Feasibility of a Novel Surface-Guided Setup Technique to Reproduce Neck Curvature Using two Regions of Interest

Abstract

Purpose

To improve the setup reproducibility of neck curvature using real-time optical surface imaging (OSI) guidance on 2 regions of interest (ROIs) to infer cervical spine (c-spine) curvature for surface-guided radiotherapy (SGRT) of head-and-neck (HN) and c-spine cancer.

Methods

A novel SGRT setup approach was designed to reproduce neck curvature with 2 ROIs: upper-chest ROI and open-face ROI. It was hypothesized that the neck curvature could be reproduced if both ROIs were aligned within ±3 mm/2˚ tolerance. This was tested prospectively in 7 volunteers using real-time 3D-OSI guidance and lateral 2D-photography verification after the 3D and 2D references were captured from the initial conventional setup. Real-time SGRT was performed to align chest-ROI and face-ROI, and the longitudinal distance between them was adjustable using a head-support slider. Verification of neck curvature anteriorly and posteriorly was achieved by overlaying edge-extracted lateral pictures. Retrospectively, the relationship between anterior surface and spinal canal alignment was checked in 11 patients using their simulation CT (simCT) and setup cone-beam CT (CBCT). After the anterior surface was rigidly aligned, the spinal canal alignment was checked and quantified using the mean-distance-to-agreement (MDA) and DICE similarity index, and surface-to-spine correlation was calculated.

Results

The reproducibility of neck curvatures using the 2xROI SGRT setup is verified and the mean neck-outline-matching difference is within ±2 mm in lateral photographic overlays. The chest-ROI alignment takes 110 ± 58 s and the face-ROI takes 60 ± 35 s. When the anterior body surface is aligned (MDA = 1.1 ± 0.6 mm, DICE = 0.96 ± 0.02,) the internal spinal canal is also aligned (MDA = 1.0 ± 0.3 mm, DICE = 0.84 ± 0.04) in 11 patients. The surface-to-spine correlation is c = 0.90 (MDA) and c = 0.85 (DICE).

Conclusion

This study demonstrates the feasibility of the novel 2-ROI SGRT setup technique to achieve reproducible neck and c-spine curvature regardless of neck visibility and availability as ROI. Staff training is needed to adopt this unconventional SGRT technique to improve patient setup.

Introduction

In radiotherapy of head and neck (HN) cancer, including oral cavity, pharynx, paranasal sinuses, and salivary glands, the local lymph nodes in the neck region are often involved and need to be treated as the target. Therefore, it is important to reproduce the cervical spine (c-spine) curvature at the patient setup, as it affects the locations of the lymph nodes. Moreover, to treat HN cancer recurrence or c-spine metastasis, reproducing the c-spine curvature is required for targeting multiple lesions or multiple vertebral bodies in hypo-fractional stereotactic body radiotherapy (SBRT).^1–3 As the targeting area decreases and the fractional dose increases, the accuracy of the setup becomes more important.^4,5 However, fast and accurate setup with real-time guidance to reproduce the c-spine curvature remains a clinical challenge as image-guided radiotherapy (IGRT) setup is incapable of providing such real-time guidance on how much the patient's neck should be adjusted (bent or stretched) compared with simulation computed tomography (simCT).^6–9 In contrast, a surface-guided radiotherapy (SGRT) setup applies a real-time optical surface imaging (OSI) technique to align patient skin surface,^10–14 which may or may not infer bony structures reliably and accurately, depending on anatomic sites and selection of region of interest (ROI). In rigid anatomies, such as the brain, clinical SGRT applications have been successful as the external-internal relationship is fixed and the open-face ROI is reliable.^11,14,15 However, for deformable anatomies, such as HN, better understanding the deformable relationship is the key to having successful SGRT applications. Whether OSI can produce useful guidance to infer internal bony structure in HN SGRT setup provides a new opportunity to address this challenging clinical need.

In recent years, SGRT has become a popular approach for fast and accurate patient setup, promising to replace the tattoo-based conventional setup.^12,14,16,17 Studies of SGRT setup include brain,^11,15 nasopharynx, ¹² breast,^18,19 prostate, ²⁰ and extremity patients. ²¹ In these clinical studies, relatively rigid surfaces are often selected as the ROI for alignment, avoiding using deformable ROI. It should be noted that SGRT may only be used for pre-alignment before radiographic imaging in IGRT setup or if sufficient treatment margin is applied to account for SGRT setup uncertainties.^10,14 For HN patient setups, the neck is highly deformable anatomy, especially for obese necks with skin folds and short necks that may not be visible, so existing SGRT techniques with a neck ROI may not be useful and incapable of inferring internal structure. Additionally, attempts have been made to use the face ROI for HN patient setup, but the uncertainty becomes location-dependent: the further away from the skull, the higher the setup uncertainty due to the c-spine curvature variation,^22,23 because the face ROI does not provide any indication of c-spine curvature and therefore may not be useful beyond brain and nasopharynx setup. ¹²

In this study, we proposed a novel approach to perform an SGRT HN setup using 2 ROIs to reproduce the neck (c-spine) curvature. First, we conducted SGRT setup experiments in seven healthy volunteers with upper-chest ROI and open-face ROI and aligned them in sequence using a 3D-OSI reference and a 2D-lateral-photography verification picture captured in the SGRT setup, illustrating the capability of aligning deformable anatomic site regardless neck visibility. Second, to demonstrate the external-internal relationship in the HN region, we evaluated cone-beam CT (CBCT) and simCT in 11 HN patients in their SBRT treatments and calculated the correlation between the anterior surface and the spinal canal, demonstrating the potential of this novel 2-ROI SGRT setup technique to infer internal structure from the external surface.

Methods

Study Design Using two Regions of Interest

In this study, AlignRT (version 5.0, VisionRT, Inc, UK) was used as the OSI system with six degrees of freedom (6DOF) for the SGRT setup experiment. We defined and applied two ROIs on relatively rigid anatomies: the upper chest and the face, avoiding the deformable neck, which may not be available for subjects with a short neck or an obese neck with skinfolds. The upper-chest ROI was defined as the chest superior to the nipples and include the shoulders while the face ROI was defined the same as the usual open-face area of the mask, superior to the lips to the forehead and covering the cheeks laterally. It was hypothesized that if both ROIs were aligned well the neck (c-spine) that links the two adjacent ROIs should fall into the same position naturally and reproducibly. A specially designed and manufactured adjustable head-support slider was used in this study through collaboration with a vendor (CDR Systems, Alberta, Canada), as shown in Figure 1. It allowed us to change the distance between the 2 ROIs, altering the neck extension and thus its curvature. The head-support slider contains 2 layers of 5mm-thick carbon fiber substrates: the bottom one was attached to the couch and the top hosted a standard headrest. A turning handle on the superior side was used to move the top layer position using a rotation-to-translation mechanism with a ruler indicator (adjustable range = ±1.5 cm).

Figure 1.

A new head-support slider device that can slide in the longitudinal direction by 3 cm using a white T-handle to adjust the top substrate position (blue arrow) with a built-in ruler (red arrow) while the bottom substrate can be fixed on the couch. Different headrests can be mounted on the top layer to support the subject's head and neck area. (A) The headrest at an inferior position, (B) the headrest at a superior position, and (C) a zoomed-up of the positioning ruler.

The upper-chest ROI covered both shoulders in an arm-down position and a shoulder strap was used to facilitate shoulder re-positioning. For the face ROI, the definition was the same as that for the stereotactic radiosurgery (SRS) patients to include the facial area above the lips within the opening area of the open-face mask. ¹⁵ Under real-time OSI guidance, the upper-chest ROI was aligned first by manually adjusting subjects, followed by the face ROI alignment via adjusting the head and the head-support slider (Figure 1).

the SGRT Setup Workflow and Neck Curvature Verification

Seven healthy volunteers (4 male and 3 female) participated in the SGRT experiment. The subjects were first “simulated” in the treatment room using a conventional setup procedure with skin marks on sticky tapes. An OSI reference image was acquired and the 2 ROIs were drawn. A lateral photographic picture was taken using a camera on a tripod at a couch-90˚ position ∼3 m away from the isocenter. After getting off the couch, the subjects were set up again using real-time AlignRT guidance for SGRT setup, in which the chest-ROI was first aligned, followed by facial-ROI alignment, under a matching tolerance of 2 mm/2˚ residual errors in any direction, except for 3 mm for lateral alignment. After the SGRT setup with both ROI alignments within the setup tolerance, a second lateral picture was taken and overlayed to the reference picture. Both photographic pictures were processed with edge extraction and overlayed with a semi-transparent setting, verifying the alignment of both anterior and posterior neck surface outlines. The reproducibility of the neck setup was estimated by comparing neck outline alignments anteriorly and posteriorly with the 2 ROI outlines (within a threshold of 2 mm).

Owing to the lack of radiographic imaging in the volunteer experiment, only photographic pictures were applied for the assessment of the neck curvature. As both ROIs were aligned within 2 mm in any direction (except for lateral shifts, which cannot be viewed in the lateral pictures), they served as internal references to verify the alignment accuracy of neck lateral outlines. Namely, by comparing the surface outline matching at the upper chest and facial regions with the neck region (on both anterior and posterior sides), the similarity between the outline matchings at various locations can be observed and assessed whether the 2 mm threshold was valid.

Statistical Analyses of Surface-to-Spinal-Canal Correlation and Predictability

To demonstrate the OSI is useful for reproducing the neck and c-spine curvatures, 11 HN hypo-fractional SBRT patients were selected to use their daily setup CBCT and simCT for a retrospective analysis. A CBCT-simCT pair with good registration per patient was chosen to assess the external-internal relationship using the MIM Research module (MIM Maestro, MIM Software, Inc. Cleveland, OH). After clinical setup data was transferred, both external body and spinal canal contours were created manually, and two rectangular contours were created to make two Boolean structures that have the same ROI on both CBCT and simCT for comparison. An anterior surface region similar to the surface ROIs was used to mimic OSI surface data, while the spinal canal was chosen to represent the c-spine ROI structure. Both the mean distance to agreement (MDA) and DICE similarity index were calculated to show the quality of the ROI matching between CBCT and simCT. The definition of DICE similarity coefficient is defined between two 3D objects (A and B) in space as shown below:

$$DICEcoefficient=2\times {\displaystyle \frac{A\cap B}{A\cup B}}$$ (1)

where A ∩ B is the overlapped volume and A ∪ B is the combined volume. The closer to DICE = 1 the higher the similarity between A and B. Further, the correlations of MDA and DICE between the anterior surface and spinal canal among the 11 patients were calculated to assess the external-to-internal relationship. The Pearson correlation coefficient is defined as:

$$Correlation(c)={\displaystyle \frac{\sum (x_i-\overline{x})(y_i-\overline{y})}{\sqrt{\sum {(x_i-\overline{x})}^2\sum {(y_i-y)}^2}}}$$ (2)

Where $x_i$ and $y_i$ are the two series of data and $\overline{x}$ and $\overline{y}$ are the corresponding mean values.

Results

Importance of Aligning Upper-Chest ROI Before Open-Face ROI in the 2-ROI SGRT Setup

The 6DOF SGRT alignment was performed on the chest ROI first within 2 mm, as shown in Figure 2A and 2B. The 2-ROI setup aligns the chest ROI first and face ROI second, using the head-support slider. The lateral view shows that the facial surface needs adjustment to improve the face ROI alignment, while the ROI remains aligned (the 6DOF values remain unchanged). Note that the person's neck is short and not useable to draw an ROI in the neck region.

Figure 2.

Visual demonstration of the upper-chest ROI remains aligned (6DOF readings) before (A) and after (B) the face ROI alignment. (A) After the upper-chest ROI is aligned, the face ROI is not aligned (∼10 mm at two red arrows), which is improved (B) (<2 mm, one red arrow) by adjusting the head and sliding the headrest. Reference (pink) and verification (green) surfaces are shown. Note that the upper-chest ROI alignment is achieved within 2 mm/2˚ as shown in (A) although it appears pink or green (the upper-chest ROI is opaque). A small green area under the chin is the marker tape getting loose.

Figure 3 shows two examples of SGRT setup in the frontal view from the 2-ROI SGRT setups. For all 7 volunteers, the 2-ROI SGRT setup is performed to have residual errors smaller than 2–3 mm for both ROIs. It should be noted that the thresholds are set to each of the six degrees of freedom, not on the magnitude (MAG). Table 1 tabulates the maximum translational (T) and rotational (R) shifts in any direction of the 6DOFs, together with the SGRT setup times (t) for each ROI.

Figure 3.

Two examples of ROI definitions (solid pink) on the upper-chest ROI and the face ROI and their alignments by the verification images (mesh green) within 2 mm (A) or 3 mm (B) in any direction (note: the thresholds are set for each of translational and rotational directions, rather than the magnitude or MAG). It is worth noting that the 2.7 mm and 2.8 mm lateral shifts in subject 2 do not affect the lateral body outline in photographic imaging. The residual setup errors in 6DOF are shown for each step.

Table 1.

the maximum Residual Translational (T in mm) and Rotational (R in ˚) Shifts After Upper-Chest ROI and Face ROI Alignments in the 7 Volunteer SGRT Setups. the SGRT Setup Time (t in s) for Each ROI Alignment is Provided

Subject	Sex	Upper-chest ROI			Face ROI
		T (mm)	R (°)	t (s)	T (mm)	R (°)	t (s)
1	M	−1.7	1.0	70	1.1	1.1	22
2	M	−1.4	−1.8	108	−1.9	0.3	120
3	M	−1.9	−1.4	230	2.3	−2.1	73
4	F	2.2	1.1	69	−2.6	1.6	52
5	F	−0.7	0.3	70	2.0	0.8	84
6	M	1.7	0.9	86	1.1	−2.0	50
7	F	2.7	1.8	135	−2.8	1.9	136
Average		0.0	0.3	110	0.0	0.2	60
St dev		2.0	1.4	58	2.1	1.6	35

Validation of the Neck Curvature Using the Lateral Photographic Pictures

Figure 4 illustrates two examples of validation of the neck curvature after the 2-ROI SGRT setup. A rigid alignment of the two edge-extracted lateral photographic pictures is applied before assessing the neck alignment in the anterior and posterior sides between the reference picture taken at the initial simulation and the post-SGRT setup picture for verification. As the outlines on the 2-ROI regions are both aligned within 2 mm, the anterior and posterior neck outlines are also aligned with a similar precision, as shown by the red arrows in Figure 4.

Figure 4.

Two examples of fused edge-extracted lateral photographic pictures (A and D) of the initial references (C and F) and the final verifications (B and E) after SGRT setup using the 2-ROI alignment technique. The alignment of upper-chest ROI and face ROI results in a well-aligned neck position on both anterior and posterior sides (red arrows) in (A) and (D). A few pieces of tape were shown on the face and neck to draw marks for laser alignment.

SGRT Setup Time Using the 2-ROI SGRT Technique

As the 2-ROI SGRT setup is very different from the conventional single-ROI SGRT setup and requires manual patient adjustments, the setup times are relatively long, as shown in Table 1. The average setup time is longer for upper-chest ROI alignment at 110 ± 58 s, as the shoulder is deformable and requires body adjustment with deformation assessment. The time for face ROI alignment is shorter at 60 ± 35 s as the ROI has been used in clinical SRS patient setup, although the headrest slider is new to the therapists. Figure 5 illustrates a real-time face ROI alignment by adjusting a subject's head and the head-support slider. The idling time (flat curves) is due to the discussion between physicists and therapists on how to proceed next.

Figure 5.

Illustration of the longest SGRT setups for the face ROI using the head-support slider and manual adjustment. Adjusting (1) the head vertical position (red), (2) the longitudinal position (green) using the head-support slider, and (3) the other degrees of freedom (DOF) via manual fine-tuning of the head position. Note that the setup time can be reduced to 61 s and faster SGRT setup can be achieved after the learning curve through more practice.

Surface-to-Spinal-Canal Correlation Based on CBCT-SimCT Setup Analysis

The clinical CBCT-simCT setups of 11 patients are used to assess the external-to-internal relationship, namely the correlation of the anterior external contours and spinal canal contours in both simCT and CBCT, as shown in Figure 6A and 6B. The spinal canal is used to represent the c-spine position and curvature. Figure 6C shows the reproducibility of CBCT setups in 5 fractions. The MDA and DICE values for the surface and canal are shown in Table 2. The DICE values are 0.96 ± 0.02 for the anterior body surface and 0.84 ± 0.04 for the spinal canal. The correlation coefficients are c = 0.90 for MDA and c = 0.85 for DICE between the two sets of 3D contours. Figure 7 illustrates the linear correlation between the anterior surface and the spinal canal using the values of MDA and DICE.

Figure 6.

Two examples (A, B) of anterior body surface and spinal canal segmentation and their relationship between simCT (gray image, blue contour) and CBCT (red image, green contour). (C) A visual example of an HN patient's five setups CBCT (red) fused with the planning CT (grey image, white body contour, dark-red canal contour), suggesting when the two anterior ROIs are aligned, the spinal canal is aligned with 2–3 mm uncertainties.

Table 2.

the MDA and DICE Comparison Between CBCT and SimCT for the Anterior Body Surface and Internal Spinal Canal. the MDA Values are Around 1 mm and DICE is Close to Unity

Patient	Sex	Age	MDA (mm)		DICE
			Surface	Canal	Surface	Canal
1	F	53	0.930	0.819	0.973	0.849
2	F	73	0.762	0.794	0.964	0.850
3	M	84	1.271	1.052	0.957	0.827
4	M	72	1.968	1.430	0.931	0.763
5	F	53	1.076	1.004	0.978	0.835
6	F	75	0.790	1.058	0.960	0.820
7	F	34	0.729	0.921	0.968	0.853
8	M	70	0.488	0.593	0.978	0.903
9	F	67	1.646	1.015	0.947	0.839
10	F	67	0.361	0.688	0.986	0.875
11	M	66	1.955	1.541	0.944	0.786
AVG			1.1	0.99	0.96	0.84
STD			0.6	0.29	0.02	0.04
Cor			0.90		0.85

Figure 7.

The linear correlation fittings between the anterior surface and spinal canal using MDA (A) and DICE (B) values illustrate the external-to-internal relationship.

Discussion

Validation of Neck Curvature Using Lateral Photographic Pictures in 7 Volunteers

The objective of this study is to demonstrate that the 2-ROI SGRT setup can reproduce the neck and c-spine curvature. We hypothesized that if the two ROIs at both immediate ends of the neck are aligned, then the c-spine curvature should be reproduced to the simulation position at the treatment setup. To validate this hypothesis, we have chosen to use photographic pictures from the lateral view to show the alignment of both anterior and posterior neck surfaces in the volunteers, due to the lack of radiographic images. As shown in Figure 4, the alignment of the edge-extracted pictures is a direct means to illustrate the neck curvature reproduction between the reference and SGRT setup. As the 2-ROI SGRT setups were performed after the subjects were off and back on the couch, the SGRT setup is independent of the initial simulation setup, similar to clinical conditions. It is worthwhile to indicate that the posterior headrest, preferably customized headrests, is an important factor in reproducing neck curvature.

From the volunteer experiment, we can only assess the external surface and use it to imply the reproducibility of c-spine curvature. To make up for the lack of radiographic verification, a retrospective analysis of 11 SBRT HN patients is evaluated to provide supporting evidence to show the relationship between ROI surface reproducibility and c-spine curvature reproducibility.

Surface-to-Spine Correlation Based on CBCT-SimCT Registration in 11 Patients

To demonstrate surface-to-spine correlation, 11 HN patients with 5-fraction SBRT treatments and daily CBCT images are selected for assessment, as shown in Figures 6 and 7 and Table 2. In the CBCT-simCT registration, both the upper chest and facial areas on the anterior surface are aligned, so the virtual SGRT setup is unbiased, unlike if only the facial area is aligned in some previous studies, leading to larger setup errors in areas further away from the face ROI.^22,23 In our clinical experience, SGRT setup using a face ROI is only valid and reliable for brain and nasopharynx setup. ¹² Over the 11 patients, the correlation between the anterior body surface and internal spinal canal is very high c = 0.90 for MDA and c = 0.85 for DICE.

As shown in Figure 6C, the customized headrest was applied for the SBRT patients with daily CBCT for setup, the clinical CBCT-simCT rigid registration shows not only anterior alignment but posterior alignment. As AlignRT surface guidance only uses the anterior ROI alignment, the posterior alignment relies on the customized headrest, which covers most of the posterior neck. Therefore, combining the 2-ROI SGRT setup with a patient-specific headrest, neck, and c-spine curvature reproducibility is feasible, providing an effective and efficient technique to realize an accurate and reproducible HN setup.

It is worthwhile to indicate that this study only focuses on the spinal canal because bony structures are often used as primary landmarks in CBCT setups. Although we did not mention the skull structure, it is well represented by the face ROI, an accurate and reliable surrogate for brain treatment setups. ¹² We did not mention the planning tumor volume (PTV) and nearby organs at risk (OARs) as they are often not clearly visible in CBCT, so the bony structures are used as the surrogate for PTV localization during IGRT setup. The long superior-inferior extent of the spinal canal traversing through the neck is also an intuitive surrogate for neck flexion and associated local lymph nodes.

Advantages of SGRT Setup Using two ROIs

One of the major advantages of this 2-ROI SGRT setup procedure is that it is the most versatile approach, as it is suitable for almost all patients, including those with a short neck (Figure 2) or an obese neck with skinfolds. The upper-chest ROI and face ROI are always available regardless of patients’ anatomy, and therefore, this approach should be universal. Additionally, the upper-chest ROI has limited deformation that can be corrected with manual manipulations while the face ROI is rather rigid. Therefore, they are useable and reliable ROIs, although the inclusion of shoulders may be unnecessary but complicates the alignment. From this study, the hypothesis of neck reproducibility has been validated to have the potential for clinical use. As the two ROIs may not include the neck surface, it is not affected by possible shrinkage of a superficial neck tumor. Additionally, as the two ROI surfaces are well supported by the bony structures directly underneath, the SGRT setup variations resulting from patient weight loss will be minimal, especially for hypo-fractional SBRT treatments that are delivered within 3–5 days.

In comparison, when using the face ROI only, studies have demonstrated that the alignment in the neck area is getting worse as it is further away from the skull.^22,23 Our study demonstrates that both ROI alignments are required to reproduce the neck and c-spine curvature, and therefore consistent with the previous findings. In other words, using face ROI for SGRT setup can only be reliably applied to brain and nasopharynx cancer patients. ¹² In contrast, the 2-ROI SGRT setup strategy can reproduce the neck/c-spine curvature, and a good surface alignment suggests a good c-spine canal alignment, leading to a good alignment of PTV and OARs as the bony structure is often used as IGRT landmarks in the clinic. Additionally, we always align the upper-chest ROI first as it is a heavier object to align and the face ROI second using the head-support slider for possible head adjustment. Therefore, when aligning the face ROI, the upper-chest ROI remains aligned (see Figure 2), so both ROIs are aligned in a single attempt for the setup.

The adjustable head-support slider is useful to achieve both ROIs alignment, as shown in Figure 2A, in which ∼1 cm misalignment of the face ROI would not be correctable without moving the head in the superior-inferior direction while maintaining upper-chest ROI alignment. The stretching or compressing of the neck in the longitudinal direction allows the 2 ROIs to align to the simulation positions, which leads to the settlement of neck (c-spine) curvature into a subject-specific natural position. As the experimental conditions are identical to the clinical conditions, the procedure can be directly applied in the clinic after sufficient staff training. To the best of our knowledge, this 2-ROI SGRT setup procedure is the first to reproduce neck curvature with real-time guidance. The immediate clinical benefits include fast and accurate SGRT setup and potential reduction of the number of setup CBCTs as CBCT setup verification is required in our clinics if CBCT shifts are greater than a clinical tolerance. ²⁴ This trend has been observed in our clinical brain SRS setups using the face ROI and is expected for the 2-ROI setup procedure as well.

SGRT Setup in Handling Deformable Anatomies

For clinical applications, a tolerance for the SGRT setup should be allowed, and in this study, we allowed a maximum of ±2 mm (±2˚) uncertainty in any direction in 6DOF for each of the ROI alignments. Although, in 3 out of 7 subjects less than ±3 mm was allowed in the lateral direction as it does not affect the anterior-posterior curvature, as shown in Figs. 3B and 5, this should be avoided. The time for both ROI setups is relatively long because it is new to therapists to align the upper chest surface with the shoulders, which can move and deform, even with the use of shoulder straps. The reason to include the shoulder in the upper-chest ROI is to align the body in the lateral direction, but may not be necessary as the upper chest is not flat but curved. Therefore, staff training and further optimization are necessary to accelerate the 2-ROI SGRT setup before clinical implementation.

For face ROI alignment, manually adjust the subject's head position first to minimize the ROI shifts, and then move the head-support slider to correct the longitudinal shift. As the face ROI is more rigid and similar to the clinical SRS setup, its alignment is straightforward to make the residual shifts within the tolerance in 6DOF by correcting head rotation first, followed by translation. ¹⁵ Overall, the alignment of the face ROI is faster than the upper-chest ROI; but both can be further improved with additional practice.

Clinical Benefits, Limitations, and Future Improvements in 2-ROI SGRT Setup

With the ability to control the neck (c-spine) curvature using the 2-ROI SGRT setup method, the deformation of the neck should be minimized, leading to a more accurate and reproducible setup for HN and c-spine patients. Therefore, it can help to achieve more accurate delivery for intensity-modulated radiotherapy (IMRT) or volumetric-modulated arc therapy (VMAT) plans with sharp dose falloff outside of the planning tumor volume. Potentially, the safety margin that accounts for neck curvature variation can be reduced. This is particularly important for hypo-fractional SBRT treatment of multi-lesion with the same isocenter or multi-vertebral bodies and potentially reduces the number of CBCT for patient setup. It will reduce the neck deformation at daily setups, so it is expected to make deformable image registration (DIR) easier with smaller positional deviation from simCT for assessment of the accumulated delivered dose from multi-fractional treatments, avoiding some DIR uncertainties in the current dose warping approach. ⁷

As the proposed method is mostly suitable for hypo-fractional SBRT treatments of HN or C-spine patients, severe patient weight loss is unlikely to occur during the 3–5 fractional treatments, especially if accelerated planning within 2 days can be achieved in the clinic. However, in case of severe patient weight loss that still occurs, the SGRT pre-CBCT setup may still provide a good estimation for the CBCT setup, because the two ROI surfaces are directly supported by bony structures on the face and upper chest underneath. Whether a patient's treatment needs adaptive re-planning can be determined using CBCT via fusion to the simCT. It should be noted that as the ROIs do not cover the neck region, neck tumor shrinkage should not affect the SGRT setup.

Although we have demonstrated the feasibility of reproducing the neck curvature using the 2-ROI SGRT setup procedure, two aspects can be further studied and improved. First, this 2-ROI SGRT approach can be further validated by conducting a prospective SGRT and IGRT patient study to provide both external neck curvature and internal c-spine curvature. Although the retrospective patient CBCT-simCT analysis demonstrates the high surface-to-spine curvature correlation, it is necessary to show the CBCT confirmation after the SGRT setup. Using the c-spine vertebral bodies as landmarks, the reproducibility of the c-spine curvature can be quantified to provide direct evidence of the c-spine curvature reproducibility.

Second, the performance of the 2-ROI SGRT setup can be further improved after therapists are trained and pass the initial learning curve. In brain SRS patient SGRT setup, it only takes about 30 s to complete the setup. In addition, as the AlignRT system upgrades from version 5.0 to version 6.3, the real-time deformation visualization tool is now available so that the SGRT upper-chest ROI setup can be accelerated. After all, the ability to reproduce deformable anatomy for patient setup is a big step forward as there are few means available to guide the setup in real-time and reproduce deformable anatomy in the current radiotherapy clinic.

Third, for conventional fractionation of HN treatments, this 2-ROI setup technique may not be necessary, as sufficient setup margins are given using conventional setup techniques, such as aligning the tragus and nose with room lasers first on a standard headrest and then placing a customized 5-point mask to immobilize the patient. No SGRT and IGRT are used as a standard of care. The SGRT setup technique is developed for hypo-fractional SBRT treatments.

Finally, in this feasibility study, only 7 volunteers and 11 patients’ data are used to demonstrate the proof of the principle of the novel approach. A clinical study with a larger patient cohort should be followed for further evaluation before widespread clinical adoption. Such a clinical study also allows us to quantitatively evaluate if patients’ weight loss is a significant factor in the hypo-fractional SBRT treatment group and how much uncertainty the SGRT setup will introduce compared with daily CBCT.

Conclusion

In this volunteer and patient study, we have demonstrated the feasibility of a novel 2-ROI SGRT setup technique to reproduce neck curvature and validate the reproducibility using both 2D-lateral photographic pictures and 3D-setup CBCT images. The surface-to-c-spine correlation coefficients are high for both MDA and DICE. With the real-time OSI guidance and adjustable head-support device, the 2-ROI SGRT setup can achieve a reproducible neck/c-spine curvature. To apply this new SGRT setup technique, sufficient staff training is required. This 2-ROI SGRT setup technique has a strong potential to improve radiotherapy of HN and c-spine cancer.

Abbreviations

SGRT

Surface-guided radiotherapy

IGRT

Image-guided radiotherapy

CBCT

cone-beam computed tomography

simCT

simulation computed tomography

HN

Head and neck

C-spine

cervical spine

OSI

optical surface imaging

ROI

region of interest

DOF

degree of freedom

MDA

mean distance to agreement

SRS

stereotactic radiosurgery

SBRT

stereotactic body radiotherapy

PTV

planning tumor volume

OAR

organ at risk