Streamlining the image-guided radiotherapy process for proton beam therapy

Abstract

Objectives:

This work evaluated the on-treatment imaging workflow in the UK’s first proton beam therapy (PBT) centre, with a view to reducing times and unnecessary imaging doses to patients.

Methods:

Imaging dose and timing data from the first 20 patients (70% paediatrics, 30% TYA/adult) treated with PBT using the initial image-guided PBT (IGPBT) workflow of a 2-dimensional kilo-voltage (2DkV), followed by cone-beam computed-tomography (CBCT) and repeat 2DkV was included. Pearson correlations and Bland-Altman analysis were used to describe correlations between 2DkV and CBCT images to determine if any images were superfluous.

Results:

229 treatment sessions were evaluated. Patient repositioning following the initial 2DkV (i2DkV) was required on 19 (8.3%) fractions. This three-step process resulted in an additional mean imaging dose of 3.4 mGy per patient, and 5.1 minutes on the treatment bed for the patient, over a whole course of PBT, compared to a two-step workflow (removing the i2DkV image). Correspondence between the mean displacements from i2DkV and CBCT was high, with R = 0.94, 0.94 and 0.80 in the anteroposterior, superiorinferior and right-left directions, respectively. Bland-Altman analysis showed very little bias and narrow limits of agreement.

Conclusions:

Removing the i2DkV, streamlining to a two-step workflow, would reduce treatment times and imaging dose, and has been implemented as standard verification protocol. For challenging cases (e.g. paediatric patients under GA), further investigations are required before the three-step workflow can be modified.

Advances in knowledge:

This is the first report assessing a preliminary imaging protocol in PBT in the UK and determining a way to reduce dose and time, which ultimately benefits the patient.

Introduction

Image guidance is an integral component of modern radiotherapy to ensure the precision and accuracy of treatment delivery.^1,2 Due to the dosimetric properties of protons (steep depth dose curve and finite range at the distal edge of the Bragg peak resulting in sensitivity to day-to-day anatomical changes), image guidance in proton beam therapy (PBT) is arguably even more critical than in photon therapy to ensure accurate treatment delivery and avoid geometric miss.³

Until recently, the standard image-guided radiotherapy (IGRT) procedure for verifying patient alignment in PBT has consisted of 2-dimensional orthogonal co-planar kilo voltage (2DkV) X-rays acquired daily in the treatment room.^3,4 This approach does not allow for the most accurate validation of the proton beam range.⁵ Although several institutions have used in-room CT for some time, recent advances of in-room volumetric image guidance using cone-beam computed tomography (CBCT) for the delineation of soft tissue target volumes and identifying anatomical changes over a course of fractionated PBT are becoming increasingly available.^5–8 However, the widespread adoption of CBCT as an established verification technique for PBT lags behind photon radiotherapy.^1,4 Moreover, strategies for image-guided proton beam therapy (IGPBT) can vary substantially between proton centres, with many institutions developing their own procedures that are closely linked to their technical implementation and delivery processes.³

High-energy PBT is novel in the UK, with the reporting institution being the first high-energy NHS PBT service. As such, there is no established national image verification guidance for proton therapy. This work describes the IGPBT strategy initially implemented at a single institution, and the evaluation undertaken to propose a reduction in imaging time and dose to patients. The objectives of this evaluation are to quantify the imaging workload and positioning reproducibility errors by evaluating imaging dose, timing, and comparison of set-up error between two imaging modalities, 2DkV and CBCT, respectively.

Methods and materials

The reporting institution’s PBT facility consists of three ProBeam gantries (Varian Medical Systems UK Ltd). All gantries use pencil beam scanning (PBS) and are equipped with an on-board imaging system consisting of two orthogonal detector panels (attached to the treatment nozzle) and two X-ray sources (embedded in the treatment floor). This equipment allows for the acquisition of 2DkV images as well as volumetric image guidance (CBCT). To implement image guidance at the opening of the PBT service, a cautious, multi-image modality, three-step IGPBT process was developed with an emphasis on CBCT. Efforts were made to limit imaging dose as low as reasonably practicable (ALARP) and the image guidance protocol initially implemented drew from local photon-based IGRT procedures, which follow UK guidance of imaging on the first three fractions, then weekly unless otherwise indicated.⁹

This work evaluated the time taken and image dose to the patients using this cautious imaging approach to determine if reductions in time and dose could be achieved. Approval for this work was granted by the local Quality Improvement and Clinical Audit committee (SE19/2487).

Patient selection

The first 20 patients treated consecutively with PBT at the authors’ institution were included in the evaluation. Treatment sites for the 20 patients included targets in the brain (n = 9), pelvis (n = 4), thoracic/lumbar spine (n = 3), cervical spine (n = 2), head and neck (n = 1), and thorax (n = 1).

Patient positioning, immobilisation and planning scans

All patients were positioned with individualised immobilisation equipment that was dependant on treatment site (Table 1). Patients undergoing PBT to the brain, head and neck, cervical spine and thorax were positioned supine on a base of skull (BoS) board with a custom-made neck rest. Those patients undergoing brain treatment were also fitted with a custom-made thermoplastic BoS mask (Figure 1) and patients undergoing head and neck, cervical spine and thorax PBT were immobilised with a thermoplastic BoS Head and Shoulders mask. Paediatric thermoplastic masks were used for paediatric patients. Patients undergoing PBT to the mid-lower spine or pelvis were immobilised using a personalised vacuum moulded bag in either a supine or prone position with a knee fix or ankle rest to aid comfort and reproducibility.

Table 1.

Patient positioning and immobilisation according to tumour site treated

Treatment site	Positioning	Immobilisation
Brain	Supine	Thermoplastic BoS mask, custom-made neck rest
Head and neck	Supine	Thermoplastic BoS Head and Shoulders mask, custom-made neck rest
Cervical spine	Supine	Thermoplastic BoS Head and Shoulders mask, custom-made neck rest
Thoracic/Lumbar/Sacral spine	Prone	Vacuum moulded bag, knee fix, ankle rest
Pelvis/Abdomen	Supine	Vacuum moulded bag, knee fix, ankle rest
Thorax	Supine	Thermoplastic BoS Head and Shoulders mask, custom-made neck rest

BoS, Base of skull.

Figure 1.

Photo of treatment positioning in thermoplastic BoS mask and custom-made neck cushion.

Set-up points were tattooed on the patients’ skin providing anatomical reference points for daily alignment to a three-point external laser system for extracranial treatments. Each patient underwent a treatment planning CT (pCT) in the PBT department (1-mm slice thickness for sites in the brain; 2-mm slice thickness for all other treatment sites; SOMATOM Confidence syngo CT VA62A, Siemens, Germany).

Planning MR sim

All patients underwent a planning Magnetic Resonance (pMR) scan in the treatment position with the MR safe immobilisation on the 70 cm diameter bore MR-simulator scanner (1–1.5 mm slice thickness for brain and head and neck; 3–4 mm slice thickness all other treatment sites; Philips Ingenia MR-RT 1.5T, Koninklijke Philips N.V., Netherlands). MR images were imported into the treatment planning system (TPS) and co-registered to aid target and organ-at-risk (OAR) delineation.

On-treatment image verification

Patient treatment set-up and positioning was verified by daily imaging immediately prior to PBT treatment delivery. CBCT imaging frequency was determined by treatment site and according to the authors’ initial institutional guidelines, which are in keeping with national image guidance recommendations.^10,11 For treatment sites in the head and neck, daily CBCT was acquired for the first five fractions followed by alternate fractions. For all other treatment sites, daily CBCTs were acquired for the first five fractions followed by weekly CBCT, providing there were no soft tissue or clinical changes impacting upon target coverage. On non-CBCT days, all treatment sites were imaged with 2DkV imaging only to eliminate gross set-up errors.

3D-imaging protocol

For treatment sessions including CBCTs, the three-step IGPBT protocol below was followed using the imaging tolerances detailed in Table 2:

1. An initial 2DkV image pair acquisition (typically an anterior and a lateral image) for gross positioning assessment, registered to the digitally reconstructed radiograph (DRR) from the pCT and auto-matched to bony anatomy.

2. A subsequent 3D CBCT to assess target volumes, OARs, and soft tissue contour changes within the proton beam pathway, using the pCT with volumes as reference to aid the match. Implementation of a correction strategy using a 6 degree of freedom couch was carried out, if the given tolerances outlined in Table 2 were exceeded.

3. A repeat verification 2DkV image pair registered and auto-matched to the DRR was acquired, to confirm all translational and rotational corrections, before delivering proton beam treatment.

Table 2.

Imaging tolerances by imaging modality (2DkV portal image or cone-beam CT)

Image modality:	Initial 2DkV image pair	CBCT	Verification 2DkV image pair
Translations: (anteroposterior, caudiocranial, mediolateral)	≤2.00 cm	≤0.10 cm	≤0.20 cm
Rotations: [pitch (x), roll (y), rotation (z)]	≤2.0o	≤0.2o	≤1.0o
Other:		Soft tissue changes, including contour

If the set-up error identified on the 3D CBCT image was within the tolerance defined in Table 2, there would be no requirement for a couch correction and therefore no requirement to acquire a repeat verification 2DkV image. If the initial 2DkV exceeded the online translational or rotational matched tolerances indicated in Table 2, patients were repositioned and 2DkV images were acquired until acceptable, before proceeding to the CBCT. The repeat verification 2DkV image pair is a requirement from the vendor to confirm accurate movement of the robotic couch after the correction shifts are applied.

Data collection

An institutional web-based patient record system, clinical web portal (CWP), was used to obtain patient characteristics. A PBT Scan Record created at the pCT stage was used to record patient treatment positioning and immobilisation. Imaging data was retrospectively collated from the ARIA Oncology Information System (version 13.7, Varian Medical Systems, USA) in the Offline Review workspace for all treatment sessions that included the three-step CBCT image-guided workflow.

The following three data items were collected and tabulated for each treatment fraction for every patient:

1. initial 2DkV imaging dose, using estimated delivered dose (mGy).

2. the elapsed time between acquiring and reviewing the initial 2DkV images and commencement of the subsequent CBCT acquisition (minutes).

3. the comparison of the magnitude of online matched values (set-up error) for the initial 2DkV and CBCT image registration in the three translational (anteroposterior, caudiocranial, mediolateral) axes and the three rotational (pitch, roll, rotation) axes.

The 2DkV imaging doses were derived from the average measured doses for the pre-set values at a source angle of 0°, anteroposterior (A/P) direction. The 2DkV image doses for the mediolateral settings were calculated from relationships determined at commissioning and listed in the institutional local documentation, ProBeam Imaging Doses and Presets.

Data analysis

Descriptive statistics were used to analyse the data and all calculations were completed using Microsoft Excel 2010 (Microsoft, Redmond, WA, USA).

The imaging dose from the initial 2DkV was accumulated over the whole course of each patient’s treatment schedule to indicate the estimated additional absorbed dose received on a three-step IGPBT workflow over a two-step process.

The elapsed time between acquiring the initial 2DkV and commencement of the subsequent CBCT acquisition was totalled over the whole treatment course per patient. A mean time was then calculated over the average of the whole course of treatment per patient and for all patients collectively.

The translational and rotational set-up uncertainty from the online matched values from 2DkV and CBCT image registration was used to calculate a Pearson’s correlation coefficient to measure the linear correlation between the two imaging modalities. This was calculated for each patient for the treatment course and across the 20 patients.

To assess agreement between the initial 2DkV and CBCT online matched values, Bland-Altman analysis was calculated. The 95% limits of agreement were calculated by using the mean and standard deviation of the differences between the two measurements.

Results

Twenty consecutive patients treated with a full course of PBT between 17 December 2018 and 24 May 2019 were evaluated; median age was 13 years (range: 2–80). Fourteen patients were paediatrics (15 years and under), four patients were within the teenage and young adult (TYA) age range (16–25 years) and two were adult patients. Seven paediatric patients underwent treatment under general anaesthesia (GA). The treatment sites evaluated were brain (n = 9), pelvis (n = 4), thoracic/lumber spine (n = 3), cervical spine (n = 2), head and neck (n = 1), and thorax (n = 1). Patient characteristics and their treatment subsites are recorded in Table 3.

Table 3.

Patient characteristics, number of fractions requiring repositioning after initial 2DkV and before CBCT

Patient	Patient age (years)	GA (Y/N)	Diagnosis	Treatment site	No. of fractions patient repositioning required
1	14	N	Pituitary	Brain	0
2	12	N	Ependymoma	Thoracic/lumbar spine	0
3	51	N	Chordoma	Lumbar/sacral spine	0
4	15	N	Low grade glioma	Brain	0
5	6	Y	Ependymoma	Brain	2
6	7	Y	Craniopharyngioma	Brain	2
7	6	Y	Rhabdomyosarcoma	Pelvis	6
8	16	N	Astrocytoma	Brain	0
9	2	Y	Ewing’s sarcoma	Cervical spine	0
10	15	N	Ewing’s sarcoma	Pelvis	2
11	80	N	Chordoma	Lumbar/sacral spine	0
12	5	Y	Ependymoma	Brain	0
13	17	N	Hard palate	Head and neck	0
14	3	Y	Rhabdomyosarcoma	Pelvis	2
15	20	N	Low grade glioma	Brain	0
16	9	N	Glioma	Cervical/thoracic spine	3
17	2	Y	Rhabdomyosarcoma	Pelvis	1
18	15	N	Ewing’s Sarcoma	Right scapula	0
19	11	N	Craniopharyngioma	Brain	1
20	18	N	Low grade glioma	Brain	0
	Median age: 13 (range: 2–80)	GA (n = 7)			19 (/229)

Patient repositioning and additional imaging

A total of 229 treatment sessions were evaluated (per patient: median 10, range 8–19). Nineteen fractions (8.3%) required repositioning following assessment of the initial 2DkV image (i.e., before CBCT acquisition). Of the 19 repositioned treatment sessions, all patients were paediatrics (0–15 years). Thirteen of the 19 fractions (68.4%) were paediatric patients (n = 5) that underwent treatment under GA (Table 3). Reasons for paediatric patient repositioning were largely (71.4%) due to patient set-up exceeding the image tolerances outlined in Table 2 for the gross error assessment, specifically in the rotational planes: pitch (X), roll (Y) and rotation (Z) axes, identified on the initial 2DkV image pairs. For adult patients, and patients within the TYA age range, no treatment sessions required modification to patient positioning after the initial 2DkV.

Eight paediatric patients required more than the prescribed initial 2DkV images. The total number of additional 2DkV image pairs acquired due to repositioning was 21 across all patients (per patient: range 1–7, median 2, mean 2.6). The treatment sites for these were brain (n = 3), pelvis (n = 4) and cervical spine (n = 1) (Table 3).

Dose and time

The initial 2DkV, used in the three-step IGPBT process, resulted in an additional mean imaging dose of 3.4mGy (range 1.7–11.2) per patient over a whole course of PBT. The use of the three-step imaging process required a mean additional time of 5.1 minutes (range 3.3–9.9) on the bed for the patient, compared to the two-step workflow.

Correlation between 2D and 3D imaging

Overall, correspondence between the mean translational displacements from the initial 2DkV and CBCT images for all treatment sites was high, with R = 0.94, 0.94 and 0.80 in the anteroposterior, superiorinferior and right-left directions, respectively. Similarly, the correlation between the mean rotational online matched discrepancies was strong, although not as strongly correlated as the translational direction. Details of the 2D and 3D correlations are noted in Table 4. These results demonstrate that set-up errors identified on both the initial 2DkV and CBCT were similar when rotations are not involved. Patients receiving PBT to the mid-lower spine or pelvis showed the greatest consistency in both translational and rotational correlation, whereas concordance was more variable in patients undergoing intercranial PBT in the pitch and roll directions specifically.

Table 4.

Pearson’s Correlation Coefficient of the Mean Translational and Rotational Displacements

	Translational			Rotational
Pearson’s Correlation Coefficient	Anteroposterior (Vertical)	Caudiocranial (Longitudinal)	Mediolateral (Lateral)	Pitch (X)	Roll (Y)	Rotation (Z)
	0.94	0.94	0.80	0.73	0.85	0.73

Figure 2 displays the Bland-Altman plots for the online matched values for the initial 2DkV and CBCT in all translation and rotational directions. The translational Bland-Altman results indicate very little bias and narrow limits of agreement (±1 mm), demonstrating a strong comparison between the set-up error for the initial 2DkV and CBCT (Table 5). Bland-Altman analysis for the online matched values in the pitch, roll and rotation direction show wider levels of agreement, (±1.9°,±1.7° and ±1.2°), indicating there was a higher variability in online matched values between 2D and 3D imaging (Table 5).

Figure 2.

Bland-Altman plots of CBCT and 2DkV portal imaging translations measured from planning CT in the vertical (top), horizontal (middle) and lateral (bottom) directions on the left. On the right rotations from the planning CT as reported on CBCT and 2DkV portal imaging are shown for the pitch (top), roll (middle) and yaw (bottom). Each patient has a different colour and the number of their values at a particular value of the graph is represented by the size of the circle.

Table 5.

Bland-Altman Analysis Summary

Translation Displacements	95% limits of agreement:
Anteroposterior (Vertical)	0.13 cm (95% CI: −0.15–−0.11 cm) to 0.13 cm (95% CI: 0.11–0.15 cm) 1.3to 1.3 mm
Caudiocranial (Longitudinal)	0.14 cm (95% CI: −0.17–−0.11 cm) to 0.15 cm (95% CI: 0.12–0.19 cm) 1.4to 1.5 mm
Mediolateral (Lateral)	0.13 cm (95% CI: −0.16–−0.09 cm) to 0.14 cm (95% CI: 0.11–0.17 cm) 1.3to 1.4 mm
Rotation Displacements	95% limits of agreement:
Pitch (X)	−0.97° (95% CI: −1.20–−0.74°) to 0.87° (95% CI: 0.64–1.10°) ±1.9°
Roll (Y)	−0.75° (95% CI: −0.91–−0.60°) to 0.87° (95% CI: 0.71–1.02°) ±1.7°
Rotation (Z)	−0.60° (95% CI: −0.72–−0.49°) to 0.64° (95% CI: 0.52–0.75°) ±1.2°

Discussion

The implementation of PBT has dramatically increased in the last 5 years. Delivery of high-energy PBT is novel in the UK, involving major investment and creating high expectations.¹² Because of its novelty in the UK, protocols and procedures for implementing still have room for fine-tuning. Although the use of CBCT is not a novel technique in radiotherapy, for PBT the wide-set adoption of CBCT is yet to be exploited.¹³ The reporting institution sought to quantify the benefits of an IGRT workflow using a mix of 2D images and CBCT. This work summarises the results of an evaluation of the initial IGPBT process implemented into clinical workflow at the first NHS PBT Centre in the UK.

The three-step IGPBT workflow was a cautious approach providing opportunities to not only verify patient positioning but to verify unfamiliar proton equipment, software and patient immobilisation confirming the reliability and reproducibility for treatment. The workflow strove to minimise the number of repeat CBCTs acquired in a (largely) paediatric population as the assessment of initial 2DkV images permitted a point in the image guidance pathway to allow for patient repositioning (following gross error assessment). This use of initial 2DkV imaging would provide only a relatively low dose delivered to the patient in comparison to the effective doses associated with CBCT^10,14 minimising the possibilities of patient positioning requiring a change following CBCT. However, it is recognised that there are several limitations of 2D imaging. Firstly, orthogonal co-planar 2D imaging does not provide sufficient anatomical information and therefore changes to the soft tissue and target volume would not be identified.¹⁵ This is perhaps of greater concern for treatment sites that may be more susceptible to daily anatomical variations, for example sites within the head and neck. It is acknowledged that there was only one head and neck case evaluated within the scope of this study, with the majority of treatment sites in the brain where soft tissue changes affecting the target volume can be marginal. Secondly, in contrast to 3D imaging, 2D imaging does not quantify both the translational and rotational displacements.¹⁶ It was anticipated that wider levels of agreement would exist for the rotational displacements due to the lack of volumetric data on co-planar imaging.

Imaging dose can be a concern in this vulnerable patient population, contributing to an already high treatment dose burden. It was therefore planned that once sufficient experience had been gained, a careful evaluation of the image guidance pathway would be carried out. This was to ensure concomitant exposure is kept ALARP with the possibility of eliminating the initial 2DkV imaging. In addition to adhering to stringent radiation protection safety regulations, evaluation of the IGPBT workflow was also necessary to ensure overall treatment time on the bed for the patient could be kept as short as possible. This evaluation was planned after the first 20 patients, meeting the suggested UK guidelines of sample population.⁹ By providing the first report assessing a preliminary imaging protocol for PBT in the UK, the authors hope to provide information for the standardisation of PBT image guidance protocols by determining a way to reduce dose and time to benefit the patient and improve service efficiency.

This report demonstrates a minority (8.3%) of all the 229 treatment sessions evaluated necessitated patient repositioning following the initial 2DkV image pair. This indicates that in 91.7% of treatment sessions, these images were superfluous and could be eliminated to minimise radiation exposure. Although imaging doses resulting from 2DkV acquisition are low, this additional dose contributes to increasing the total radiation burden.¹⁷ This is of particular concern in the paediatric population, whereby potential late-radiation effects and its prevalence among childhood cancer survivors is critical.¹⁸ Furthermore, the increased time to carry out the IGPBT workflow may reduce the patient’s tolerance for maintaining the treatment position.¹⁹

PBT delivery often takes longer to deliver than photon radiotherapy due to the complex multiple layers of the target field which require scanning beams to use magnets to move the proton beam precisely, to “paint” the area to be treated.²⁰ Shortening the IGPBT workflow from a three-step process to a two-step by removing the initial 2DkV and commencing image guidance with CBCT would reduce the overall treatment time for the patient on the bed. Indeed, the results of this study indicate a mean reduction of up to almost 10 minutes for one particular patient evaluated. Furthermore, the two-step workflow would facilitate the effort to adhere to the ALARP principle, by reducing mean additional imaging doses of up to 3.4 mGy per patient on average.

All of the treatment sessions that required a change to patient positioning were paediatric patients (range 2–15 years) with the majority (68.4%) having their treatment under GA. No treatment sessions for patients within the TYA and adult age group required a change to patient set-up after the initial 2DkV. Discrepancies in the rotational set-up error exceeding the defined image tolerances were notably prevalent for paediatrics under GA immobilised in a thermoplastic base of skull (BoS) mask where there were challenges in achieving the correct head position in the rotational planes. The inconsistencies in correct head alignment may be associated with the daily variability of airway positioning, resulting in differing degrees of flexion and extension of the patient’s neck and evident in the discrepancies of the pitch (X) direction. Similarly, all three paediatric patients under GA having treatment to the pelvis and immobilised in a personalised vacuum moulded bag, presented with some variation in the rotation of the pelvis. The differences in pelvic alignment could be attributed to the daily variation of positioning an anaesthetised paediatric patient in the vacuum moulded bag as a result of manual handling. Despite these considerations for repositioning the paediatric patient undergoing PBT under GA, the overall number of treatment sessions that required a modification to patient set-up is still minor. Nevertheless, in light of the small sample of patients under GA evaluated, and with consideration that the PBT clinical service was still in its infancy at the time of carrying out the project, there remains some trepidation when altering the imaging pathway for such patients. It was, therefore, deemed necessary to undertake additional investigations on a specific cohort of patients treated under GA before modifying the IGPBT process when the PBT service is not as novel.

A minority of repeat initial 2DkV images that did not exceed the set-up tolerance, but still required patient repositioning were due to suboptimal patient positioning unsuitable to proceed with treatment. This was predominantly associated with patients having PBT to the pelvis and immobilised supine in a vacuum bag, where the femur was displaced laterally and inconsistent with the planning CT scan. Where required, patient repositioning allowed the radiographers to correct the head or pelvis position and verify treatment position with another orthogonal image pair, before proceeding to CBCT. It is worth noting that none of the online matched values for the initial 2DkV images exceeded the translational and rotational thresholds of 2 cm and 2 degrees, respectively.

Pearson’s Correlation Coefficient demonstrated strong concordance of the set-up error identified on the initial 2DkV and CBCT images for the treatment sites evaluated. Bland-Altman analysis of the translational online matched values further corroborates a strong translational agreement between 2DkV and CBCT imaging. Although some caution was exercised when drawing conclusions about the correlation between 2D and 3D imaging, the results show that for most cases, on the three-step image workflow, the initial 2DkV remains redundant and can be removed from the image guidance pathway, streamlining the IGPBT workflow as the CBCT would still identify daily patient set-up variations, target volume delineation and soft tissue changes. The results of this evaluation have also led to further work investigating the possibility of increasing the frequency of CBCTs while optimising the acquisition parameters to keep the dose as low as possible. Moreover, further investigations are ongoing to further examine the agreement of 2D and 3D imaging in specific tumour sites and in comparison to photon practices.

A limitation of this work is that it was carried out when the PBT service was newly opened. Although this may have had an impact on the overall treatment times due to new processes, software and treatment techniques, image analysis is not a novel concept in radiotherapy and the radiographers were well-versed in image-matching and analysis. Therefore, the ‘newness’ of the service should not have affected the image analysis time. The study sample was small and appears to show an overrepresentation of treatment sites in the brain and an underrepresentation of other treatment sites. For example, only one head and neck case was evaluated. A lack of familiarity and experience the radiographers had with the GA equipment in the PBT treatment room when the study was carried out may account for some of the inconsistencies in the positioning of paediatric patients under GA. Despite these limitations, the evidence was convincing enough to modify the initial practice and prompt on-going evaluations of IGPBT practice evolution including comparisons to local photon-based IGRT in similar treatment sites.

Conclusion

The results of this evaluation have provided a quantifiable evidence-base supporting a change in clinical workflow and streamlining the image guidance process for PBT requiring less time to deliver and less imaging dose to the patient. A two-step IGPBT workflow has now been implemented at the authors’ institution and is standard verification protocol. For challenging cases (e.g. paediatric patients under GA), further investigations are required before the three-step workflow can be modified.