- Email: [email protected]
Image Guidance for Proton Therapy
Clinical Oncology xxx (2018) 1e6 Contents lists available at ScienceDirect
Clinical Oncology journal homepage: www.clinicaloncologyonline.net
Overview
Image Guidance for Proton Therapy R.I. MacKay *y * Manchester
Academic Health Science Centre, Institute of Cancer Sciences, Manchester Cancer Research Centre (MCRC), The University of Manchester, Manchester, UK y Radiotherapy Related Research, The Christie NHS Foundation Trust, Manchester, UK Received 26 January 2018; received in revised form 31 January 2018; accepted 4 February 2018
Abstract Image-guided radiotherapy has an established role in all forms of radiotherapy treatment delivery. Proton therapy seeks to deliver superior dose distributions through utilising the Bragg peak to target tumour and avoid sensitive normal tissue. The Bragg peak and sharp falloff in dose delivered by proton therapy necessitate careful treatment planning and treatment delivery. The dose distribution delivered by proton therapy is particularly sensitive to uncertainty in the prediction of proton range during treatment planning and deviations from the planned delivery during the course of the fractionated treatment. Realising the superior dose distribution of proton therapy requires increased diligence and image guidance has a key role in ensuring that treatments are planned and delivered. This article will outline the current status of image guidance for proton therapy, particularly highlighting differences with regard to high-energy X-ray therapy, and will look at a number of future improvements in image-guided proton therapy. Ó 2018 Published by Elsevier Ltd on behalf of The Royal College of Radiologists.
Keywords: Image guidance; Proton
Introduction Radiotherapy in all its forms involves depositing a potentially lethal dose of radiation to the tumour to cure patients of cancer. In external beam therapy, where radiation beams are targeted from outside the patient, delivering the prescribed dose to the defined target volume while sparing surrounding normal tissue is of upmost importance. In practice the accurate and precise delivery of radiotherapy throughout a course of treatment requires technical treatment planning, to identify disease and target the radiation dose, along with on-treatment monitoring to ensure the patient is treated as planned. Imaging in many forms is a key tool used in both treatment planning and treatment delivery. The term ‘image-guided radiotherapy’ is often defined as the use of imaging at the pretreatment and treatment stage that leads to an action to improve or verify the accuracy of radiotherapy [1]. This article will examine the difference in image guidance for proton therapy and Author for correspondence: The Christie NHS Foundation Trust, Manchester, UK. E-mail address: [email protected].
high-energy X-ray therapy for both treatment planning and treatment delivery. The advantage of proton therapy through the delivery of dose by the Bragg peak is well known. To utilise this advantage in terms of patient outcome, particular attention must be applied to predicting the range of the proton beam at which the Bragg peak delivers its dose, during pretreatment planning and during treatment delivery. In the pretreatment planning, imaging is required for target and normal tissue delineation and also as the basis for the dose calculation to assess where the Bragg peak delivers its dose. During the delivery of fractionated radiotherapy, anatomical changes in the patient in proton therapy alter the position of the Bragg peak and change the dose distribution, potentially leading to dose inhomogeneity in the target and/or overdose of normal tissue. There are differences between proton and high-energy X-ray therapy that lead to image guidance being used in different ways. In treatment planning there is additional difficulty in using computed tomography (CT) images in dose calculation algorithms for proton therapy. In the treatment room, cone beam CT and full diagnostic CT are beginning to be used, but developments have been slow
https://doi.org/10.1016/j.clon.2018.02.004 0936-6555/Ó 2018 Published by Elsevier Ltd on behalf of The Royal College of Radiologists.
Please cite this article in press as: MacKay RI, Image Guidance for Proton Therapy, Clinical Oncology (2018), https://doi.org/10.1016/ j.clon.2018.02.004
2
R.I. MacKay / Clinical Oncology xxx (2018) 1e6
compared with other forms of radiotherapy. There are also additional modes of imaging associated with directly imaging the dose deposition of protons that are termed ‘range verification’ and in the future the proton community is looking towards the advantages of imaging with the proton itself.
Imaging for Treatment Planning Many imaging modalities used in treatment planning for proton therapy are the same as those used for conventional radiotherapy, and in many of the steps of the planning process, for instance volume delineation, imaging systems, such as CT, magnetic resonance imaging and positron emission tomography (PET), can be used in the same manner for proton therapy as they are for high-energy Xrays. However, there are differences in the way that images can be used in proton therapy treatment planning. The primary imaging modality in radiotherapy is the CT scan and it is used to provide anatomical information on the patient and also as the primary image upon which a dose calculation can be made. To allow a CT image to be used in a dose calculation, the Hounsfield units (HU) that form the CT image must first be used to derive other parameters. In high-energy X-ray therapy, HU are related to electron or physical density, which are the parameters required to allow the calculation of the dose deposited by megavoltage photons. In proton therapy, the HU must be converted into proton stopping powers before the dose calculation can be carried out. The process of deriving stopping powers from HU is complex, as it aims to predict the behaviour of one type of particle (the proton, a charged particle) from the behaviour of a completely different particle (the massless, chargeless photon used to acquire the kilovoltage CT image). There is, therefore, an uncertainty associated with the derived stopping powers, which is consequently a source of uncertainty within the dose calculation itself. The conventional method for calibrating a CT scanner for proton therapy is the stoichiometric method [2], which is a two-step process. In step one, the scanner is characterised by taking scans of tissue substitute materials with known compositions and parameterising the relationship between the measured HU and the material properties. In step two this parameterisation is applied to a set of published reference tissue compositions, to generate a relationship between the stopping powers (as calculated from the tissue composition) and the associated HU values (as calculated from the relationships derived in step one). As noted, this calibration is more complex than that for high-energy X-rays and more than one calibration curve may be required, for instance for different sizes of patient. Given the importance of range in proton therapy, a final check of proton range in real animal tissue imaged on a CT scanner is often made. Even with great care, the stoichiometric method of calibration still results in uncertainty in the proton dose calculation, and the related prediction of proton range. It is for this reason that there is considerable interest in dual-
energy CT for proton therapy. Although proton therapy is not the prime motivation for the commercial development of dual-energy CT, the ability of scanners to improve tissue characterisation has the potential to improve the estimation of proton stopping powers and is of great interest. The degree of improvement is the topic of much current research [3] and depends on a number of factors, including the design of the dual-energy scanner, the particular method used to determine stopping power from the dual-energy imaging and the implementation of dual-energy scans in the treatment planning system. However, several researchers have reported a clinically significant reduction in range uncertainty and incorporating dual-energy CT scans into treatment planning systems will allow this to be realised clinically.
Kilovoltage Imaging for Patient Set-up Proton therapy was an early implementer of image guidance in the treatment room. The increased dosimetric consequence of changes in patient position necessitated inroom imaging to verify patient set-up on a daily basis. In the early implementation of proton therapy this was achieved through the greater use of kilovoltage imaging in the treatment room. Kilovoltage imaging systems have been routinely used in conjunction with fixed beam and gantrybased proton therapy treatments. Typically, twodimensional kilovoltage imaging would be used before delivery of the treatment field. A two-dimensional kilovoltage image taken at set-up can be compared against a digital reconstructed radiograph produced from the planning CT scan [4]. The angles for verification must be chosen at the planning stage. Often the angles will correspond to the beam directions, but other angles may be chosen specifically to assist with patient set-up. In addition to the patient position on the couch, particular attention must be paid to the position of bony anatomy such as the pelvis, as this will affect the range on the proton field. Kilovoltage imaging is an effective way to ensure that the patient is in as similar position as practicable to that in which the proton therapy treatment plan was prepared. However, two-dimensional kilovoltage imaging has an inherent limitation, in that it is difficult to assess the effect of anatomical changes that occur during treatment on the planned dose distribution.
Volumetric Imaging in the Treatment Room To a certain degree, X-ray therapy was image guided before the evolution of volumetric imaging in the treatment room. For many years, megavoltage imaging was used in the treatment room to correct patient set-up and some stereoscopic kilovoltage imaging systems were used clinically [5]. However, it was the advent of volumetric CT imaging in the linear accelerator bunker that sparked the revolution of image-guided radiotherapy. Volumetric imaging in the
Please cite this article in press as: MacKay RI, Image Guidance for Proton Therapy, Clinical Oncology (2018), https://doi.org/10.1016/ j.clon.2018.02.004
R.I. MacKay / Clinical Oncology xxx (2018) 1e6
treatment room for high-energy X-ray therapy can be provided by a CT scanner in the treatment room or by cone beam CT. In proton therapy, with the additional clinical need for image guidance, one might expect that the technologies used in X-ray radiotherapy would have found early application. However, in contrast to kilovoltage image guidance, proton therapy was not an early implementer of CT in the treatment room. To understand why, one needs to consider some of the practical aspects of both cone beam CT and full diagnostic CT. Full diagnostic CT in the treatment room offers the same quality CT as used for treatment planning. In theory this looks like an ideal tool for patient set-up and adaptive radiotherapy planning. However, many practicalities in the application of CT in the room can limit its applicability. The application of CT scanners in the treatment room requires effective transfer of the patient from the CT scanner to the position of treatment. Any uncertainty that arises due to the movement of the patient between imaging and treatment negates the effectiveness of image guidance. Crucial to effective workflow is rigid registration of the CT scanner to the treatment unit so that a common co-ordinate system can be defined for imaging and treatment. In X-ray therapy, the transfer was usually achieved by rails between the CT and linear accelerator. After imaging, the patient would be translated from the CT bore to the linear accelerator isocentre, a move that required a 180 degree rotation. In proton therapy, robotic couches are generally used and an in-room CT system must translate the patient from the CT scanner to the treatment machine. The practicalities of positioning the patient in the bore and translating and positioning in the treatment position at a speed that is safe, comfortable and does not induce motion in the patient is challenging. The desire to utilise CT for monitoring patient set-up and as the tool for adaptive therapy still leads many to attempt to overcome these problems. CT has been incorporated into CT gantry rooms and some centres have used a CT scanner out of the bunker for daily patient set-up. In instances where CT is used out of the room, patients must be moved from the scanner into the treatment room on a motorised treatment couch [6]. In high-energy X-ray therapy, CT in the treatment room was rapidly usurped by cone beam CT and, provided effective cone beam systems become available in proton therapy, one would expect a similar trend. The cone beam CT systems that are widely applied in radiotherapy do not provide diagnostic quality images. It is recognised that cone beam images suffer with respect to imaging artefacts in comparison to fan beam CT. The artefacts in both fan beam and cone beam CT are many and varied but it is apparent that cone beam is inferior with respect to aliasing, noise, scatter and movement artefacts. Such artefacts have an adverse impact on the potential application of cone beam for both proton and X-ray radiotherapy [7]. In proton therapy, the artefacts typically found in cone beam images affect the predicted range of the protons in the patient, producing greater dosimetric errors. For this reason it has long been felt that if cone beam imaging was of
3
comparable quality to that used in X-ray radiotherapy, it would be more difficult to utilise in proton therapy [8]. However, many would argue that with appropriate cone beam correction methods or the use of software that will perform deformable registration of cone beam images to planning scans, the utility of cone beam images in proton therapy can be improved to the point where they can be used in the process of assessing the dosimetric consequences of patient changes and adaption of the treatment [9]. The quality of images has certainly been a factor in the speed in which proton therapy equipment manufacturers have developed cone beam solutions integrated with their treatment units. Another factor hampering the speed of manufacturer development is the very low number of proton treatment rooms in comparison to X-ray radiotherapy and the consequential difficulty in recouping the development costs of cone beam for proton therapy. However, many major proton therapy manufacturers have now implemented or are developing cone beam solutions. The imaging systems are either integrated with the gantry, on a standalone robotic arm or integrated with the treatment couch. It is an area where there is still considerable scope for improvement in both hardware and software. One area that researchers have only recently begun to look at is dualenergy cone beam CT, which has the potential to improve tissue contrast and reduce beam hardening [10,11].
Imaging for Range Verification In X-ray radiotherapy, anatomical imaging in combination with dose modelling is the best method to determine the likely dose to the patient. The deposition of dose from an X-ray beam cannot be imaged directly. Proton therapy offers the intriguing prospect of being able to image the dose deposited in the patient by a more direct means. There are two methods that are widely studied to produce meaningful on-treatment imaging of dose deposition. PET can measure the gamma pair emitted by the annihilation of positrons produced by unstable isotopes created by inelastic interaction of the proton beam with patient tissues. A conventional PET scanner can measure the positron emission of a number of different isotopes with varying half-lives. The PET signal that is imaged is not directly a measurement of dose, so the PET image cannot be directly compared with the three-dimensional dose distribution predicated from treatment planning but can be compared with a Monte Carlo simulation predicted distribution of positron emitters [12]. The feasibility of using PET as an effective verification tool for proton therapy depends on a number of factors. When using a commercial PET scanner that may be some distance from the treatment room it will take time to remove the patient from the treatment couch, move them to the PET imager and set up for imaging in the treatment position. During this time the PET signal diminishes due to the half-life of the radioactive isotopes produced and biological washout of the PET activity. As the PET signal
Please cite this article in press as: MacKay RI, Image Guidance for Proton Therapy, Clinical Oncology (2018), https://doi.org/10.1016/ j.clon.2018.02.004
4
R.I. MacKay / Clinical Oncology xxx (2018) 1e6
diminishes the scanning time must be increased to collect a usable PET image. Different proton fields in the treatment delivery may be delivered minutes apart and so when imaged the activity will be dependent on the order of treatment delivery. As the time between imaging and treatment increases it becomes more difficult to predict the expected PET distribution using Monte Carlo simulation. These difficulties have led to attempts to perform PET scanning in the room [13]. Either through a standalone PET scanner in the treatment room or a PET scanning system integrated with the treatment gantry. Feasibility studies have shown that a standalone scanner in the treatment room can cut down the gap between treatment and imaging to 2.5 min and the scan time to 5 min, reduced from an estimated 30 min when the scanner is out of the room. Even with these timescales it was found that the modelling of the biological washout produced considerable uncertainty. In addition, imaging would have a significant impact on patient throughput in a busy clinical centre. To streamline this process further, ‘in beam’ PET has been developed, where PET detectors are positioned around the treatment beam so the patient can be imaged in the treatment position. Planar PET detectors have been integrated with a gantry to image the isocentre position [14]. In this system, imaging can be initiated immediately after the delivery of the treatment beam and acquisition times of 200 s were used. Rather than compare the planar image with a predicted PET distribution, the image from subsequent fractions was compared with the first fraction and time trends associated with changes in the tumour were observed. The other method being explored for range verification imaging is measurement of the prompt gamma emission produced by inelastic interactions of protons with the target nuclei. The gamma emission is a result of the excited nucleus returning to its ground state. Gamma emission from a prompt gamma interaction is more immediate than that from positron decay and is directly correlated with dose. There are a number of ways that have been used to measure the prompt gamma signal, including a single scintillator, gamma camera knife edge slit, multi slit detector and a Compton camera [15,16]. Currently, only one system has been developed commercially and tested clinically [17]. A prototype knife edge slit camera developed by Oncoray (Dresden, Germany) and IBA (Louvain, Belgium) was used to measure interactional range in passive scattered proton therapy with an accuracy of 2 mm. The same system was later used to verify spot scanning proton therapy, where range of the individual energy layers could be verified to an accuracy of 1.3 mm or better [18]. The prototype trolley-based system shows the feasibility of clinically meaningful prompt gamma measurement. Due to the inherent problem of range uncertainty in proton therapy and its potential clinical consequence, both the PET and prompt gamma methods of range verification have received considerable attention from researchers. However, there is still much work to do in developing an effective system to image range in clinical proton therapy.
The inherent difficulties around isotope decay and biological washout for PET verification have prevented progress in recent years. Prompt gamma has been developed to a commercial prototype and seems to have more clinical potential. For widespread clinical use there is still much hardware development required to produce an integrated easy to use measurement system. As a clinical tool, range imaging will provide valuable feedback that will reduce systematic uncertainties in treatment planning and flag deviations in patient set-up or anatomy that should trigger plan adaption.
Proton Computed Tomography and Proton Radiography Proton CT and proton radiography are both imaging techniques that attempt to use proton beams to image the patient. Both techniques rely on increasing the proton energy beyond what would usually be used to treat the tumour so that the proton beam can pass through the area of the patient being imaged. These techniques offer great promise to reduce uncertainties through the improved measurement of stopping power and have been in development for many years. However, currently of all the techniques discussed here, these techniques are perhaps the furthest from clinical application. In proton radiography, a transmission image can be formed by measuring the proton position at the entrance and exit of the patient and its residual energy after transmission through the patient [19]. In proton CT, similar technology can be used to take a series of images from different angles in an analogous manner to X-ray CT but the resulting volumetric image can be reconstructed from the measured energy loss of the protons as they pass through the patient. One of the obvious difficulties in developing proton imaging is the cost of the equipment needed to accelerate and detect the protons that form the images. However, in proton imaging for proton therapy the equipment to accelerate the protons is already a part of the treatment system. Despite this there are some important technological considerations. Typically, proton therapy requires maximum proton energy of around 250 MeV. This is high enough to target deepseated tumours but not always high enough for proton imaging where the protons need to pass through the complete patient with a residual energy that can be measured. Although smaller patients, particularly paediatrics, could be imaged with conventional clinical equipment, there has been much interest in increasing the maximum energy of accelerators [20,21] or boosting the energy specifically for imaging. Further to the accelerating technology, imaging requires proton tracking technology to measure the position and residual energy of the protons that are transmitted through the patient. A number of groups have developed proton imaging using position sensitive detectors and range telescopes [19]. In the UK, the Proton Radiotherapy Verification and Dosimetry Applications (PRaVDA) consortium has been at
Please cite this article in press as: MacKay RI, Image Guidance for Proton Therapy, Clinical Oncology (2018), https://doi.org/10.1016/ j.clon.2018.02.004
R.I. MacKay / Clinical Oncology xxx (2018) 1e6
the forefront of the development of a solid state system to perform proton CT. The prototype system has been used to image phantoms and the group has investigated proton CT based on many parameters, including proton stopping power, proton scattering power, proton attenuation power and proton straggling power. The integration of such a system with a clinical spot scanning proton therapy system is of considerable interest and hopefully will be the next phase of development. The first clinical application of proton CT may be to improve the characterisation of proton stopping powers derived from the CT scan used in pretreatment imaging. However, in the future, proton CT could be the primary image used in treatment planning and eventually could provide daily in-room 3D images of stopping power. Such images would be the ideal images for adaptive proton therapy.
Image Guidance in UK Centres In the UK, the National Health Service centres are being built with a range of image-guided equipment. For pretreatment planning, CT and magnetic resonance imaging have been integrated on the clinical treatment floor to image proton therapy patients. The centres will be able to perform dual-energy CT, but will probably implement the standard stoichiometric calibration before the first treatment. In the treatment room, two kilovoltage imaging systems are incorporated into the treatment gantry to allow two-dimensional planar and three-dimensional cone beam imaging in the treatment position. In-room cone beam CT will be used for patient set-up and to flag when adaption of the treatment plan is necessary. In many cases the adaption of the treatment plan will involve repeating the pretreatment CT imaging when any significant revision is required. The requirement for one in-treatment diagnostic CT scan has been incorporated into the treatment workflow. This is the starting point for a clinical service that should be in place for 20 years. Obviously it is hoped that many of the image guidance methods listed in this article will become a clinical reality over the lifetime of the centres. Both institutions have active research programmes in image guidance for radiotherapy.
Conclusion The need for image guidance in proton therapy is clear. It is perhaps surprising that with the number of potential developments to improve proton therapy treatment and delivery that clinical application seems limited, particularly with the attention given to image guidance for high-energy X-rays. As the number of proton centres grows so will the demand for improved image guidance. Researchers in more institutions will turn their attention to image guidance and manufacturers will have improved resources to translate
5
systems into their clinical products. Such developments can only help to realise the clinical proton advantage offered by the Bragg peak, ensuring that improved dose distributions are planned and delivered.
References [1] National Cancer Action Team National Radiotherapy Implementation Group Report. Image guided radiotherapy (IGRT) guidance for implementation and use 2012. [2] Schneider U, Pedroni E, Lomax A. The calibration of CT Hounsfield units for radiotherapy treatment planning. Phys Med Biol 1996;41(1):111e124. €r E, Lalonde A, Royle G, Lu HM, Bouchard H. The potential of [3] Ba dual-energy CT to reduce proton beam range uncertainties. Med Phys 2017;44(6):2332e2344. [4] Schulte RW. Strategies for image-guided proton therapy of cancer. Eur Oncol Dis 2007;1(2):116e118. [5] Shirato H, Shimizu S, Kitamura K, Nishioka T, Kagei K, Hashimoto S, et al. Four-dimensional treatment planning and fluoroscopic real-time tumor tracking radiotherapy for moving tumor. Int J Radiat Oncol Biol Phys 2000;48:435e442. [6] Devicienti S, Strigari L, D’Andrea M, Benassi M, Dimiccoli V, Portaluri M. Patient positioning in the proton radiotherapy era. J Exp Clin Cancer Res 2010;29(1):47. [7] Schulze R, Heil U, Gross D, Bruellman DD, Dranischnikow E, Schwanecke U, et al. Artefacts in CBCT: a review. Dentomaxillofac Radiol 2011;40(5):265e273. https://doi.org/10.1259/ dmfr/30642039. [8] Park YK, Sharp GC, Phillips J, Winey BA. Proton dose calculation on scatter-corrected CBCT image: feasibility study for adaptive proton therapy. Med Phys 2015;42(8):4449e4459. [9] Veiga C, Janssens G, Teng C-L, Baudier T, Hotoiu L, McClelland JR, et al. First clinical investigation of cone beam computed tomography and deformable registration for adaptive proton therapy for lung cancer. Int J Radiat Oncol Biol Phys 2016;95(1):549e559. [10] Peng H, Kanehira T, Takao S, Matsuura T, Umegaki K, Shirato H, et al. Feasibility study of using dual-energy cone beam computed tomography (DE-CBCT) in proton therapy treatment planning. Int J Radiat Oncol Biol Phys 2016;96:E697. [11] Kovacs DG, Rechner L, Berthelsen AK, Aznar MC. Dual energy cone beam computed tomography for image guided radiation therapy. Int J Radiat Oncol Biol Phys 2016;96:E689. ~ a S, Paganetti H. Monte Carlo patient [12] Moteabbed M, Espan study on the comparison of prompt gamma and PET imaging for range verification in proton therapy. Phys Med Biol 2011; 56(4):1063. ~ a S, Daartz J, Liebsch N, Ouyang J, Paganetti H, [13] Zhu X, Espan et al. Monitoring proton radiation therapy with in-room PET imaging. Phys Med Biol 2011 Jul 7;56(13):4041e4057. [14] Nishio T, Miyatake A, Ogino T, Nakagawa K, Saijo N, Esumi H. The development and clinical use of a beam ON-LINE PET system mounted on a rotating gantry port in proton therapy. Int J Radiat Oncol Biol Phys 2010;76(1):277e286. https://doi. org/10.1016/j.ijrobp.2009.05.065. [15] Min C. Prompt gamma measurements for locating the dose falloff region in the proton therapy. Appl Phys Lett 2006;89: 183517. [16] Smeets J, Roellinghoff F, Prieels D, Stichelbaut F, Benilov A, Busca P, et al. Prompt gamma imaging with a slit camera for real-time range control in proton therapy. Phys Med Biol 2012; 57(11):3371e3405. https://doi.org/10.1088/0031-9155/57/11/ 3371.
Please cite this article in press as: MacKay RI, Image Guidance for Proton Therapy, Clinical Oncology (2018), https://doi.org/10.1016/ j.clon.2018.02.004
6
R.I. MacKay / Clinical Oncology xxx (2018) 1e6
[17] Richter C, Pausch G, Barczyk S, Priegnitz M, Keitz I, Thiele J, et al. First clinical application of a prompt gamma based in vivo proton range verification system. Radiother Oncol 2016;118(2):232e237. https://doi.org/10.1016/j.radonc.2016. 01.004. [18] Xie Y, Bentefour EH, Janssens G, Smeets J, Vander Stappen F, Hotoiu L, et al. Prompt gamma imaging for in vivo range verification of pencil beam scanning proton therapy. Int J Radiat Oncol Biol Phys 2017;99(1):210e218. https://doi.org/10. 1016/j.ijrobp.2017.04.027.
[19] Poludniowski G, Allinson NM, Evans PM. Proton radiography and tomography with application to proton therapy. Br J Radiol 2015;88(1053):20150134. https://doi.org/10.1259/bjr. 20150134. [20] Apsimon R, Burt G, Pitman S, Owen H. ProBE e proton boosting extension for imaging and therapy 2016. [21] Al Harbi N, Lee SY. Design of a compact synchrotron for medical applications. Rev Sci Instr 2003;74:2540e2545. https://doi.org/10.1063/1.1561598.
Please cite this article in press as: MacKay RI, Image Guidance for Proton Therapy, Clinical Oncology (2018), https://doi.org/10.1016/ j.clon.2018.02.004
Clinical Oncology journal homepage: www.clinicaloncologyonline.net
Overview
Image Guidance for Proton Therapy R.I. MacKay *y * Manchester
Academic Health Science Centre, Institute of Cancer Sciences, Manchester Cancer Research Centre (MCRC), The University of Manchester, Manchester, UK y Radiotherapy Related Research, The Christie NHS Foundation Trust, Manchester, UK Received 26 January 2018; received in revised form 31 January 2018; accepted 4 February 2018
Abstract Image-guided radiotherapy has an established role in all forms of radiotherapy treatment delivery. Proton therapy seeks to deliver superior dose distributions through utilising the Bragg peak to target tumour and avoid sensitive normal tissue. The Bragg peak and sharp falloff in dose delivered by proton therapy necessitate careful treatment planning and treatment delivery. The dose distribution delivered by proton therapy is particularly sensitive to uncertainty in the prediction of proton range during treatment planning and deviations from the planned delivery during the course of the fractionated treatment. Realising the superior dose distribution of proton therapy requires increased diligence and image guidance has a key role in ensuring that treatments are planned and delivered. This article will outline the current status of image guidance for proton therapy, particularly highlighting differences with regard to high-energy X-ray therapy, and will look at a number of future improvements in image-guided proton therapy. Ó 2018 Published by Elsevier Ltd on behalf of The Royal College of Radiologists.
Keywords: Image guidance; Proton
Introduction Radiotherapy in all its forms involves depositing a potentially lethal dose of radiation to the tumour to cure patients of cancer. In external beam therapy, where radiation beams are targeted from outside the patient, delivering the prescribed dose to the defined target volume while sparing surrounding normal tissue is of upmost importance. In practice the accurate and precise delivery of radiotherapy throughout a course of treatment requires technical treatment planning, to identify disease and target the radiation dose, along with on-treatment monitoring to ensure the patient is treated as planned. Imaging in many forms is a key tool used in both treatment planning and treatment delivery. The term ‘image-guided radiotherapy’ is often defined as the use of imaging at the pretreatment and treatment stage that leads to an action to improve or verify the accuracy of radiotherapy [1]. This article will examine the difference in image guidance for proton therapy and Author for correspondence: The Christie NHS Foundation Trust, Manchester, UK. E-mail address: [email protected].
high-energy X-ray therapy for both treatment planning and treatment delivery. The advantage of proton therapy through the delivery of dose by the Bragg peak is well known. To utilise this advantage in terms of patient outcome, particular attention must be applied to predicting the range of the proton beam at which the Bragg peak delivers its dose, during pretreatment planning and during treatment delivery. In the pretreatment planning, imaging is required for target and normal tissue delineation and also as the basis for the dose calculation to assess where the Bragg peak delivers its dose. During the delivery of fractionated radiotherapy, anatomical changes in the patient in proton therapy alter the position of the Bragg peak and change the dose distribution, potentially leading to dose inhomogeneity in the target and/or overdose of normal tissue. There are differences between proton and high-energy X-ray therapy that lead to image guidance being used in different ways. In treatment planning there is additional difficulty in using computed tomography (CT) images in dose calculation algorithms for proton therapy. In the treatment room, cone beam CT and full diagnostic CT are beginning to be used, but developments have been slow
https://doi.org/10.1016/j.clon.2018.02.004 0936-6555/Ó 2018 Published by Elsevier Ltd on behalf of The Royal College of Radiologists.
Please cite this article in press as: MacKay RI, Image Guidance for Proton Therapy, Clinical Oncology (2018), https://doi.org/10.1016/ j.clon.2018.02.004
2
R.I. MacKay / Clinical Oncology xxx (2018) 1e6
compared with other forms of radiotherapy. There are also additional modes of imaging associated with directly imaging the dose deposition of protons that are termed ‘range verification’ and in the future the proton community is looking towards the advantages of imaging with the proton itself.
Imaging for Treatment Planning Many imaging modalities used in treatment planning for proton therapy are the same as those used for conventional radiotherapy, and in many of the steps of the planning process, for instance volume delineation, imaging systems, such as CT, magnetic resonance imaging and positron emission tomography (PET), can be used in the same manner for proton therapy as they are for high-energy Xrays. However, there are differences in the way that images can be used in proton therapy treatment planning. The primary imaging modality in radiotherapy is the CT scan and it is used to provide anatomical information on the patient and also as the primary image upon which a dose calculation can be made. To allow a CT image to be used in a dose calculation, the Hounsfield units (HU) that form the CT image must first be used to derive other parameters. In high-energy X-ray therapy, HU are related to electron or physical density, which are the parameters required to allow the calculation of the dose deposited by megavoltage photons. In proton therapy, the HU must be converted into proton stopping powers before the dose calculation can be carried out. The process of deriving stopping powers from HU is complex, as it aims to predict the behaviour of one type of particle (the proton, a charged particle) from the behaviour of a completely different particle (the massless, chargeless photon used to acquire the kilovoltage CT image). There is, therefore, an uncertainty associated with the derived stopping powers, which is consequently a source of uncertainty within the dose calculation itself. The conventional method for calibrating a CT scanner for proton therapy is the stoichiometric method [2], which is a two-step process. In step one, the scanner is characterised by taking scans of tissue substitute materials with known compositions and parameterising the relationship between the measured HU and the material properties. In step two this parameterisation is applied to a set of published reference tissue compositions, to generate a relationship between the stopping powers (as calculated from the tissue composition) and the associated HU values (as calculated from the relationships derived in step one). As noted, this calibration is more complex than that for high-energy X-rays and more than one calibration curve may be required, for instance for different sizes of patient. Given the importance of range in proton therapy, a final check of proton range in real animal tissue imaged on a CT scanner is often made. Even with great care, the stoichiometric method of calibration still results in uncertainty in the proton dose calculation, and the related prediction of proton range. It is for this reason that there is considerable interest in dual-
energy CT for proton therapy. Although proton therapy is not the prime motivation for the commercial development of dual-energy CT, the ability of scanners to improve tissue characterisation has the potential to improve the estimation of proton stopping powers and is of great interest. The degree of improvement is the topic of much current research [3] and depends on a number of factors, including the design of the dual-energy scanner, the particular method used to determine stopping power from the dual-energy imaging and the implementation of dual-energy scans in the treatment planning system. However, several researchers have reported a clinically significant reduction in range uncertainty and incorporating dual-energy CT scans into treatment planning systems will allow this to be realised clinically.
Kilovoltage Imaging for Patient Set-up Proton therapy was an early implementer of image guidance in the treatment room. The increased dosimetric consequence of changes in patient position necessitated inroom imaging to verify patient set-up on a daily basis. In the early implementation of proton therapy this was achieved through the greater use of kilovoltage imaging in the treatment room. Kilovoltage imaging systems have been routinely used in conjunction with fixed beam and gantrybased proton therapy treatments. Typically, twodimensional kilovoltage imaging would be used before delivery of the treatment field. A two-dimensional kilovoltage image taken at set-up can be compared against a digital reconstructed radiograph produced from the planning CT scan [4]. The angles for verification must be chosen at the planning stage. Often the angles will correspond to the beam directions, but other angles may be chosen specifically to assist with patient set-up. In addition to the patient position on the couch, particular attention must be paid to the position of bony anatomy such as the pelvis, as this will affect the range on the proton field. Kilovoltage imaging is an effective way to ensure that the patient is in as similar position as practicable to that in which the proton therapy treatment plan was prepared. However, two-dimensional kilovoltage imaging has an inherent limitation, in that it is difficult to assess the effect of anatomical changes that occur during treatment on the planned dose distribution.
Volumetric Imaging in the Treatment Room To a certain degree, X-ray therapy was image guided before the evolution of volumetric imaging in the treatment room. For many years, megavoltage imaging was used in the treatment room to correct patient set-up and some stereoscopic kilovoltage imaging systems were used clinically [5]. However, it was the advent of volumetric CT imaging in the linear accelerator bunker that sparked the revolution of image-guided radiotherapy. Volumetric imaging in the
Please cite this article in press as: MacKay RI, Image Guidance for Proton Therapy, Clinical Oncology (2018), https://doi.org/10.1016/ j.clon.2018.02.004
R.I. MacKay / Clinical Oncology xxx (2018) 1e6
treatment room for high-energy X-ray therapy can be provided by a CT scanner in the treatment room or by cone beam CT. In proton therapy, with the additional clinical need for image guidance, one might expect that the technologies used in X-ray radiotherapy would have found early application. However, in contrast to kilovoltage image guidance, proton therapy was not an early implementer of CT in the treatment room. To understand why, one needs to consider some of the practical aspects of both cone beam CT and full diagnostic CT. Full diagnostic CT in the treatment room offers the same quality CT as used for treatment planning. In theory this looks like an ideal tool for patient set-up and adaptive radiotherapy planning. However, many practicalities in the application of CT in the room can limit its applicability. The application of CT scanners in the treatment room requires effective transfer of the patient from the CT scanner to the position of treatment. Any uncertainty that arises due to the movement of the patient between imaging and treatment negates the effectiveness of image guidance. Crucial to effective workflow is rigid registration of the CT scanner to the treatment unit so that a common co-ordinate system can be defined for imaging and treatment. In X-ray therapy, the transfer was usually achieved by rails between the CT and linear accelerator. After imaging, the patient would be translated from the CT bore to the linear accelerator isocentre, a move that required a 180 degree rotation. In proton therapy, robotic couches are generally used and an in-room CT system must translate the patient from the CT scanner to the treatment machine. The practicalities of positioning the patient in the bore and translating and positioning in the treatment position at a speed that is safe, comfortable and does not induce motion in the patient is challenging. The desire to utilise CT for monitoring patient set-up and as the tool for adaptive therapy still leads many to attempt to overcome these problems. CT has been incorporated into CT gantry rooms and some centres have used a CT scanner out of the bunker for daily patient set-up. In instances where CT is used out of the room, patients must be moved from the scanner into the treatment room on a motorised treatment couch [6]. In high-energy X-ray therapy, CT in the treatment room was rapidly usurped by cone beam CT and, provided effective cone beam systems become available in proton therapy, one would expect a similar trend. The cone beam CT systems that are widely applied in radiotherapy do not provide diagnostic quality images. It is recognised that cone beam images suffer with respect to imaging artefacts in comparison to fan beam CT. The artefacts in both fan beam and cone beam CT are many and varied but it is apparent that cone beam is inferior with respect to aliasing, noise, scatter and movement artefacts. Such artefacts have an adverse impact on the potential application of cone beam for both proton and X-ray radiotherapy [7]. In proton therapy, the artefacts typically found in cone beam images affect the predicted range of the protons in the patient, producing greater dosimetric errors. For this reason it has long been felt that if cone beam imaging was of
3
comparable quality to that used in X-ray radiotherapy, it would be more difficult to utilise in proton therapy [8]. However, many would argue that with appropriate cone beam correction methods or the use of software that will perform deformable registration of cone beam images to planning scans, the utility of cone beam images in proton therapy can be improved to the point where they can be used in the process of assessing the dosimetric consequences of patient changes and adaption of the treatment [9]. The quality of images has certainly been a factor in the speed in which proton therapy equipment manufacturers have developed cone beam solutions integrated with their treatment units. Another factor hampering the speed of manufacturer development is the very low number of proton treatment rooms in comparison to X-ray radiotherapy and the consequential difficulty in recouping the development costs of cone beam for proton therapy. However, many major proton therapy manufacturers have now implemented or are developing cone beam solutions. The imaging systems are either integrated with the gantry, on a standalone robotic arm or integrated with the treatment couch. It is an area where there is still considerable scope for improvement in both hardware and software. One area that researchers have only recently begun to look at is dualenergy cone beam CT, which has the potential to improve tissue contrast and reduce beam hardening [10,11].
Imaging for Range Verification In X-ray radiotherapy, anatomical imaging in combination with dose modelling is the best method to determine the likely dose to the patient. The deposition of dose from an X-ray beam cannot be imaged directly. Proton therapy offers the intriguing prospect of being able to image the dose deposited in the patient by a more direct means. There are two methods that are widely studied to produce meaningful on-treatment imaging of dose deposition. PET can measure the gamma pair emitted by the annihilation of positrons produced by unstable isotopes created by inelastic interaction of the proton beam with patient tissues. A conventional PET scanner can measure the positron emission of a number of different isotopes with varying half-lives. The PET signal that is imaged is not directly a measurement of dose, so the PET image cannot be directly compared with the three-dimensional dose distribution predicated from treatment planning but can be compared with a Monte Carlo simulation predicted distribution of positron emitters [12]. The feasibility of using PET as an effective verification tool for proton therapy depends on a number of factors. When using a commercial PET scanner that may be some distance from the treatment room it will take time to remove the patient from the treatment couch, move them to the PET imager and set up for imaging in the treatment position. During this time the PET signal diminishes due to the half-life of the radioactive isotopes produced and biological washout of the PET activity. As the PET signal
Please cite this article in press as: MacKay RI, Image Guidance for Proton Therapy, Clinical Oncology (2018), https://doi.org/10.1016/ j.clon.2018.02.004
4
R.I. MacKay / Clinical Oncology xxx (2018) 1e6
diminishes the scanning time must be increased to collect a usable PET image. Different proton fields in the treatment delivery may be delivered minutes apart and so when imaged the activity will be dependent on the order of treatment delivery. As the time between imaging and treatment increases it becomes more difficult to predict the expected PET distribution using Monte Carlo simulation. These difficulties have led to attempts to perform PET scanning in the room [13]. Either through a standalone PET scanner in the treatment room or a PET scanning system integrated with the treatment gantry. Feasibility studies have shown that a standalone scanner in the treatment room can cut down the gap between treatment and imaging to 2.5 min and the scan time to 5 min, reduced from an estimated 30 min when the scanner is out of the room. Even with these timescales it was found that the modelling of the biological washout produced considerable uncertainty. In addition, imaging would have a significant impact on patient throughput in a busy clinical centre. To streamline this process further, ‘in beam’ PET has been developed, where PET detectors are positioned around the treatment beam so the patient can be imaged in the treatment position. Planar PET detectors have been integrated with a gantry to image the isocentre position [14]. In this system, imaging can be initiated immediately after the delivery of the treatment beam and acquisition times of 200 s were used. Rather than compare the planar image with a predicted PET distribution, the image from subsequent fractions was compared with the first fraction and time trends associated with changes in the tumour were observed. The other method being explored for range verification imaging is measurement of the prompt gamma emission produced by inelastic interactions of protons with the target nuclei. The gamma emission is a result of the excited nucleus returning to its ground state. Gamma emission from a prompt gamma interaction is more immediate than that from positron decay and is directly correlated with dose. There are a number of ways that have been used to measure the prompt gamma signal, including a single scintillator, gamma camera knife edge slit, multi slit detector and a Compton camera [15,16]. Currently, only one system has been developed commercially and tested clinically [17]. A prototype knife edge slit camera developed by Oncoray (Dresden, Germany) and IBA (Louvain, Belgium) was used to measure interactional range in passive scattered proton therapy with an accuracy of 2 mm. The same system was later used to verify spot scanning proton therapy, where range of the individual energy layers could be verified to an accuracy of 1.3 mm or better [18]. The prototype trolley-based system shows the feasibility of clinically meaningful prompt gamma measurement. Due to the inherent problem of range uncertainty in proton therapy and its potential clinical consequence, both the PET and prompt gamma methods of range verification have received considerable attention from researchers. However, there is still much work to do in developing an effective system to image range in clinical proton therapy.
The inherent difficulties around isotope decay and biological washout for PET verification have prevented progress in recent years. Prompt gamma has been developed to a commercial prototype and seems to have more clinical potential. For widespread clinical use there is still much hardware development required to produce an integrated easy to use measurement system. As a clinical tool, range imaging will provide valuable feedback that will reduce systematic uncertainties in treatment planning and flag deviations in patient set-up or anatomy that should trigger plan adaption.
Proton Computed Tomography and Proton Radiography Proton CT and proton radiography are both imaging techniques that attempt to use proton beams to image the patient. Both techniques rely on increasing the proton energy beyond what would usually be used to treat the tumour so that the proton beam can pass through the area of the patient being imaged. These techniques offer great promise to reduce uncertainties through the improved measurement of stopping power and have been in development for many years. However, currently of all the techniques discussed here, these techniques are perhaps the furthest from clinical application. In proton radiography, a transmission image can be formed by measuring the proton position at the entrance and exit of the patient and its residual energy after transmission through the patient [19]. In proton CT, similar technology can be used to take a series of images from different angles in an analogous manner to X-ray CT but the resulting volumetric image can be reconstructed from the measured energy loss of the protons as they pass through the patient. One of the obvious difficulties in developing proton imaging is the cost of the equipment needed to accelerate and detect the protons that form the images. However, in proton imaging for proton therapy the equipment to accelerate the protons is already a part of the treatment system. Despite this there are some important technological considerations. Typically, proton therapy requires maximum proton energy of around 250 MeV. This is high enough to target deepseated tumours but not always high enough for proton imaging where the protons need to pass through the complete patient with a residual energy that can be measured. Although smaller patients, particularly paediatrics, could be imaged with conventional clinical equipment, there has been much interest in increasing the maximum energy of accelerators [20,21] or boosting the energy specifically for imaging. Further to the accelerating technology, imaging requires proton tracking technology to measure the position and residual energy of the protons that are transmitted through the patient. A number of groups have developed proton imaging using position sensitive detectors and range telescopes [19]. In the UK, the Proton Radiotherapy Verification and Dosimetry Applications (PRaVDA) consortium has been at
Please cite this article in press as: MacKay RI, Image Guidance for Proton Therapy, Clinical Oncology (2018), https://doi.org/10.1016/ j.clon.2018.02.004
R.I. MacKay / Clinical Oncology xxx (2018) 1e6
the forefront of the development of a solid state system to perform proton CT. The prototype system has been used to image phantoms and the group has investigated proton CT based on many parameters, including proton stopping power, proton scattering power, proton attenuation power and proton straggling power. The integration of such a system with a clinical spot scanning proton therapy system is of considerable interest and hopefully will be the next phase of development. The first clinical application of proton CT may be to improve the characterisation of proton stopping powers derived from the CT scan used in pretreatment imaging. However, in the future, proton CT could be the primary image used in treatment planning and eventually could provide daily in-room 3D images of stopping power. Such images would be the ideal images for adaptive proton therapy.
Image Guidance in UK Centres In the UK, the National Health Service centres are being built with a range of image-guided equipment. For pretreatment planning, CT and magnetic resonance imaging have been integrated on the clinical treatment floor to image proton therapy patients. The centres will be able to perform dual-energy CT, but will probably implement the standard stoichiometric calibration before the first treatment. In the treatment room, two kilovoltage imaging systems are incorporated into the treatment gantry to allow two-dimensional planar and three-dimensional cone beam imaging in the treatment position. In-room cone beam CT will be used for patient set-up and to flag when adaption of the treatment plan is necessary. In many cases the adaption of the treatment plan will involve repeating the pretreatment CT imaging when any significant revision is required. The requirement for one in-treatment diagnostic CT scan has been incorporated into the treatment workflow. This is the starting point for a clinical service that should be in place for 20 years. Obviously it is hoped that many of the image guidance methods listed in this article will become a clinical reality over the lifetime of the centres. Both institutions have active research programmes in image guidance for radiotherapy.
Conclusion The need for image guidance in proton therapy is clear. It is perhaps surprising that with the number of potential developments to improve proton therapy treatment and delivery that clinical application seems limited, particularly with the attention given to image guidance for high-energy X-rays. As the number of proton centres grows so will the demand for improved image guidance. Researchers in more institutions will turn their attention to image guidance and manufacturers will have improved resources to translate
5
systems into their clinical products. Such developments can only help to realise the clinical proton advantage offered by the Bragg peak, ensuring that improved dose distributions are planned and delivered.
References [1] National Cancer Action Team National Radiotherapy Implementation Group Report. Image guided radiotherapy (IGRT) guidance for implementation and use 2012. [2] Schneider U, Pedroni E, Lomax A. The calibration of CT Hounsfield units for radiotherapy treatment planning. Phys Med Biol 1996;41(1):111e124. €r E, Lalonde A, Royle G, Lu HM, Bouchard H. The potential of [3] Ba dual-energy CT to reduce proton beam range uncertainties. Med Phys 2017;44(6):2332e2344. [4] Schulte RW. Strategies for image-guided proton therapy of cancer. Eur Oncol Dis 2007;1(2):116e118. [5] Shirato H, Shimizu S, Kitamura K, Nishioka T, Kagei K, Hashimoto S, et al. Four-dimensional treatment planning and fluoroscopic real-time tumor tracking radiotherapy for moving tumor. Int J Radiat Oncol Biol Phys 2000;48:435e442. [6] Devicienti S, Strigari L, D’Andrea M, Benassi M, Dimiccoli V, Portaluri M. Patient positioning in the proton radiotherapy era. J Exp Clin Cancer Res 2010;29(1):47. [7] Schulze R, Heil U, Gross D, Bruellman DD, Dranischnikow E, Schwanecke U, et al. Artefacts in CBCT: a review. Dentomaxillofac Radiol 2011;40(5):265e273. https://doi.org/10.1259/ dmfr/30642039. [8] Park YK, Sharp GC, Phillips J, Winey BA. Proton dose calculation on scatter-corrected CBCT image: feasibility study for adaptive proton therapy. Med Phys 2015;42(8):4449e4459. [9] Veiga C, Janssens G, Teng C-L, Baudier T, Hotoiu L, McClelland JR, et al. First clinical investigation of cone beam computed tomography and deformable registration for adaptive proton therapy for lung cancer. Int J Radiat Oncol Biol Phys 2016;95(1):549e559. [10] Peng H, Kanehira T, Takao S, Matsuura T, Umegaki K, Shirato H, et al. Feasibility study of using dual-energy cone beam computed tomography (DE-CBCT) in proton therapy treatment planning. Int J Radiat Oncol Biol Phys 2016;96:E697. [11] Kovacs DG, Rechner L, Berthelsen AK, Aznar MC. Dual energy cone beam computed tomography for image guided radiation therapy. Int J Radiat Oncol Biol Phys 2016;96:E689. ~ a S, Paganetti H. Monte Carlo patient [12] Moteabbed M, Espan study on the comparison of prompt gamma and PET imaging for range verification in proton therapy. Phys Med Biol 2011; 56(4):1063. ~ a S, Daartz J, Liebsch N, Ouyang J, Paganetti H, [13] Zhu X, Espan et al. Monitoring proton radiation therapy with in-room PET imaging. Phys Med Biol 2011 Jul 7;56(13):4041e4057. [14] Nishio T, Miyatake A, Ogino T, Nakagawa K, Saijo N, Esumi H. The development and clinical use of a beam ON-LINE PET system mounted on a rotating gantry port in proton therapy. Int J Radiat Oncol Biol Phys 2010;76(1):277e286. https://doi. org/10.1016/j.ijrobp.2009.05.065. [15] Min C. Prompt gamma measurements for locating the dose falloff region in the proton therapy. Appl Phys Lett 2006;89: 183517. [16] Smeets J, Roellinghoff F, Prieels D, Stichelbaut F, Benilov A, Busca P, et al. Prompt gamma imaging with a slit camera for real-time range control in proton therapy. Phys Med Biol 2012; 57(11):3371e3405. https://doi.org/10.1088/0031-9155/57/11/ 3371.
Please cite this article in press as: MacKay RI, Image Guidance for Proton Therapy, Clinical Oncology (2018), https://doi.org/10.1016/ j.clon.2018.02.004
6
R.I. MacKay / Clinical Oncology xxx (2018) 1e6
[17] Richter C, Pausch G, Barczyk S, Priegnitz M, Keitz I, Thiele J, et al. First clinical application of a prompt gamma based in vivo proton range verification system. Radiother Oncol 2016;118(2):232e237. https://doi.org/10.1016/j.radonc.2016. 01.004. [18] Xie Y, Bentefour EH, Janssens G, Smeets J, Vander Stappen F, Hotoiu L, et al. Prompt gamma imaging for in vivo range verification of pencil beam scanning proton therapy. Int J Radiat Oncol Biol Phys 2017;99(1):210e218. https://doi.org/10. 1016/j.ijrobp.2017.04.027.
[19] Poludniowski G, Allinson NM, Evans PM. Proton radiography and tomography with application to proton therapy. Br J Radiol 2015;88(1053):20150134. https://doi.org/10.1259/bjr. 20150134. [20] Apsimon R, Burt G, Pitman S, Owen H. ProBE e proton boosting extension for imaging and therapy 2016. [21] Al Harbi N, Lee SY. Design of a compact synchrotron for medical applications. Rev Sci Instr 2003;74:2540e2545. https://doi.org/10.1063/1.1561598.
Please cite this article in press as: MacKay RI, Image Guidance for Proton Therapy, Clinical Oncology (2018), https://doi.org/10.1016/ j.clon.2018.02.004