ASSESSMENT OF A RADIOTHERAPY PATIENT CRANIAL

IMMOBILIZATION DEVICE USING DAILY ON-BOARD KILOVOLTAGE

IMAGING

JOSEPH HARMON, DEREK VAN UFFLEN, SUSAN LARUE

The purpose of this study was to utilize state-of-the-art on-board digital kilovoltage (kV) imaging to determine

the systematic and random set-up errors of an immobilization device designed for canine and feline cranial

radiotherapy treatments. The immobilization device is comprised of a custom made support bridge, bite block,

vacuum-based foam mold and a modified thermoplastic mask attached to a commercially available head rest

designed for human radiotherapy treatments. The immobilization device was indexed to a Varian exact couch-

top designed for image guided radiation therapy (IGRT). Daily orthogonal kV images were compared to

Eclipse treatment planning digitally reconstructed radiographs (DRRs). The orthogonal kV images and DRRs

were directly compared online utilizing the Varian on-board imaging (OBI) system with set-up corrections

immediately and remotely transferred to the treatment couch prior to treatment delivery. Off-line review of 124

patient treatments indicates systematic errors consisting of þ 0.18mm vertical, þ 0.39mm longitudinal and

�0.08mm lateral. The random errors corresponding to 2 standard deviations (95% CI) consist of 4.02mm

vertical, 2.97mm longitudinal and 2.53mm lateral and represent conservative CTV to PTV margins if kV OBI

is not available. Use of daily kV OBI along with the cranial immobilization device permits reduction of

the CTV to PTV margins to approximately 2.0mm. Veterinary Radiology & Ultrasound, Vol. 50, No. 2, 2009,

pp 230–234.

Key words: immobilization device, IGRT, kilovoltage, on-board imaging, radiation therapy, repositioning

accuracy.

Introduction

EFFECTIVE PLANNING AND delivery of external beam ra-

diation requires thorough knowledge and understand-

ing of patient set-up uncertainties. ICRU1 Report no. 62

(supplement to ICRU Report no. 50) defines a planning

target volume (PTV) generated by adding a safety margin

to the clinical tumor volume (CTV) to account for inter-

fractional and intrafractional movement.

When considering intracranial radiotherapy using robust

immobilization methods, a stable treatment couch and

proper anesthesia, intrafractional variability is negligible.

However, interfractional movement can be significant,

depending on the design of the immobilization device,

repeatability and stability of couch position, accuracy of

set-up lasers and type/availability of portal imaging. The

advantage of kV vs. MV imaging for daily set-up includes

less excess dose accumulation, less interobserver alignment

variability,2,3 improved soft tissue contrast and, depending

on the specific imager design, the potential for improved

spatial resolution. Minimizing the CTV to PTV margin

benefits the patient by minimizing the volume of healthy

tissue irradiated, thus reducing complications. This is of

particular importance in cranial radiotherapy treatments

due to the proximity of the treated volume to critical

structures such as the spinal cord, brainstem, parotid

glands, optic nerves, and inner ear. The reduced PTV

margin may also permit dose escalation.

The goal of this study was to determine the appropriate

PTV margin for intracranial treatments using a newly de-

veloped indexable cranial fixation device. The fixation

device was designed for a wide range of patient sizes, the

ability to permit anesthesia ventilation during treatment,

ease of initial and daily set-up, compatibility with the in-

dexed couch top, and suitability for kV imaging. Tools

used for determining the repositioning accuracy and pre-

cision of the device include orthogonal kV images obtained

with on-board imaging (OBI).

Materials and Methods

Nineteen dogs and three cats were treated with the cra-

nial fixation device with a total of 124 individual treatment

sessions being delivered. Patient fractionation ranged

from 1 to 21 fractions and included three-D radiation

therapy, stereotactic radiosurgery, intensity-modulated ra-

diation therapy, and intensity-modulated radiosurgery.

Address correspondence and reprint requests to Joseph Harmon, at theabove address. E-mail: joseph.harmon@colostate.edu

Received June 5, 2008; accepted for publication September 15, 2008.doi: 10.1111/j.1740-8261.2009.01522.x

From the Environmental & Radiological Health Sciences, ColoradoState University, 1681 Campus Delivery, Fort Collins, CO 80523.

230

The distribution of patient tumor locations included 10

involving the pituitary, two involving the nasal area, and

the remaining 10 distributed throughout the brain.

Components of the immobilization device include a

baseplate, vacuum bead style moldable cushion, custom

upper jaw support bridge, bite block and thermoplastic

mask. The baseplate (Civco model MT-20100CF�) is made

of carbon fiber and attaches (indexes) to a wide variety of

couch tops using a locking bar. A carbon fiber laminate

support bridge was designed and built to rigidly attach to

the base plate using an inner set of pins already existing in

the base plate. For human use, the inner set of pins in the

baseplate are used for a removable headrest. Because pre-

vious animal immobilization studies4–6 have illustrated the

importance of using a bite block for reproducible cranial

fixation, the support bridge was designed with three holes

across the top to permit a custom thermoplastic bite block

(made from thermoplastic pellets, Civco model MT-APS

3A�) to be easily attached and removed between treatments.

Together, the support bridge and bite block secure the

patient’s maxilla during treatment in the prone, head-first,

position. The evacuated bead style cushion (Civco model

MT-VL-37) adapts to the shape of the patient and base

plate during evacuation and supports the patient’s mandi-

ble and neck, as well as providing additional lateral posi-

tioning stability for the upper torso. Finally, a

thermoplastic mask (Civco model MT-APU) is affixed to

the base plate over the top of the immobilization assembly

to secure the patient to the bite block and cushion. The

mask is modified by cutting a triangle-shaped section out

of the cranial end thus permitting undisturbed access to

ventilation tubing during application or removal of the

mask (Fig. 1).

Each patient received an initial CT simulation scan using

a Picker PQ-2000 scanner, typically using a 2mm slice

thickness to balance the need for image contrast and res-

olution of the digitally reconstructed radiograph (DRR).

To ensure the base plate was properly aligned with the

longitudinal axis of the scanning couch, a Varian carbon

fiber OBI couch insert was attached to the Picker couch top

using custom designed adapters. The use of the OBI couch

insert permitted indexing during the scanning process iden-

tical to that during the treatment process. During the CT

simulation process, CT set-up lasers were used to identify a

relative set-up (simulation) isocenter. The location of the

simulation isocenter was marked on the mask and also

identified by use of 1.5mm metal bead markers (Spee-D-

Mark model SDM-BB15�).All treatment plans were performed using a Varian

Eclipse treatment planning system (TPS) using software

version 8.1 and the Analytical Anisotropic Algorithm for

the photon dose distributions. During the planning process,

the treatment isocenter was determined and both dorsal and

lateral DRRs generated for the treatment position. DRRs

were optimized to enhance visualization of bony structures.

On the first day of treatment, using a Varian Trilogy

with 6MV photon beam, each patient was placed within

the immobilization device in the same manner used during

simulation. The treatment room set-up lasers and patient

mask with BBs were then used to place the simulation

isocenter at the location of the treatment room (linac)

isocenter. Appropriate couch shifts (vertical, lateral, and

longitudinal) generated by the treatment planning system

were then applied to the couch to place the treatment plan

isocenter at the treatment room isocenter.

Orthogonal digital radiographs were used to compare

the actual treatment position to the position indicated by

the DRR of the treatment planning system. Typically,

pulsed fluoroscopy mode with automatic brightness con-

trol was used to quickly select optimum technique settings

for the relevant anatomy. The pulsed fluoroscopy images

contain 1024� 768 pixels of image data collected in a

2 � 2 binned mode with an aSi imaging panel. A high

resolution, single exposure, 1 � 1 binned mode is also

available providing 2048� 1536 for select patients that re-

quire higher resolution. For this study, all images were

obtained in the 2 � 2 binned mode to quickly acquire

images. Required couch shifts were determined using

extensive manual and/or automatic matching tools built

into the Varian OBI kV–kV analysis software. Evaluation

Fig. 1. (A) Illustration of cranial immobilization system components 365 � 274mm (96 � 96DPI). (B) Example of cranial immobilization system in use.

�Civco, Orange City, IA.

231CRANIAL IMMOBILIZATION DEVICEVol. 50, No. 2

of patient set-up was determined by comparing unique bony

landmarks using both image overlay and ‘‘spyglass’’ meth-

ods of image comparison. The spyglass is particularly useful

as it permits one to see the DRR within a user defined,

moveable, rectangular portal within the acquired kV OBI

image. Rotation of the patient, if required, was also an op-

tion available during the matching process. Figure 2 illus-

trates the kV–kV matching workspace with an exaggerated

shift. As illustrated in the left lateral view, there is a need to

raise the treatment couch top to overlay the anatomy shown

in the actual kV image with that in the treatment plan

DRR. Similarly, as illustrated in the dorsal view, there is a

need to shift the treatment couch top to the right to overlay

the actual kV image and treatment plan DRR anatomy.

Once a suitable image match was found, the corre-

sponding couch shifts were remotely transferred to the

Fig. 2. Exaggerated example of vertical and lateral shift between digitally reconstructed radiograph (DRRs) and kV radiographs. Spyglass tool is shown.

Fig. 3. Example of daily treatment position corrections for a single patient based on orthogonal kV imaging.

232 HARMON ET AL. 2009

treatment couch. The final couch position for the first

treatment session was used as the reference condition for

all subsequent treatment sessions. Thus, on each subse-

quent treatment session, the patient was placed in the im-

mobilization device, indexed to the original couch notches

and the couch returned to the session 1 coordinates. A set

of orthogonal images are then acquired, compared with the

original TPS DRRs, and any required shifts are applied

remotely. The Varian ARIA patient management system

stores all paired images and records each session’s shift

values for follow-up analysis.

Results

A review of the couch shift required to place the patient

in the correct treatment position was performed for all

patient treatment sessions. An example of the typical couch

shifts during a fractionated treatment course is shown in

Fig. 3. Each shift represents the difference between the

reference (day 1) couch position and the final treatment

couch position. In this example, the vast majority of the

shifts fall within 2mm. A summary of all 124 treatment

sessions is shown in Table 1. The average shift and rotation

values represent the overall set-up accuracy and reflect any

systematic errors during set-up. Regardless of the immo-

bilization device design, systematic errors are possible if the

imaging device is not properly calibrated. Thus, a rigorous

quality assurance procedure is followed based on manu-

facturer specifications and published protocols7 to insure

submillimeter accuracy. The precision of the immobiliza-

tion system represents the random errors associated with

the immobilization system. The precision values presented

in Table 1 correspond to two standard deviations (the 95%

confidence interval).

Because the soft tissue in the neck area does not permit

rigid fixation as the bite block does, we evaluated the data

to determine if a trend existed relating magnitude of av-

erage error with distance of treatment isocenter from the

support bridge. Figure 4 presents the position error as a

function of distance of treatment isocenter from the infe-

rior edge of the support bridge. The data does reveal a

slight trend of increasing error (from approximately

0.5mm error at a distance of 4 cm to approximately

1mm error at a distance of 10 cm) in the vertical and lon-

gitudinal direction but not with rotation or lateral direc-

tion. Because the device has been used for a wide range of

patient sizes, we also investigated average error as a func-

tion of patient weight. Figure 5 presents average position

error as a function of patient weight and indicates very

little correlation between patient weight and error except

possibly in the vertical direction where the error actually

decreases from approximately 1mm at 2kg to approxi-

mately 0mm at 43kg.

Table 1. Summary of Radiotherapy Immobilization Device Patient Set-Up Accuracy and Precision Results (Based on 124 Patient Treatments)

MeasureVertical(mm)

Longitudinal(mm)

Lateral(mm)

Rotation(1)

Average error (accuracy) þ 0.18 þ 0.39 �0.08 �0.0152s (precision) 4.02 2.97 2.53 0.32

Fig. 4. Treatment position corrections vs. distance from support bridge to treatment isocenter.

233CRANIAL IMMOBILIZATION DEVICEVol. 50, No. 2

Discussion

The results of this study indicate submillimeter patient set-

up accuracy (systematic error) and precision (random error)

on the order of 2–4mm. These values represent a conser-

vative estimate of the magnitude of daily set-up variation

corrected for using kV OBI. To treat the same group of

patients at our institution using the cranial immobilization

device without the use of kV OBI, one would need to ensure

PTV margins account for these set-up variations to ensure

the PTV receives the prescribed dose at least 95% of the

time. Because we were not using lasers for the daily patient

set-up, the variability in couch shift corrections reflect a

combination of the accuracy and precision of the digital

couch readout and variability in patient movement within

the immobilization device. Routine quality assurance testing

indicates the couch repositioning variability to be � 1mm.

Thus, the remaining variability presented in Table 1 reflects

patient movement within the immobilization device.

The data presented in Fig. 4 indicate a slight improvement

in overall accuracy the closer the treatment isocenter is to the

support bridge. The trend is most prominent in the vertical

and longitudinal shift axes and not significant for the lateral

shift and rotation correction. The vertical shift error trend is

not unexpected due to the fact the support bridge with bite

block is secure and acts as a pivot point with variation in

patient soft tissue placement within the vacuum foam mold

causing minor vertical displacement variation. Less correla-

tion is noted between positioning error and patient weight as

demonstrated in Fig. 5 thus permitting use of the device on a

wide range of patient sizes.

We have found the cranial fixation device to be easy to

use and, so far, one size support bridge has worked well

with all patients treated. Use of this cranial immobilization

device at an institution with accurate and precise couch

repositioning but without access to kV OBI would require

CTV to PTVmargins ranging from 2 to 4mm to ensure the

PTV received the prescribed dose at least 95% of the time.

Utilizing both daily kV OBI and this immobilization device

permits one to comfortably reduce the CTV to PTV

margin to 2mm for most patients.

REFERENCES

1. International Commission on Radiation Units and Measurements.Prescribing, recording, and reporting photon beam therapy (supplement toICRU Report 50), ICRU Report 62. Bethesda, MD: ICRU Publications,1999.

2. Pisani L, Lockman D, Jaffray D, et al. Setup error in radiother-apy: on-line correction using electronic kilovoltage and mega-voltage radiographs. Int J Radiat Oncol Biol Phys 2000;47:825–839.

3. Mechalakos J, Hunt M, Lee M, Hong L, Ling C, Amols H. Using anonboard kilovoltage imager to measure setup deviation in intensity-modu-lated radiation therapy for head-and-neck patients. J Appl Clin Med Phys2007;8:28–44.

4. Kippenes H, Gavin PR, Sande RD, Rogers D, Sweet V. Compar-ison of the accuracy of positioning devices for radiation therapy of canineand feline head tumors. Vet Radiol Ultrasound 2000;41:371–376.

5. Kippenes H, Gavin PR, Sande RD, Rogers D, Sweet V. Accuracy ofpositioning the cervical spine for radiation therapy and the relationship toGTV, CTV and PTV. Vet Radiol Ultrasound 2003;44:714–719.

6. Bley CR, Blattmann H, Roos M, Sumova A, Kaser-Hotz B. Assess-ment of a radiotherapy patient immobilization device using single plane portradiographs and a remote computed tomography scanner. Vet RadiolUltrasound 2003;44:470–475.

7. Yoo S, Kim G, Hammoud R, et al. A quality assurance program forthe on-board imager. Med Phys 2006;33:4431–4447.

Fig. 5. Treatment position corrections vs. mass of patient. Also shown is frequency distribution of patient mass.

234 HARMON ET AL. 2009