Co-registration of cone beam CT and planning CT in head and neck IMRT dose estimation: a feasible adaptive radiotherapy strategy

Abstract

Objective:

To investigate if cone beam CT (CBCT) can be used to estimate the delivered dose in head and neck intensity-modulated radiotherapy (IMRT).

Methods:

15 patients (10 without replan and 5 with replan) were identified retrospectively. Weekly CBCT was co-registered with original planning CT. Original high-dose clinical target volume (CTV1), low-dose CTV (CTV2), brainstem, spinal cord, parotids and external body contours were copied to each CBCT and modified to account for anatomical changes. Corresponding planning target volumes (PTVs) and planning organ-at-risk volumes were created. The original plan was applied and calculated using modified per-treatment volumes on the original CT. Percentage volumetric, cumulative (planned dose delivered prior to CBCT + adaptive dose delivered after CBCT) and actual delivered (summation of weekly adaptive doses) dosimetric differences between each per-treatment and original plan were calculated.

Results:

There was greater volumetric change in the parotids with an average weekly difference of between −4.1% and −27.0% compared with the CTVs/PTVs (−1.8% to −5.0%). The average weekly cumulative dosimetric differences were as follows: CTV/PTV (range, −3.0% to 2.2%), ipsilateral parotid volume receiving ≥26 Gy (V26) (range, 0.5–3.2%) and contralateral V26 (range, 1.9–6.3%). In patients who required replan, the average volumetric reductions were greater: CTV1 (−2.5%), CTV2 (−6.9%), PTV1 (−4.7%), PTV2 (−11.5%), ipsilateral (−10.4%) and contralateral parotids (−12.1%), but did not result in significant dosimetric changes.

Conclusion:

The dosimetric changes during head and neck simultaneous integrated boost IMRT do not necessitate adaptive radiotherapy in most patients.

Advances in knowledge:

Our study shows that CBCT could be used for dose estimation during head and neck IMRT.

Radiotherapy alone or chemoradiation is used in the primary treatment of head and neck squamous cell carcinomas (HNSCCs). Intensity-modulated radiotherapy (IMRT) allows dose conformality and organ-at-risk (OAR) avoidance in the complex head and neck anatomical region and is therefore the standard of care.¹ IMRT has been proven to reduce the risk of late xerostomia with no detrimental effect on local control or survival.² Changes in gross target volume (GTV) due to treatment response or patient contour secondary to weight loss can lead to dosimetric inaccuracies.^3,4 Because of the greater conformality of IMRT, these changes can have more severe dosimetric impact.¹ This is pertinent to patients receiving curative radiotherapy for HNSCC as delivery is protracted over 6–7 weeks but is based on a single anatomical snapshot acquired during the planning CT scan. Studies have shown that GTVs shrink by 1.8–3.9% per day and parotids shrink by 0.6–1.1% per day.^3,5,6 This coupled with weight loss during treatment could result in a loose immobilization device, thus increasing the set-up uncertainty.^1,3 These progressive anatomical and volumetric changes may have dosimetric consequences and subsequent clinical impact despite the theoretical advantages of IMRT. Clinical significance is determined not only by the level of dosimetric difference, typically defined as ±5%, but also the impact that these may have on the respective target volumes and OARs.

Adaptive radiotherapy (ART) has been advocated to overcome some of the problems associated with IMRT in head and neck cancers by modifying treatment plans in response to patient-specific changes during radiotherapy.⁷ This strategy involves modification of the original radiotherapy treatment volumes and plans in response to progressive tumour or normal tissue changes.^7,8 Although ART can be performed online or offline, the latter strategy is probably more applicable in the busy clinical setting. ART requires the availability of image-guided radiotherapy using on-board imaging such as kilovoltage (kV) or megavoltage (MV) cone beam CT (CBCT), or in-room kV CT to detect tumour or normal tissue changes. However, not all patients derive a clinically significant benefit from ART.^9,10 In addition, this strategy is labour and resource intensive, which could affect its widespread clinical implementation. As such, attempts should be made to identify the subset of patients who will benefit from ART as this could potentially improve disease control and reduce toxicities, without imposing a great stress on current radiotherapy resources.¹¹

Image-guided IMRT using kV CBCT is routinely used in our centre for HNSCC patients receiving curative radiotherapy. We hypothesize that the co-registration of regular CBCT and planning CT scans could be used to estimate delivered doses in these patients. This may allow identification of a cohort of patients with significant volumetric and, more importantly, dosimetric changes who may benefit from ART.

METHODS AND MATERIALS

Patients

Following institutional review approval, 15 consecutive head and neck cancer patients who required bilateral neck irradiation and who were treated with curative intent using simultaneous integrated boost IMRT between March 2012 and September 2012 were included in this retrospective exploratory study to assess the feasibility of this dose estimation method. 10 patients did not require a treatment replan, whereas 5 required one. Treatment replans were initiated because of poor set-up (n = 1), weight loss (n = 3) and a tight-fitting immobilization shell owing to facial swelling (n = 1). Patient characteristics are shown in Table 1.

Table 1.

Patient and treatment characteristics

Patient characteristics	All patients (N = 15), n (%)	No replan (N = 10), n	Replan (N = 5), n
Age (years)
Mean (range)	51 (36–72)	50 (36–60)	54 (42–72)
Gender
Male	11 (73)	7	4
Female	4 (27)	3	1
Primary tumour sites
Tonsil	6 (40)	4	2
Base of tongue	5 (34)	3	2
Hypopharynx	2 (13)	2	0
Nasopharynx	2 (13)	1	1
Tumour stage
I	0	0	0
II	1 (7)	0	1
III	1 (7)	0	1
IV	13 (86)	9	4
Concurrent chemotherapy
Yes	14 (93)	9	5
No	1 (7)	1	0

Patients received neoadjuvant cisplatin/carboplatin (n = 1) and 5-fluorouracil chemotherapy.

Radiotherapy

All patients were immobilized using a thermoplastic shell and shoulder depressors (Figure 1). Planning CT scans were obtained at 2-mm slice thickness following intravenous contrast injection. Gross, clinical and planning target volumes (GTVs, CTVs and PTVs) were created according to institutional protocol. Primary and nodal GTVs were delineated using diagnostic imaging, clinical examination and endoscopy information. CTVs included the GTV and areas of high risk to be treated to a curative dose (CTV1) and the node-negative neck to be treated to a prophylactic dose (CTV2). A uniform 4-mm expansion was applied to create the PTV, limited to no closer than 5 mm to the skin surface. Dose prescriptions were 65–70 Gy in 30–35 fractions for PTV1 and 54–56 Gy in 30–35 fractions for PTV2. OAR such as parotids, brainstem (BS) and spinal cord (SC) were contoured. SC planning organ-at-risk volumes (SC PRVs) were formed using a 3-mm uniform expansion. All IMRT plans were generated using the Monaco® (Elekta, Stockholm, Sweden) inverse treatment planning system which utilized a Monte Carlo dose calculation algorithm.

Figure 1.

Head and neck immobilization using a thermoplastic shell and shoulder depressors.

Image acquisition and co-registration

Per institutional protocol, all patients were imaged daily using on-board kV CBCT (Synergy® XVI; Elekta Stockholm, Sweden) for the first three fractions of their treatment. The system then calculated the systematic error, and this was applied for the subsequent fractions. Further CBCTs were acquired on fractions 4 and 5, and if patient set-up was within tolerance a CBCT was then acquired weekly. The weekly CBCTs acquired during treatment were as follows (number of patients with CBCT taken at fraction n and the range of fraction numbers during which CBCTs were acquired for each week, where appropriate): fraction 1 (n = 15), fraction 6 (n = 15), fraction 11 (n = 11; range, 9–12), fraction 16 (n = 15; range, 15–16), fraction 21 (n = 12; range, 20–22), fraction 26 (n = 12; range, 25–27) and fraction 31–32 (n = 2). All CBCT images were reconstructed to 3-mm slice thickness and transferred to the focal treatment planning system (Elekta). Automated rigid image registration was used to align each CBCT to the original planning CT, using the C2 vertebra as the reference point because of its proximity to the high-dose CTV and the greater risk of overdosing the SC with any misalignment at this region. This was adjusted manually if the automated image fusion was not optimal. Following image registration, the original planning CTVs, OARs (parotids, BS and SC) and external body contours were copied onto each per-treatment CBCT. The per-treatment CTVs were modified to account for significant anatomical changes only (i.e. CTV extending beyond the skin) by a single oncologist (CY). CTVs were not specifically modified to account for tumour shrinkage. Following modification, a 4-mm expansion was applied to form the per-treatment PTV. The parotids, BS, SC and external body contours were modified to account for anatomical changes. All modified per-treatment CBCT contours, including external body contours, were copied onto the original planning CT, which was then imported to the Monaco treatment planning system for dose calculation. The original IMRT plan was applied on the original planning CT overlaid by the weekly modified per-treatment target volumes, OARs and external body contour to generate a per-treatment dose–volume histogram (DVH).

Volumetric and dosimetric analyses

Doses received by 95% (D₉₅) and 2% (D₂) of CTV1, CTV2, PTV1 and PTV2 were extracted from the DVH. The CTVs and PTVs were also obtained. The following additional parameters were analysed: SC and SC PRV, D_max and D₂; BS, D_max and D₂; and ipsilateral and contralateral parotids, volume, volume receiving ≥26 Gy (V₂₆) and D_mean.

Statistical analysis

The cumulative dose at each week was calculated by the summation of the calculated dose delivered by the original plan prior to each weekly CBCT and the total dose delivered for the remaining fractions using the new per-treatment plan following each CBCT. Weekly percentage volumetric and cumulative dosimetric changes were defined as:

$$\left(\frac{\text{Per}\text{-}\text{treatment\hspace{0.17em}volume\hspace{0.17em}or\hspace{0.17em}cumulative\hspace{0.17em}dose}-\text{original\hspace{0.17em}volume\hspace{0.17em}or\hspace{0.17em}dose}}{\text{Original\hspace{0.17em}volume\hspace{0.17em}or\hspace{0.17em}dose}}\right)\hspace{0.17em}\times \mathrm{\hspace{0.17em}100}$$

As there were only two patients who had 7 weeks of treatment, the cumulative dose analysis was limited to the initial six weekly per-treatment plans. The delivered dose was derived by the summation of the calculated delivered dose each week based on all the weekly per-treatment plans, inclusive of the Week 7 per-treatment plans for two patients. For patients who required a replan, cumulative and delivered doses were calculated using the original dosimetry prior to the start of the new treatment plan and using the dosimetry derived from the new plan for subsequent fractions.

A subgroup analysis of the patients who required treatment replan was performed. Volumetric changes at the point of treatment replan were compared with original volumes. Delivered doses up to the time of treatment replan were calculated by the summation of the calculated weekly delivered doses prior to replan and compared with original planning dosimetry.

Statistical analysis was performed using the SPSS® software v. 21.0 (IBM Corporation, Armonk, NY).

RESULTS

Volumetric changes

The greatest average weekly volumetric reductions were noted in ipsilateral parotids (absolute range, −0.9 to −7.4 cm³; relative range, −2.9 to −27.0%) and contralateral parotids (absolute range, −1.1 to −6.1 cm³; relative range, −4.1 to −23.3%) (Figure 2). Although CTVs and PTVs shrank throughout treatment, these changes were small with an average weekly absolute and relative volume reduction ranging from −17.3 to −2.8 cm³/−4.9 to −2.2% (CTV1), −11.6 to −5.1 cm³/−6.9 to −3.7% (CTV2), −69.2 to 35.6 cm³/−5.9 to −1.0% (PTV1) and −31.0 to −9.6 cm³/−5.3 to 0.7% (PTV2), respectively (Figure 2).

Figure 2.

Mean percentage volumetric differences for clinical target volume (CTV), planning target volume (PTV) and parotids.

Dosimetric changes

The average weekly cumulative differences in D₉₅ and D₂ for CTV1 (D₉₅: range, 0.1–0.3%; D₂: range, 0.5–1.8%), CTV2 (D₉₅: −0.6 to 0.6%; D₂: −3.0 to −0.3%), PTV1 (D₉₅: −0.8 to −0.3%; D₂: 0.6–2.2%) and PTV2 (D₉₅: −1.9 to −0.8%; D₂: −0.04 to 1.3%) were small and clinically insignificant (Figure 3a,b).

Figure 3.

(a) Mean percentage weekly doses received by 95% (D₉₅) differences for target volumes. (b) Mean percentage weekly doses received by 2% (D₂) differences for target volumes. (c) Mean percentage weekly dosimetric differences for ipsilateral and contralateral parotids. (d) Mean percentage weekly dosimetric differences for brainstem (BS), spinal cord (SC) and SC planning organ at risk volume. CTV, clinical target volume; D_max, maximum dose; D_mean, mean dose; PTV, planning target volume; V₂₆, volume receiving ≥26 Gy.

There was an increase in the dose received by the parotids with time, with higher weekly mean cumulative V₂₆ in the contralateral parotid (range, 1.9–6.3%) than in the ipsilateral gland (range, 0.5–3.2%). Changes in the parotid D_mean were less consistent (Figure 3c). BS, SC and SC PRV had lower weekly mean cumulative D_max than the original plan. However, these normal tissues received a higher weekly mean cumulative D₂ than the original plan although these changes were small (0.1–2.8%) and not clinically significant (Figure 3d).

The percentage differences in the calculated delivered doses to target volumes were not consistently lower than the original plan (Figure 4a). The range of differences in delivered doses for D₉₅ and D₂ were −6.7 to 4.8% and −7.4 to 5.9%, respectively.

Figure 4.

(a) Percentage differences in delivered doses to target volumes. (b) Percentage differences in delivered doses to organs at risk. Top of boxplot, 75% quartile; bottom of boxplot, 25% quartile, top of whisker, maximum value; bottom of whisker, minimum value. BS, brainstem; CTV, clinical target volume; D₂, doses received by 2%; D₉₅, doses received by 95%; D_max, maximum dose; D_mean, mean dose; PRV, planning organ-at-risk volume; PTV, planning target volume; SC, spinal cord.

There were small increases in the delivered D_mean to the ipsilateral and contralateral parotids: 1.9% (mean)/1.3% (median) and −1.0% (mean)/1.7% (median), respectively (Figure 4b). By contrast, D_max was lower for SC, SC PRV and BS with an average decrease of 5.2%, 3.5% and 4.7%, respectively (Figure 4b). SC, SC PRV and BS D₂ increased although these changes were not clinically significant with an average difference ranging from 1.9% to 3.0% (Figure 4b).

Subgroup analysis for patients with treatment replan

Five patients underwent repeat planning CT scans after a mean dose of 30.70 Gy (range, 21.67–41.17 Gy) and started treatment on the new plan at a mean dose of 40.41 Gy (range, 30.34–46.00 Gy). The average weight change prior to treatment replan was −2.0%.

There were greater volumetric reductions in CTV2 (mean, −6.9%) and PTV2 (mean, −11.5%) than in CTV1 (mean, −2.5%) and PTV1 (mean, −4.7%) prior to treatment replan (Figure 5a). The average parotid shrinkage was 10.4% and 12.1% in the ipsilateral and contralateral parotids, respectively.

Figure 5.

Volumetric and dosimetric differences in patients with treatment replan showing the mean and 95% confidence intervals. (a) Percentage volumetric differences; (b) percentage target volume dosimetric differences; (c) percentage organ-at-risk dosimetric differences. BS, brainstem; CTV, clinical target volume; D₂, doses received by 2%; D₉₅, doses received by 95%; D_max, maximum dose; D_mean, mean dose; PRV, planning organ-at-risk volume; PTV, planning target volume; SC, spinal cord.

There were no clinically significant dosimetric differences in target volume coverage (Figure 5b) with the majority of patients having <2.5% variation in D₉₅ and D₂. There was generally an increase in SC, SC PRV and BS D₂, with most of these changes being <3.5% (Figure 5c). By contrast, D_max was lower than the original plan. The changes in D_mean and V₂₆ ranged from 4.7% to −7.4% (Figure 5c).

DISCUSSION

Our study showed that it is feasible to use co-registered kV CBCT acquired as part of treatment verification to planning CT for dose estimation in the delivery of HNSCC IMRT using a different methodology from those that had been previously published.^9,10,12 The mean weekly cumulative CTV/PTV dosimetric differences were small, with an average difference ranging from −3.0% to 2.2%. Similarly, the average difference in delivered doses to the target volumes were within ±5% for most patients and not clinically significant. Even in patients who required treatment replans, the dosimetric differences were not large, although this is limited by the small number of patients in this subgroup analysis. We included patients who underwent curative concurrent chemo-IMRT with bilateral neck irradiation because this cohort would theoretically derive the most benefit from ART compared with patients receiving unilateral neck irradiation or adjuvant radiotherapy. Our findings suggest that most patients undergoing curative IMRT do not require ART routinely, as although there is a shrinkage of target volumes and parotids this did not lead to clinically relevant reduction in PTV coverage or significant overdosing of critical normal structures. Nonetheless, treatment replanning was necessary in some patients because of excessive set-up uncertainty and/or weight loss.

The use of kV CBCT for dose calculation has been investigated previously.^9,12–15 Previous studies showed that the conversion of Hounsfield units to electron density for direct dose calculation using kV CBCT is feasible with estimated dose differences between planning CT and kV CBCT of up to 5%.^12,13,15 This involves a pixel correction strategy that requires validation with phantom studies and the generation of CT number–electron density curves. Our study used a simpler dose estimation method in HNSCC IMRT, which could be implemented in most institutions by adapting the original target volumes, OARs and external body contour based on an up-to-date CBCT anatomy. These updated contours were overlaid on the original planning CT for dose calculation. This eliminates the need for the more resource-intensive Hounsfield unit conversion. This dose estimation method could be used as an initial screening strategy to identify patients who will benefit from ART, thus reducing the number of IMRT replans and planning workload and making ART a viable strategy in routine clinical practice.¹⁶ Although the image quality of kV CBCT is improved compared with MV CBCT, it is still inferior to kV CT because of its greater scatter radiation component, which is dependent on the volume imaged, and beam hardening effect resulting in poorer image contrast and increased artefacts.^17–19 Thus, accurate target and organ delineations are hard to achieve using kV CBCT, making direct dose calculations using CBCT difficult to implement.

Our study is limited by the small number of patients and the retrospective nature of the analysis. A prospective study will benefit from a standardized study-specific CBCT acquisition protocol. Accurate delineation of target volumes and OARs is difficult owing to poorer CBCT image quality. Thus, we evaluated CTV and PTV rather than GTV, as these volumes are less prone to delineation uncertainty. Our modification of volumes was performed in response to anatomical changes. We did not modify CTV or PTV coverage based on tumour shrinkage, as the role of reducing treatment volumes during radiotherapy is unclear at present and may compromise local control. There could be an element of intraobserver variability in volume modifications, although interobserver variability is not applicable in this study, as all modifications were performed by a single oncologist. We used rigid image registration followed by manual optimization in this study. Deformable registration could be useful in this context and could eliminate interobserver variation in contour modification in the clinical setting.

Our findings were similar to those of Ho et al,⁹ who investigated the dosimetric impact of weight loss or tumour shrinkage on parotids and other normal structures in 10 patients with locally advanced oropharyngeal cancers treated with chemo-IMRT prospectively. The authors used a shading correction algorithm on each weekly kV CBCT, and the corrected CBCT was used for subsequent dose calculation. Despite weight loss and parotid shrinkage, no significant dosimetric difference was observed in the parotids, BS and SC.

Another prospective study assessed the role of ART in 22 patients with locally advanced oropharyngeal squamous cell carcinoma using an in-house deformable registration algorithm.²⁰ All patients required at least one replan and eight patients had two replans. After a median follow-up of 31 months, there was no local failure and one nodal relapse. Acute toxicities were comparable to those in a conventional IMRT cohort. The authors implemented zero-margin CTV to PTV expansion for all adaptive planning, although a 3–4 mm expansion was used in the initial treatment planning. Despite that, this study suggests that ART is feasible and safe to be used in clinical practice.

CONCLUSIONS

Most patients with HNSCC undergoing curative IMRT do not require ART routinely despite volumetric reduction of their tumours and parotids. Co-registration of kV CBCT to planning CT for estimation of actual dose received during curative IMRT is feasible in these patients and has the potential to be used as part of the ART strategy.

FUNDING

CY receives funding support from the National Medical Research Council, Singapore.