the beam to a larger spot and fill in the spots in the rest of the target.^54,56

System self-diagnostics and self-calibration—automatic warm-up and checking overnight

Artificial intelligence and automation must play a much larger role in modern and new proton therapy systems.⁵⁷ While a proton system is indeed fully automated, we are still doing manual quality assurance checks by putting various external detectors in the beam merely to measure again what the control system just measured during beam delivery. Modern systems are already equipped with collision avoidance systems that prevents any mechanical collisions between hardware components of the system.⁵⁸ By simply putting a test object on the treatment couch, the entire system can be put in qualityassurance mode and it can cycle through checks during the night shift when patients are not treated. The control system can steer the beam systematically onto beam position monitors and test that the interdependencies of the different monitoring circuits are intact. The image guidance system can interact with optical tracking systems to ensure that all the systems involved in patient positioning are tracking with each other. The well-being of the system can then be validated by scrutinizing the data in the recorded log files.

Discussion

To look at the 2030 proton therapy landscape from a practical perspective we must consider three pathways:

1. Existing (legacy) facilities still in operation.

2. New facilities that will be developed using traditional approaches and employing single or multiple rooms from one accelerator;

3. New facilities that will not be bound by existing paradigms.

When looking at the prerequisites listed above, it is clear that it will be expensive to retrofit existing facilities with many of the features on the list, but we argue that such facilities should at least be capable of being retrofitted with features listed in Table 1.

Table 1.

The list of prerequisites for proton systems in 10 years that can be retrofitted to existing facilities

#	Feature	Existing facilities	New facilities —paradigm bound	New facilities to be designed
1	Proton arc therapy	?	X	X
2	Rapid beam delivery enabling FLASH therapy	?	?	X
3	Proton beam-based Image guidance—in-vivo range verification and proton beam imaging	X	X	X
4	Online adaptive therapy—on the fly. Bring the beam to the patient	?	X	X
5	Beam control based on the patient’s respiration cycle—full 4D beam delivery	X	X	X
6	Fast trimming apertures		X	X
7	Beam focusing conditions —variable spot size	?	X	X
8	System self-diagnostics and self-calibration—automatic warm-up and checking overnight	X	X	X

4D, four-dimensional.

Aquestion mark indicates that it might already be possible to upgrade anexisting facility, but it might not be cost effective or feasible.

Several vendors developed third generation proton therapy systems during the last decade as recently reviewed by Farr et al.¹ The biggest problem with upgrading existing facilities is that in most cases the system will have to be taken out of service for extended periods of time. This is extremely expensive and adds to the total cost for the upgrade.

The prerequisites discussed above has one common theme: a new approach with respect to what happens around the treatment isocenter. It calls for a fresh look at the proton therapy system starting at the treatment isocenter. We must rethink the beam delivery process from the isocenter and not the accelerator. Most of the listed technological improvements require space around the isocenter to allow for easier deployment. Having an open isocenter is one example that was proposed by Cameron et al in 2010.⁵⁹ Several vendors planned for this by designing their systems with improved patient positioning and image guidance in mind. The superconducting gantry developed by ProNova uses a cantilever design that obviates the need of a tunnel.⁶⁰ The half gantry design, first proposed and used at the Paul Scherrer Institute⁶¹ and now employed by IBA, Mevion and Protom also provide an open isocenter but have the disadvantage that the patient needs to be rotated through 180° to get beams entering from the other side of the patient.⁶² If the half-gantries become equipped with sophisticated optical guidance systems, this slight disadvantage will be mitigated. Open isocenters will allow for more effective optical image guidance.

Next generation systems must be designed such that all the features listed can be accommodated. This will allow for proton therapy systems to function at a truly differentiating level and set proton therapy apart from the most advanced X-ray therapy systems. Apart from the benefit to cancer patients, these technologies will also make proton therapy deliveries more efficient with obvious operational benefits. The latter will help to justify the higher costs of proton therapy systems.