a non-invasive immobilization system and related quality assurance

Int. J. Radiation Oncology Biol. Phys., Vol. 43, No. 2, pp. 455– 467, 1999
Copyright © 1999 Elsevier Science Inc.
Printed in the USA. All rights reserved
0360-3016/99/$–see front matter
PII S0360-3016(98)00398-8
PHYSICS CONTRIBUTION
A NON-INVASIVE IMMOBILIZATION SYSTEM AND RELATED QUALITY
ASSURANCE FOR DYNAMIC INTENSITY MODULATED RADIATION
THERAPY OF INTRACRANIAL AND HEAD AND NECK DISEASE
JEN-SAN TSAI, PH.D.,* Mark J. ENGLER, PH.D.,* Marilyn N. LING, M.D.,* JULIAN K. WU, M.D.,†
BRADLEY KRAMER, M.D.,* THOMAS DIPETRILLO, M.D.,* AND DAVID E. WAZER, M.D.*
Departments of *Radiation Oncology and †Neurosurgery, New England Medical Center Hospital and Tufts University School of
Medicine, 750 Washington Street, Boston, MA
Purpose: To develop and implement a non-invasive immobilization system guided by a dedicated quality
assurance (QA) program for dynamic intensity-modulated radiotherapy (IMRT) of intracranial and head and
neck disease, with IMRT delivered using the NOMOS Corporation’s Peacock System and MIMiC collimator.
Methods and Materials: Thermoplastic face masks are combined with cradle-shaped polyurethane foaming
agents and a dedicated quality assurance program to create a customized headholder system (CHS). Plastic
shrinkage was studied to understand its effect on immobilization. Fiducial points for computerized tomography
(CT) are obtained by placing multiple dabs of barium paste on mask surfaces at intersections of laser projections
used for patient positioning. Fiducial lines are drawn on the cradle along laser projections aligned with nasal
surfaces. Lateral CT topograms are annotated with a crosshair indicating the origin of the treatment planning
and delivery coordinate system, and with lines delineating the projections of superior-inferior field borders of the
linear accelerator’s secondary collimators, or with those of the fully open MIMiC. Port films exposed with and
without the MIMiC are compared to annotated topograms to measure positional variance (PV) in superiorinferior (SI), right-left (RL), and anterior posterior (AP) directions. MIMiC vane patterns superposed on port
films are applied to verify planned patterns. A 12-patient study of PV was performed by analyzing positions of
10 anatomic points on repeat CT topograms, plotting histograms of PV, and determining average PV.
Results and Discussion: A 1.5 6 0.3 mm SD shrinkage per 70 cm of thermoplastic was observed over 24 h.
Average PV of 1.0 6 0.8, 1.2 6 1.1, and 1.3 6 0.8 mm were measured in SI, AP, and RL directions, respectively.
Lateral port films exposed with and without the MIMiC showed PV of 0.2 6 1.3 and 0.8 6 2.2 mm in AP and
SI directions. Vane patterns superimposed on port films consistently verified the planned patterns.
Conclusion: The CHS provided adequately reproducible immobilization for dynamic IMRT, and may be
applicable to decrease PV for other cranial and head and neck external beam radiation therapy. © 1999
Elsevier Science Inc.
Non-invasive immobilization, Polyurethane cradle, Alpha cradle, Thermoplastic mask, Aquaplast, Conformal
radiotherapy, Intensity-modulated radiotherapy, IMRT, Topograms, Quality assurance.
The anatomic accuracy of radiotherapy depends on the
ability of a positioning system to reproduce patient geometry among simulation, computerized tomography (CT)
scanning, and treatment procedures. Geometric accuracy is
especially critical for the radiotherapeutic modalities of
intensity-modulated radiotherapy (IMRT) (1–7) and radiosurgery (SRS) (8 –13), where escalated or large single doses
may be delivered. The setup accuracy of the NOMOS
Corporation’s dynamic Peacock IMRT System (also called
Corvus Version 1.12; NOMOS, The IMRT Corp., Sewick-
ley, PA) is improved by invasive or non-invasive immobilization systems, with the former using screws surgically
placed in the cranium (1– 4, 8 –16), and the latter, external
devices (1, 6, 7, 17–26). Thus far, over 1500 patients with
cranial or head and neck disease immobilized non-invasively have been treated with the Peacock system (1, 6).
Rapid growth of these applications suggests detailed study
of the effectiveness of non-invasive patient immobilization
and positioning.
Thermoplastics (Aquaplast; WFR/Aquaplast Corp., Wyckoff, NJ) (23) and polyurethane cradles (Smithers Medical
Products, Inc., Akron, OH) are widely used as non-invasive
Reprint requests to: Jen-Sai Tsai, Ph.D., Department of Radiation Oncology #246, New England Medical Center, 750 Washington Street, Boston, MA 02111. E-mail: [email protected]
Presented at the 38th Annual Meeting of the American Association of Physicists in Medicine, Philadelphia, PA, July 22–26,
1996.
Current address of M.N.L.: Department of Radiation Oncology,
Washington University, Seattle, WA 98136.
Acknowledgments—The authors acknowledge extensive support
from the NOMOS Corporation. We thank Richard Behrman,
Ph.D., for useful discussions about CT imaging.
Accepted for publication 13 September 1998.
INTRODUCTION
455
456
I. J. Radiation Oncology
●
Biology
●
Physics
positioning devices. To increase immobilization reproducibility among simulator, CT, and treatment procedures,
these two devices are applied here simultaneously to create
a “customized headholder” (CH) (Fig. 1a). The use of a CH
has already been described with conventional external beam
radiation (25, 26).
In our clinic, it has been observed that the face mask
shrinks after fabrication, and is flexible during applications.
Hence, mask shrinkage and flexibility is studied in an attempt to minimize its effect on immobilization reproducibility. Studies of immobilization systems and positional
reproducibility are affected by immobility of the patient’s
anatomy of interest (AOI), ability to locate AOIs accurately,
and adaptability of the immobilization technique to a specific radiotherapeutic modality. In this investigation, the CH
is augmented and tested with dedicated quality assurance
(QA) to create a CH system (CHS) in an attempt to optimize
positional reproducibility when delivering dynamic IMRT.
METHODS AND MATERIALS
Coordinate and fiducial systems
Coordinate and fiducial systems are generally applied to
implement non-invasive immobilization with the IMRT system, and specifically, for a 12-patient study of immobilization reproducibility. The treatment planning and delivery
coordinate system axes are denoted by (X, Y, Z), and ideally
coincide with the RL (right-left), AP (anterior-posterior),
and SI (superior-inferior) anatomic directions. Any of the
three directions are denoted by “r.” Thus, coordinate pairs
of points on two-dimensional (2D) simulator radiographs,
CT scans, CT topograms (or “scout” images), and megavoltage port films in orthogonal transverse, coronal, and
lateral planes are specified as (AP, RL), (RL, SI), and (AP,
SI), respectively. The linear accelerator (linac) gantry axis
(G), ideally identical to the Z axis of the Peacock system, is
also ideally coincident with the SI axis, and intercepts the
treatment planning and delivery coordinate system origin
(O). Anatomic fiducial points applied in the CHS include
anterior (A), right and left lateral (R and L), cranial vertex
(V), inferior edge of the nasal septum (IENS), and posterior
edges of the tragi (ET). Primary fiducial lines and planar
curves include the transverse fiducial line (TFL) in a transverse plane (“tra” view), sagittal (SFL) in a sagittal plane
(“lat” view), and coronal (CFL) in a coronal plane (“cor” or
AP view). Secondary fiducial curves include the sagittal
midline of the nasal surface (SMNS), the mid-curve of the
eyebrows (EB), the abutment of the eyelids (EL), the abutment of the lips (AL), and the external curve of the ears
(EA).
Second and third coordinate systems on the CH baseplate
are dedicated to coronal and lateral CT topograms applied in
a 12-patient study of patient immobilization reproducibility.
Metrics for immobilization reproducibility
A metric for immobilization, reproducibility is specified
as positional variance (PV) of r, PV(r), with ranges signifying one standard deviation (SD). Radiographically deter-
Volume 43, Number 2, 1999
mined average absolute values of PVs are denoted by
,Drcor. and ,Drlat., for example, where Dr signifies
variance in (RL, SI), or (AP, SI) directions. Absolute values
of PV are plotted in histograms and their average values
,Dr. are tabulated. Values Dr averaged over both CT
(coronal and lateral) topograms and port films are termed
global positional variance Dgr to characterize the CHS.
Detection of setup errors in topographic and port film
studies
To analyze PV and model patient setup, three translational degrees of freedom were recognized relative to the X,
Y, and Z coordinate axes. Three rotational degrees of freedom were recognized relative to axes coincident with or
parallel to the X, Y, and Z axes. Thus, a total of 6 degrees
of freedom were extractable from the combination of a 2D
transverse CT image with coronal and lateral topograms.
Each image may be used to detect two translational and one
rotational parameter. Thus, three orthogonal images, each
containing three parameters, provide up to nine parameters
for characterizing 6 degrees of freedom, and one of the
images may be omitted. The transverse CT was omitted in
the 12-patient study of PV.
Although AP (coronal) port films were desired to capture
possible PV in the RL dimension, it was impractical to
insert cassettes directly under the patient. Anatomy surrounding the sagittal midplane would have been obscured in
films exposed below the table by the steel spline supporting
the table. A posterior span of up to 70° was omitted from
treatment planning and delivery to avoid irradiating through
the center spline. Nevertheless, rigorous QA is implemented
prior to treatment to make sure of the coincidence of the
SFL with the sagittal laser, thus minimizing possible PV in
the RL direction.
CT topograms and port films of an anthropomorphic
phantom (Alderson Rando Phantom; Radiology Support
Devices Inc., Long Beach, CA) were obtained to check the
consistency of the 2D scale of the topogram and the magnification factor of the port film for their mutual PV verification. Dimensions on the topogram and demagnified dimensions on port films were compared with actual
dimensions of phantom structure.
Posterior polyurethane cradle
The cradle is fabricated in the simulator room with polyurethane foaming agents and a 16.6 3 25.4 3 7.6 cm
styrofoam box lacking anterior and inferior sides. The container is placed inside a plastic bag with the open end
pointing in the patient’s inferior direction. The bag and
container are pressed tightly into the U-shaped bay of the
CH baseplate. Foaming agents are mixed and spread evenly
along the inner surfaces of the bag inside and inferior to the
styrofoam box. The patient’s head, with IENS approximately perpendicular to the table top, is lowered onto the
bag into the container. After the foaming agents expand and
harden, the cradle surface is trimmed to allow a U-frame
containing softened thermoplastic to be drawn downward
along the sides of the patient’s face. A 15 3 20 3 0.08 cm
A non-invasive immobilization system
● J.-S. TSAI et al.
Fig. 1. Alignment of the cradle. (a) Anthropomorphic phantom head resting on its cradle and before mask placement.
(b) Transverse view of the SFL and laser light projection on the superior external surface of the cradle head holder
aligned with the SMNS (left) and lateral view of the TFL and laser light projection on the lateral external surface of the
cradle container aligned with the IENS (right). (c) Top view of the TFL and SFL fiducials and laser light projections
aligned with relative to the cranium. (d) Transverse view of misalignment of the SFL with the SMNS, creating the
variance DRLtra. (e) Lateral view of possible misalignment of the TFL with the SMNS, creating the variance DSIlat (left),
and possibly undetected misalignment cause by cranial rotation in the sagittal plane (right). (f) Top view of correct
alignment; rotation in the coronal plane is avoided by mechanical resistance.
457
458
I. J. Radiation Oncology
●
Biology
●
Physics
piece of rubber gauze is placed on the cradle to compensate
for anticipated mask shrinkage. The SFL is drawn on the
cradle along the sagittal ceiling laser projection aligned with
the SMNS. The table is adjusted so that the vertical side
wall lasers are tangential to the IENS (Fig. 1b). The right
and left sides of the TFL are drawn on the lateral cradle
surfaces along the projections of the side wall lasers. If the
dose distribution may be improved with the head in the
flexed position, the baseplate is set so that the ETs are
aligned to intercept the intersection points of the side wall
laser projections. The three fiducial lines (TFL, CFL, and
TFL) are applied to reposition the patient’s head at the
original cradle fabrication position during CT and treatment
procedures (Fig. 1c).
Anterior thermoplastic face mask
To reduce the beam attenuation by the baseplate, it was
modified by cutting several parallel slots in the SI direction.
The modified baseplate and the attached cradle are placed
on the table top so that the room lasers match the TFL, CFL,
and SFL on the cradle surface (Fig. 1c). The patient is
positioned on the cradle such that IENS and SMNS are
aligned with the TFL and SFL, respectively. For the flexed
head setup, the table is adjusted so that the inferiormost
edges of the ear lobes or posteriormost ET coincide with
horizontal side wall lasers.
The U-frame is held with two hands so that the sheet is
coplanar to the table top and its geometric center is directly
above the patient’s nose, and is then lowered over the
patient’s face until reaching the baseplate, where it is fixed
in position with four plastic clamps, to the extent of what a
patient can reasonably tolerate. The thermoplastic is molded
around the chin, the bridge of the nose, and the supra-orbital
ridges. Despite the tightness of the mask, the patient’s
cranium may move because of the deformation of soft tissue
pressed against the mask. To minimize PV of the patient
relative to the mask, underlying EB, EL, ML, and EA
curves are traced on the outer surface of the mask with a
fine-tipped felt pen, with the intention of matching the
tracings to the anatomy during subsequent mask placements. After the mask hardens, a hole is cut around the nasal
area. Sharp edges of the nose hole in the mask are taped to
avoid skin abrasion. The nose hole is necessary for visualization of lasers along the IENS and SMNS. In addition, the
hole allows for increased patient tolerance of the tightness
of the mask.
Mask shrinkage
To study mask shrinkage, four routinely used thermoplastic mesh sheets, 23 3 30.5 cm in size, were heated in a 66°C
water bath and stretched to 70 cm, an average stretched
length in applications around the anterior and lateral head.
Length was then measured as a function of time. In addition,
three masks shaped to the Rando head were used to study
shrinkage by measuring the height from the U-frame to the
nasal tip of the mask as a function of time.
Mask shrinkage after fabrication for patients is compen-
Volume 43, Number 2, 1999
sated for by insertion of the rubber gauze during fabrication
and omission of the gauze during CT and treatments. To
determine the effect on PV of neglecting mask shrinkage,
some topograms were obtained with the same gauze used
during mask fabrication left in the cradle during CT.
QA procedures common to the simulator, CT scanner,
and linear accelerator
Stability of lasers and the isocenter of the treatment
machine are critical to the integrity of patient positioning
and treatment. Thus, laser alignments in the simulator, CT
scanner, and IMRT linac are inspected regularly and adjusted when necessary. Additional QA helps minimize linac
isocenter motion (1, 9, 28 –30). As part of our radiosurgery
program, the location of the three major rotational axes of
the linac are tested with a 3D mechanical pointer and a
3-axis micrometer for their mechanical stability (28) and
with multiple film exposures of a target simulator (9, 12,
29 –31). Table tops for simulator, CT, and linac procedures
are inspected with a level and adjusted when necessary.
To maximize positional reproducibility of the patient’s
head and the outer soft tissue anatomy relative to the mask,
a best coincidence of the mask tracings with underlying
anatomy is sought. If no tracings match their underlying
anatomy, or more than half the tracings deviate by more
than 3 mm from underlying anatomy, the mask is removed
and reattached until the 3 mm criterion is met. To maximize
positional reproducibility of the CH among CT and treatments, , 1 mm tolerance is allowed for coincidence of the
TFL, CFL, SFL, IENS, and SMNS fiducials with laser lines,
and for coincidence of laser line intersections with fiducial
points (A, R, L, V).
Peacock IMRT gantry axis and isocenter placement
The IMRT system has been described in detail elsewhere
(1–7). However, certain features of the CH were developed
to overcome limitations of isocenter placement during treatment planning. The isocenter is set anatomically in a software module termed “Image Registration.” For non-invasively immobilized patients, the isocenter is digitized on a
CT image manually without tools to display the relationship
of the isocenter coordinates to external fiducial points
needed for patient setup in CT and treatment. The isocenter
is placed: (a) so as to prevent collision of the MIMiC with
the table or patient, as the MIMiC surface proximal to the
patient is only 36 cm from the gantry axis; (b) at a point less
than 10 cm, the MIMiC field half width, from the most
distal target, meaning that the IMRT beam can reach and
cover the most distal target; and (c) within the target to
avoid suboptimal dose distributions that may arise from
underdosing at matchplanes of adjacent arced fields (5).
After image registration is approved, the user moves on to
delineate AOIs in the “Anatomy” software module, to prescribe dose in the “Prescription” module, and to examine
and approve plans in the “Display Results” module. If the
isocenter placement allows collisions, or is suboptimal for
treatment planning, it is necessary to revoke all of the
A non-invasive immobilization system
● J.-S. TSAI et al.
459
information entered and reviewed in all of the software
modules from Image Registration through Display Results,
thereby wasting considerable time and effort (more recent
software allows changing isocenter in “Prescription”). Thus,
two sets of 3– 4 additional fiducial points are applied for CT
scanning so that the above described software limitations
are eliminated. These additional fiducials are specified as
(R, L, V)i, where i 5 1, 2, 3, or 4. Fiducial points and planar
curves are drawn on pieces of nylon tape affixed to the
mask.
Patient setup for CT scanning
The patient’s head is adjusted in the cradle such that the
SMNS is aligned with the SFL and the sagittal laser projections and simultaneously the IENS is tangential to both
side wall lasers and is aligned with the TFL as it was during
cradle fabrication. These adjustments are intended to insure
that head position replicates the original setup in the simulator. Under such circumstances, a possible rotation of the
head around the sagittal axis of the body is corrected most
easily by detection in a transverse view (Fig. 1d). A possible
translation of the head in the SI direction relative to its
original position in the cradle is corrected most easily by
detection in a lateral view (Fig. 1e). The only possible
remaining misalignment is shown in Fig. 1d, and is subsequently detected by comparing a lateral topogram with a
lateral port film. Because the cradle conforms to the posterior and lateral cranial surfaces, and the SFL is aligned with
SMNS, PVs in the RL and SI directions are unlikely, as
illustrated in Fig. 1f. Once the head is positioned in the
cradle as in the original setup in the simulator, the mask is
attached and the EB, EL, AL, and EA curves drawn on the
mask are verified to match the corresponding anatomic
structures.
Prior to CT scanning, dabs of barium paste are placed on
the mask fiducial points A and (R, L, V)i for subsequent
placement of the origin O which automatically defines the
axis G at X 5 Y 5 0 (Fig. 2a). To insure that at least one
transverse CT image shows all fiducial points A and (R, L)i,
about five scans, 1-mm thick, are obtained in the region of
the transverse plane defined by points A and (R, L)i.
Treatment planning
The origin O is assigned along a line RLi at its intersection with a line dropped from point A and perpendicular to
RLi. The origin satisfies the following criteria arising from
system limitations described above: (a) Potential collision is
avoided (Fig. 2b); (b) the target most distal from G is within
the MIMiC field half width of 10 cm; and (c) G is within the
target to avoid underdose at matchplanes of unopposed
anterior spans of adjacent arced beam segments.
Twelve-patient study of PV using CT topograms
To study PV, 2–5 CT topograms were repeated for each
of 12 patients over the course of 1 h, and of 4 patients over
the course of more than 3 days for assessing the influence of
Fig. 2. (a) Transverse, sagittal, and coronal views of the CH with
fiducial points A and (R, L, V)i marked with barium sulfate paste
for CT imaging. (b) Patient setup relative to Peacock IMRT
hardware, highlighting the IMRT field projected through the gantry rotational axis (G), the gantry’s range of rotation, and possible
collision between the patient and the MIMiC assembly.
prolonged mask shrinkage. Anatomic points Qi and Qi9,
where i 5 1, 2, . . . , 5, are digitized on topograms. For each
patient, the internal points were selected as those with the
most unique radiographic appearance within their anatomy.
The origins of the topogram coordinates are points on the
baseplate of the CH, and are designated as Q0 and Q09. For
clarity of display in the figures, subscripts of points are
displayed as normally sized next to their symbol Q or Q9.
Internal anatomic fiducial points on the coronal topogram
are in the sphenoparietal suture (Q1), frontozygomatic suture (Q2), supra-orbital margin (Q3), and zygomaticomaxillary sutures (Q4, Q5) as shown in Fig. 3a. Fiducial points on
the lateral topogram are in the frontal sinus (Q19), anterior
sella (Q29), anterior nasal spine (Q39), coronoid process of
the mandible (Q49), and mastoid process (Q59) (Fig. 3b).
These points were chosen for ease and uniqueness of topographic identification. In the figures, subscripts of points are
displayed next to their symbol for clarity. When points were
not uniquely identified on a topogram, they were not used in
the statistical analysis of PV.
Coordinate pairs (RL, SI) and (AP, SI) of points Q0, Qi,
Q09, and Qi9 from coronal and lateral topograms were dig-
460
I. J. Radiation Oncology
●
Biology
●
Physics
Volume 43, Number 2, 1999
Dri, j, cor 5 max ?Qi, j, cor(r) 2 Qi, j, cor(r)?, and
(1a)
Dri, j, lat 5 max ?Qi, j, lat9(r) 2 Qi, j, lat9(r)?,
(1b)
Variances were determined from topograms obtained
in , 1 h and . 3 days of CT scan interval, and were plotted
in histograms. Five-point variances were sought per topogram set per patient. The variance measurement was slightly
increased by a reading and digitization uncertainty of 0.5
mm inherent to CT pixel size. Average variance was specified as:
,Drcor. 5
O O f(r
) Dri, j, cor /
O O f(r
) Dri, j, lat /
i, j, cor
i
and
,Drlat. 5
j
i
i, j, lat
i
j
O O f(r
i, j, cor
)
(2a)
),
(2b)
j
O O f(r
i, j, lat
i
j
where f(ri, j) is the frequency of variance Dri, j. Variance
was also plotted in histograms.
Fig. 3. (a) Coronal topogram, coordinate system Q0, and five
anatomic points, Q1–Q5. (b) Lateral topogram with points Q09–
Q59.
itized at the CT console. Dri, j was defined as the maximum
variance of either coordinate of point Qi, or Qi9 among
repeat topograms of patient j:
Verification of treatment setup by comparing the IMRT
port film to the CT topogram
Lateral port film verification prior to IMRT is obtained to
monitor the consistency between CT and IMRT setup. Two
methods of obtaining port films were evaluated. In both, the
gantry is set horizontally at an angle of 90° or 270° and the
table, at a treatment position z. With the first method, a port
film is exposed to 3 MU of radiation with the MIMiC fully
open to a field size of 20 3 3.4 cm. Without disturbing the
patient setup, the MIMiC assembly is removed and the film
is exposed to an additional 3 MU with a field size of 30 3
30 cm. In the event that the craniocaudal edges of the 20 3
3.4 cm strip fail to reveal unique bony anatomy near the
target, the same port film is exposed to a third exposure with
the secondary collimators set to 30 cm wide 3 independent
jaw 5 or 10 cm long.
Removing the MIMiC posed a risk to the patient position
and was thought to be overly time consuming for routine
use. Therefore, a second method is employed using three
adjacent 20 3 3.4 cm MIMiC fields. The resulting film is
then compared to the lateral topogram (Fig. 4). A special
cassette holder was mounted on the side rails of the table so
that the port film could move with the patient and capture
three strips of anatomy irradiated by fully open MIMiC
fields. The cassette is fixed at a known distance (SFD),
usually 127 cm, from the radiation source, projecting a
4.3 cm field length to the film. Of the three exposures, the
middle one is taken at a central table position identical to
that of a treatment to insure imaging a substantial portion of
the target. The two neighboring exposures are made at table
positions 4.3 cm from the central position, for an SFD of
127 cm. The 4.3 cm shift avoids overexposure by overlap of
adjacent divergent field edges at the film. The total length of
the three-exposure film is then back projected 10.2 cm to the
G axis, enough to capture substantial AOI including the
A non-invasive immobilization system
● J.-S. TSAI et al.
461
Fig. 5. Shrinkage of thermoplastic sheets as a function of time.
Fig. 4. Port film obtained with three fully opened MIMiC fields
matched at the film, where 4.3 cm is the port film table index
interval, and 3.4 cm is the treatment table position interval: anatomy exposed at the gantry axis (upper left), coronal view of three
diverging MIMiC beams (upper right), and reconstruction without
unirradiated anatomy (lower left).
target. Due to the desired matching of the three strip fields
at the film cassette, two segments of about 0.9 cm of
anatomy at the axis G between the adjacent diverging beams
are missing from the image (Fig. 4). By accurately locating
the edges of the three exposures relative to the CT coordinate origin, a verification of the IMRT setup is obtained. If
a discrepancy of anatomy on the topogram and port film
exceeds 2 mm, another port film is exposed and the setup is
adjusted until the discrepancy is under 2 mm.
To image and verify the MIMiC vane pattern at horizontal gantry positions, the MIMiC controller is programmed to
deliver a narrow arc and open the MIMiC to a patientspecific vane pattern over a span of , 2° including the
horizontal gantry position. The MIMiC remains shut during
the first 6 –12° of rotation so that stability of gantry rotational speed, as monitored by clinometers on the controller,
satisfies programmed criteria.
RESULTS AND DISCUSSION
Shrinkage of the mask
Shrinkage of thermoplastic up to 95 h after initial elongation is shown in Fig. 5. Shrinkage of 1.5 6 0.3 mm SD in
1 day was determined from all four sheets. Shrinkage of
0.9 6 0.2 mm in height was observed in the masks fabricated for the Rando phantom. After the first day, a maximum of 0.5 mm additional shrinkage was observed in the
ensuing 3 days, after which no shrinkage was observed.
Because shrinkage is most significant during the first day, at
least 1 day is required between mask fabrication and the
treatment planning CT.
Patient setup reproducibility from CT topograms
Figures 6a and 6b show PV histograms taken from measurements of the coronal and lateral CT topograms of five
patients immobilized with the same gauze as that used for
cradle fabrication, implying that the mask shrank before the
CT scans were obtained. Using Eq. 2a, data from Fig. 6a
yield ,DSIcor. and ,DRLcor. of 1.2 6 1.0, and 1.6 6 1.2
mm, respectively. Similarly, from Eq. 2b and Fig. 6b,
,DSIlat. and ,DAPlat. are 1.4 6 1.2, and 1.3 6 1.3 mm
respectively. Figures 6c and 6d show histograms from
twelve patients’ coronal and lateral topograms scanned
within 1 h, without gauze, and at least 1 day after mask
fabrication with gauze, for example, with mask shrinkage
offset by removal of the gauze. From Eq. 2a and Fig. 6c,
,DSIcor. and ,DRLcor. are 1.1 6 0.9 and 0.8 6 0.7 mm,
respectively. From Eq. 2b and Fig. 6d, ,DSIlat. and
,DAPlat. are 1.2 6 1.3 and 1.2 6 1.4 mm, respectively.
The agreement of the average discrepancy of ,DSIcor.
of 1.1 6 0.9 mm for Fig. 6c with the average discrepancy of
,SIlat. of 1.2 6 1.3 mm for Fig. 6d suggests that the
lateral port film is adequate at this time for IMRT setup
verification without consuming additional time and resources on a coronal port film. The maximum deviation of
5 mm occurred with a patient having long hair, suggesting
that positioning of the hair may be important in the use of
the CH. Another source of error in this study is attributed to
deficient visualization of internal anatomic points Qi and Qi9
in the attempt to identify the same point on successive
topograms.
Results with the CHS also indicate that the variance seen
on the lateral topogram is slightly larger than that seen on
the coronal topogram. This may be attributed to setting up
the coronal alignment first, thus making the lateral alignment secondary. In other words, the coronal alignment
accuracy may be viewed somewhat as an asymptote of the
system’s overall experimental error, i.e. no worse than lateral or transverse view alignments. This served as another
rationalization for omitting the coronal port film with negligible deterioration of the global QA.
462
I. J. Radiation Oncology
●
Biology
●
Physics
Volume 43, Number 2, 1999
Fig. 6. Variance histograms: (a) ,DSIcor. and ,DRLcor. from coronal topograms obtained , 1 h from mask
fabrication with gauze in place; (b) ,DSIlat. and ,DAPlat. from lateral topograms obtained , 1 h from mask
fabrication with gauze in place; (c) ,DSIcor. and ,DRLcor. from coronal topograms obtained more than 1 day from
mask fabrication; (d) ,DSIlat. and ,DAPlat. from lateral topograms obtained more than 1 day from mask fabrication.
A non-invasive immobilization system
● J.-S. TSAI et al.
463
Fig. 7. (a) Lateral topogram with three lines at predetermined
treatment table coordinates. (b) Port film exposed firstly with the
fully open MIMiC field of 20 3 3.4 cm, secondly with the MIMiC
removed and the field size slightly increased with an independent
jaw, and thirdly with a field size of 30 3 30 cm. (c) Vane patterns
and relative pencil beam intensities as a function of 10° bins of
gantry rotation near the horizontal gantry position. Percent black
shows the fraction of time during which the vane is open over a
10° span of gantry rotation (top); three adjacent port film exposures pieced together show anatomy relative to the edges (white
lines) of adjacent fields; a vane pattern at the horizontal gantry
position is superimposed on the central of the three exposures. (d)
The CT topogram with pre-calculated location of the Z coordinates
at the edges of anatomy in the image plane; two areas labeled with
asterisks are expected missing image plane anatomy. (e) Lateral
view of anatomy, with isodose lines showing the target.
464
I. J. Radiation Oncology
●
Biology
●
Physics
Volume 43, Number 2, 1999
Lateral topograms obtained over an interval of 3 days for
four patients showed ,DSIlat. of 1.0 6 0.6 mm, and
,DAPlat. of 1.7 6 1.0 mm. The former agrees with value
1.2 6 1.3 mm measured within one CT session, and the
latter is slightly larger than the value 1.2 6 1.4 mm measured within one CT session. This difference, still within the
reading and digitization error of 0.5 mm, and the agreement
of ,DSIlat. between 1.0 6 0.6 mm over 3 days interval
and (1.2 6 1.3, 1.4 6 1.2 mm) in one CT session support
the QA protocol for mask shrinkage and patient setup, for
example, mask shrinkage has little effect on the positional
variance in the SI direction and therefore accounts for small
DSIlat, but does affect AP variance due to changes of the
mask vertical dimension manifested in DAPlat.
In the PV study, CT topograms were used in place of
simulator films because: (a) the IMRT treatment planning
and delivery are based on the CT images; (b) CT topograms
provide high quality images and thus a high degree of
spatial resolution; and (c) CT topograms in computer storage are in numerical forms (coordinates), thus facilitating
the analysis.
Port film verification of patient anatomy imaged in the CT
topogram
Anatomic dimensions obtained from topograms and demagnified port films agreed with the actual dimensions of
the anthropomorphic phantom head and neck within 6 1%
SD. Thus, the combined uncertainty in comparing topograms with port films is estimated to be 6 1.4%.
The lateral topogram with three delineator lines at coordinates pre-calculated for the sake of port filming is shown
in Fig. 7a. The three exposure port film of the anatomy of
Fig. 7a together with surrounding anatomy revealed by
exposure of a larger field without the MIMiC on the linac,
is shown in Fig. 7b. Figures 7c and 7d show three consecutive images with the MIMiC fully opened and port images
pieced together. Two narrow strips of missing anatomy are
indicated with asterisks in Fig. 7d, as shown schematically
in Fig. 4. The location of specific anatomic structure from
these lateral topograms and port films are compared with
ruler measurements to the delineator lines and field edges.
Figure 7c also reveals a radiation vane pattern at the
anatomic location relative to the tumor shown in Fig. 7e.
This example illustrates a vane pattern corresponding to the
beam at a horizontal gantry angle with expected patterns
shown in the insert of Fig. 7c. For some patients, horizontal
vane pattern printouts may not be available. In these cases,
the vane pattern is interpolated from printouts of patterns at
gantry positions 6 5° from the horizontal gantry position.
From Figs. 7c, 7d, and 7e, the patient’s setup between CT
and IMRT and the vane pattern’s agreement with IMRT
planned vane pattern, as well as its radiation coverage
relative to the tumor, are simultaneously verified.
Figure 8 shows the discrepancy histograms from a sample
of 28 patients’ lateral port films compared with their lateral
topograms in both the SI and AP directions. The SI and AP
discrepancy histograms yield average discrepancies of ?0.2?
Fig. 8. Histograms of the SI (top) and AP (bottom) variance of
lateral port film relative to topogram.
6 1.3 , and ?0.8? 6 2.2 mm, respectively. Discrepancies
larger than 3.5 mm may arise from negligence in alignment
of the sagittal and side wall lasers with the IENS and SMNS
prior to mask attachment to the patient’s face. Occasional
discrepancies of 2.0 mm at the target are tolerated for
fractionated IMRT, and rare discrepancies . 2.0 mm have
been seen, in which case the setup is adjusted until a port
film fulfilling the 2 mm criterion is obtained. For example,
two patients were immobilized using face masks fabricated
for conventional therapy in conjunction with a standard
plastic head cup. No fiducial lines relative to the nasal
septum and lasers were drawn on the plastic head cup; nor
were eyebrows, eyelids, or mouth lip markings on the mask
with the patient’s. As a result, discrepancies . 2 mm
between the CT topogram and IMRT port film were observed. For these cases, new CHs were made, and patients
were re-planned.
Setup variance assessed from topograms and port films
are given in Table 1. Craniocaudally, the overall discrepancy is 1.0 6 0.8 mm, based on the calculation sketched in
Fig. 9, where ,DgSI. 5 1.0 6 0.8 mm SD is the global
mean discrepancy in the SI direction, i.e., the sum of Eq. 2
A non-invasive immobilization system
● J.-S. TSAI et al.
465
Table 1. Positional variance in millimeter of the CHS
determined from topograms
1/2
Duration Gauze* ,DSIlat. ,DAPlat. ,DSIcor. ,DRLcor.
,1 h
,1 h
. 3 days
Port films
Average
variance
1
2
2
N/A
1.4 6 1.2
1.2 6 1.3
1.0 6 0.6
0.2 6 1.3
1.3 6 1.3
1.2 6 1.4
1.7 6 1.0
0.8 6 2.2
1.2 6 1.0
1.1 6 0.9
–
–
1.6 6 1.2
0.8 6 0.7
–
–
1.0 6 1.2 1.3 6 1.5
1.2 6 1.0
1.2 6 1.2
*1Gauze: mask shrank in the intervening time prior to implementation. 2Gauze: masks’ shrinkage is offset during implementation.
over all topograms. Globally averaged ,DgAP. and
,DgRL. were 1.2 6 1.1, and 1.3 6 0.8 mm, respectively.
Analogous values reported from other facilities for noninvasive immobilization are 2.0 mm from Thornton et al.
(32), 2.2 6 1.4 mm from Verhey et al. (23), and 1.02 mm
from Sweeney et al. (24) using a special vacuum dental cast
in conjunction with a hydraulic arm.
Advantages and disadvantages of non-invasive relative to
invasive immobilization
To optimize immobilization for IMRT, a comparison
between the invasive and non-invasive immobilization setups is useful. The non-invasive CHS has the following
advantages over invasive skull screw immobilization:
1. Avoids the cost and inconvenience of surgical procedure.
2. Allows easy airway access and emergency release with
less risk of injury.
3. Avoids possible medical complications, such as infection
around head screws.
4. May be used for treatment at a number of anatomic sites.
5. Reduces patient discomfort.
Disadvantages are:
1. Slightly more PV than PV of rigid headscrew immobilization (2 mm vs. 1 mm) (1).
2. The CHS assumes a constant patient geometry during the
planning CT and over the course of therapy. The mask
shrinkage (Figs. 5, 6a– d) study demonstrated at least 1
day was needed before the CT. In one case, because of
facial swelling in response to high dose chemotherapy,
the mask would not fit on a patient, and a new mask, CT,
and treatment plan had to be obtained.
Implementing the present non-invasive immobilization
system in conjunction with our alignment QA procedures
lessens the need for additional devices such as biteblocks
(20, 23, 32), vacuum dental cast (24), or earplugs for reinforcing immobilization reproducibility.
QA for placement of the gantry axis and origin in the
treatment planning and delivery systems
The results and consequence of QA in origin O assignment relative to the target during the IMRT treatment plan-
Fig. 9. Illustration of the calculation of the mean discrepancy in
craniocaudal (SI) direction from Table 1. Sample ranges of the
absolute values of individual variances include negative values
which are a byproduct of the estimated measurement error (top).
,DgSI. is the average of the absolute values of variance in the SI
direction and sSI is one standard deviation (bottom).
ning is shown in the Appendix. The use of the multiple
barium dots (Fig. 2a) has allowed us to achieve optimal
positioning of O/G relative to target.
CONCLUSION
A customized styrofoam/polyurethane cradle combined
with a thermoplastic mask and guided by a carefully designed QA procedure, termed the customized headholder
system (CHS), has been reliably implemented for noninvasive immobilization of dynamic IMRT of intracranial,
head and neck, and esophageal targets. The present noninvasive immobilization assures safe, easy, and reproducible immobilization to within a standard deviation of 2 mm.
The CHS need not be limited to dynamic IMRT, and may be
applicable to other external beam radiation therapy.
466
I. J. Radiation Oncology
●
Biology
●
Physics
Volume 43, Number 2, 1999
REFERENCES
1. Tsai J-S, Wazer DE, Ling MN, et al. Dosimetric verification
of the dynamic intensity-modulated radiation therapy of 92
patients. Int J Radiat Oncol Biol Phys 1998; 40:1217–1230.
2. Carol MP. Conformal radiosurgery: Stereotactic surgery and
radiosurgery. In: Purdy JA, Emami B, editors. 3D Radiation
Treatment Planning and Conformal Therapy. Madison: Medical Physics Publishing; 1993. p. 249–266.
3. Carol MP. Integrated 3-D conformal multivane intensity modulation delivery system for radiotherapy. In: Hounsell AR,
editor. Proceedings of the Eleventh International Conference
on the Use of Computers in Radiation Therapy. Stockport,
Chesire, UK: Handley Printers, Ltd.; 1994. p. 172–173.
4. Carol MP. Peacock: A system for planning and rotational
delivery of intensity-modulated fields. Int J Imag Syst Tech
1995; 6:56–61.
5. Carol MP, Grant WH, Bleier AR, et al. The field-matching
problem as it applies to the Peacock three-dimensional conformal system for intensity modulation. Int J Radiat Oncol
Biol Phys 1996; 34:183–187.
6. Verellen D, Linthout N, Van Den Gerge D, et al. Initial
experience with intensity-modulated conformal radiation therapy for treatment of the head and neck region. Int J Radiat
Oncol Biol Phys 1997; 39:99–114.
7. Engler MJ, Tsai J-S, Vora S, et al. Application of dynamic
intensity modulated radiation therapy to the treatment of prostate cancer. (Abstr.) Int J Radiat Oncol Biol Phys 1997; 39
(Suppl.):193.
8. Colombo F, Beneditti A, Pozza F, et al. External stereotactic
irradiation by linear accelerator. Neurosurgery 1985; 16:154–
160.
9. Engler MJ, Curran BH, Tsai J-S, et al. Fine tuning of linear
accelerator accessories for stereotactic radiotherapy. Int J Radiat Oncol Biol Phys 1994; 28:1001–1008.
10. Hartmann GH, Schlegel W, Sturm V, et al. Cerebral radiation
surgery using moving field irradiation at a linear accelerator
facility. Int J Radiat Oncol Biol Phys 1985; 11:1185–1192.
11. Houdek PV, Schwade JG, Serago CF, et al. Computer controlled stereotactic radiotherapy system. Int J Radiat Oncol
Biol Phys 1992; 22:175–180.
12. Lutz WR, Winston KR, Maleki N. A system for stereotactic
radiosurgery with a linear accelerator. Int J Radiat Oncol Biol
Phys 1987; 14:373–381.
13. Podgorsak EB, Olivier A, Pla M, et al. Physical aspects of
dynamic stereotactic radiosurgery. Appl Neurophysiol 1987;
50:263–268.
14. Gall K, Verhey L, Wagner M. Computer-assisted positioning
of radiotherapy patients using implanted radiopaque fiducials.
Med Phys 1993; 20:1153–1159.
15. Jones D, Christopherson DA, Washington JT, et al. A frameless method for stereotactic radiotherapy. Br J Radiol 1993;
66:1142–1150.
16. Podgorsak EB, Souhami L, Caron J-L, et al. A technique for
fractioned stereotactic radiotherapy in the treatment of intracranial tumors. Int J Radiat Oncol Biol Phys 1993; 27:1225–
1230.
17. Delannes M, Daly NJ, Bonnet J, et al. Fractioned radiotherapy
of small inoperable lesions of the brain using a noninvasive
stereotactic frame. Int J Radiat Oncol Biol Phys 1991; 21:
749–755.
18. Gademann G, Schlegel W, Debus J, et al. Fractioned stereotactically guided radiotherapy of head and neck tumors: A
report on clinically use of a new system in 195 cases. Radiother Oncol 1993; 29:205–213.
19. Gill SS, Thomas DGT, Warrington AP, et al. Relocatable
frame for stereotactic external beam radiatherapy. Int J Radiat
Oncol Biol Phys 1991; 20:599–603.
20. Hamilton RJ, Kuchnir FT, Pelizzari CA, et al. Repositioning
accuracy of a noninvasive head fixation system for stereotactic
radiotherapy. Med Phys 1996; 23:1909–1917.
21. Hariz MI, Henricksson R, Löfroth P-O, et al. A noninvasive
method for fractionated stereotactic irradiation of brain tumors
with linear accelerator. Radiother Oncol 1990; 17:57–72.
22. Kooy HM, Dunbar SF, Tarbell NJ, et al. Adaption and verification of the relocatable Gill-Thomas-Cosman frame in stereotactic radiotherapy. Int J Radiat Oncol Biol Phys 1994;
30:685–691.
23. Verhey LJ, Goitein M, McNulty P, et al. Precise positioning of
patients for radiation therapy. Int J Radiat Oncol Biol Phys
1982; 8:289–294.
24. Sweeney R, Bale R, Vogele M, et al. Repositioning accuracy:
Comparison of a noninvasive head holder with thermoplastic
mask for fractionated radiotherapy and a case report. Int J
Radiat Oncol Biol Phys 1998; 41:475–483.
25. Bentel GC, Marks LB, Sherouse GW, et al. A customized
head and neck support system. Int J Radiat Oncol Biol Phys
1995; 32:245–248.
26. Bentel GC, Marks LB. Comparison of two head and neck
immobilization systems. Int J Radiat Oncol Biol Phys 1997;
38:867–873.
27. Sherouse GW, Bourland JD, Reynold K, et al. Virtual simulation in the clinical setting: Some practical considerations. Int
J Radiat Oncol Biol Phys 1990; 19:1059–1065.
28. Tsai J-S, Curran B, Sternick E, et al. The measurement of
linear accelerator isocenter motion using a three micrometer
device and an adjustable pointer. Int J Radiat Oncol Biol Phys
1996; 34:189–195.
29. Tsai J-S, Buck BA, Svensson GK, et al. Quality assurance in
stereotactic radiosurgery using a standard linear accelerator.
Int J Radiat Oncol Biol Phys 1996; 34:189–195.
30. Tsai J-S. Analyses of multi-irradiation film for system alignments in stereotactic radiotherapy (SRT) and radiosurgery
(SRS). Phys Med Biol 1996; 41:1597–1620.
31. Drzymala RE, Klein EE, Simpson JR, et al. Assurance of high
quality linac-based stereotactic radiosurgery. Int J Radiat Oncol Biol Phys 1994; 30:459–472.
32. Thornton AF, Ten Haken RK, Gerhardsson A, et al. Threedimensional motion analysis of an improved head immobilization system for simulation CT, MRI, and PET imaging.
Radiother Oncol 1991; 20:224–228.
33. Rosenthal SJ, Gall KP, Jackson M, et al. A precision cranial
immobilization system for conformal stereotactic fractionated
radiation therapy. Int J Radiat Oncol Biol Phys 1995; 33:
1239–1245.
APPENDIX
Influence of the origin location relative to the target to
the quality of IMRT
Figures A-1 and A-2 show one example of suboptimal
origin and axis assignment. The collimator pattern on the
left fails criterion (c) from the Peacock IMRT gantry axis
and isocenter placement subsection in Methods and Materials above: here the gantry axis G and origin O are posterior
to the anterior-most target so as to be out of reach of the
MIMiC, that is, the IMRT field. As a result, the isodose
distribution and dose volume histogram (DVH) are suboptimal (Fig. A-2). The DVH shows that to cover the entire
target volume, an isodose surface of 78% needs to be
prescribed, leading to maximum dose of 513 cGy. To im-
A non-invasive immobilization system
● J.-S. TSAI et al.
467
Fig. A-1. One-mm thick transverse CT view of fiducial points A
and L; by placing the gantry axis and origin using the posteriormost L3, the MIMiC is unable to irradiate the anteriormost target
at all gantry angles.
Fig. A-3. The gantry axis and origin at L3 in Fig. A-1, which is
moved anteriorly to L1, 4 cm toward the target and within reach of
the MIMiC field.
prove the plan, G and O were moved 4 cm anteriorly, as
shown in Fig. A-3, to the fiducial points (R, L, V)1. Resulting improvement to collimator pattern, dose distribution,
and DVH are shown in Fig. A-4, where the 84% isodose
line completely covers the whole target with a lower maximum dose of 476 cGy.
Fig. A-2. Gantry axis and origin placement from Fig. A-1 results
in suboptimal MIMiC patterns. Fifteen of the patterns including
vanes at the two lateral extremes of the MIMiC appear truncated.
Fig. A-4. Optimal dynamic IMRT collimator pattern, isodose
distribution, and DVH obtained from gantry axis and origin placement of Fig. A-3.