A simple approach of three-isocenter IMRT planning for craniospinal irradiation
A simple approach of three-isocenter IMRT planning for craniospinal irradiation
Zheng Wang 0
Wei Jiang 0
Yuanming Feng 2
Yang Guo 0
Zheng Cong 0
Bin Song 0
Yu Guo 1
0 Department of Radiation Oncology, Huanhu Hospital , Tianjin , China
1 Department of Biomedical Engineering, Tianjin University , Tianjin , China
2 Department of Radiation Oncology, Tianjin Medical University Cancer Hospital , Tianjin , China
Purpose: To develop a new IMRT technique to simplify the process and improve efficiency in radiotherapy treatment planning for craniospinal irradiation (CSI) treatment. Methods: Image data of 9 patients who received CSI treatment in 2012 were used, the prescription was 36Gy in 20 fractions. Two treatment plans were created for each patient, one was with the new technique called three-isocenter overlap-junction (TIOJ) IMRT and the other was with the three-isocenter jagged-junction (TIJJ) IMRT technique. The comparative study was conducted using the parameters of heterogeneity index (HI), conformity index (CI), and doses to the organs at risk (OARs). Results: Comparing the TIOJ IMRT plans with the TIJJ IMRT plans, the average homogeneity index is 0.071 0.003 and 0.077 0.002, respectively, and the averaged conformity number is 0.80 0.012 and 0.80 0.009, respectively. There are no significant differences (p > 0.05). Both plans provide satisfactory sparing for the OARs. Conclusions: The TIOJ IMRT technique for CSI treatment planning can create similar plans as with the TIJJ IMRT technique, but the new technique greatly simplifies the steps required to manually set field widths and boundaries and improved efficiency.
Craniospinal irradiation; Field junction; Field edge matching; Overlap-junction IMRT
Craniospinal irradiation (CSI) has become an important
treatment method for primary tumors. Commonly
treated tumors include medulloblastoma, high-risk
germcell tumors, and some radio-sensitive secondary malignant
tumors of the meninges. Emerging radiotherapy
techniques, such as three-dimensional conformal
radiotherapy (3DCRT) and intensity modulated radiation
therapy (IMRT), have gradually replaced the traditional
large field radiotherapy technology used in CSI
treatment. CSI involves complex anatomical structures and
requires complex treatment planning, which often
entails setting multiple isocenters and matching a large
number of fields to obtain satisfactory plans. IMRT
technology can offer better comformity Index (CI) and
homogeneity index (HI) than traditional multi-field
3DCRT in complex target areas. Inverse treatment
planning with IMRT reduces the difficulty of planning
and implementation as well. These two advantages are
particularly important in CSI. Helical tomotherapy
[1,2] and radiotherapy techniques based on protons 
have also been used in CSI. Another emerging
radiotherapy technique, volumetric modulated arc therapy
(VMAT), has also been applied in CSI treatment [4,5].
In comparison to other IMRT techniques [6-8], the
three-isocenter jagged-junction (TIJJ) IMRT recently
proposed by Cao et al.  achieves similar CI and HI and
simplifies planning and implementation processes. Yet the
treatment plan still involves adjustment of a large number
of staggered fields, and the planning process is time
consuming. Reducing the complexity of treatment plans and
shortening treatment time will make the treatment more
reliable and improve the overall treatment quality. For this
purpose, we have developed a simplified IMRT technique
called three-isocenter overlap-junction (TIOJ) IMRT, and
presented in this manuscript. The goal is to simplify the
implementation of the treatment plan, ensure satisfactory
CI and HI, and reduce the time needed for planning and
implementing CSI treatment.
Material and methods
Image data of nine patients who received CSI treatment in
2012 were used for this study (Table 1). Two CSI treatment
plans were made for each patient, one was TIOJ IMRT as
defined in this manuscript and the other was TIJJ as defined by
Cao et al. . The two plans were compared using the
parameters of HI, CI, and doses to the organs at risk (OARs).
Among the nine patients evaluated, three were
averagerisk medulloblastoma patients who received reduced dose
(23.4 Gy in 13 fractions) while the others received 36 Gy
in 20 fractions. To make the plans comparable, we created
the plans with the same prescription of 36 Gy in 20
fractions for all patients. The study using patient data was
approved by the Ethics Committee of Tianjin Huanhu
Patient position and simulation
All patients were set up in a supine position with both
hands naturally and comfortably placed at their sides. A
thermoplastic facial mask and a body mask were used to
fix the patient. After the immobilization, Six BrainLAB
(BrainLAB AG, Feldkirchen, Germany) real-time infrared
reflective marker balls were placed on the surface of the
thermoplastic masks. The simulation CT images were
acquired using a Brilliance CT Big Bore (Philips Medical
Systems, Cleveland, OH, USA). Scan range was from the
top of the head to the proximal femur, which included the
entire torso and arms. Slice thickness was 3 mm.
Target delineation: Delineation of planning target volume
(PTV) and OARs were both based on CT images. PTVcns
included the brain (PTV brain) and spinal cord (PTVspinal).
PTV brain included the whole brain, the meninges and
3 mm beyond their external boundary. PTVspinal included
C1 through S3, and 5 mm beyond their external boundary.
Lens, optic nerves, eyes, thyroid, heart, lungs, liver, and
kidneys were delineated as OARs for comparison.
Table 1 Patient demographics
Patient demographics at time of CT scan acquisition.
Abbrevations: PTV Planning target volume, M male, F female, GCT Germ cell
tumor, MM Medublastoma.
IMRT plans for all patients were generated using
Eclipse treatment planning system (Eclipse TPS 10.0.24,
Varian Medical Systems Inc., Palo Alto, CA, USA). A
6MV (Varian 6EX) linear accelerator equipped with a
120-leaf multileaf collimator (MLC) was used to
implement the treatment plan. The BrainLAB ExacTrac was
used for setup before treatment, moving the table for
different isocenters and monitoring the mobility of the
patient during treatment.
TIOJ IMRT plan
The three isocenters of the beams were placed in the
TIOJ plan such that they had the same distance of
100 cm to the source, and were denoted as A, B, C,
respectively. As shown in Figure 1a, the line connecting
the three isocenters (A, B and C) was parallel to and
above the midline of the patient in the sagittal plane. All
of the PTVspinal were located dorsal to this connecting
line. The positions of A, B and C on this connecting line
were determined by the following steps. Point A was set
as the midpoint along the rostral-caudal direction of the
PTVbrain in the sagittal plane. If the length of the PTVcns
was less than 80 cm, B and C were set to make A, B,
and C 25 cm apart to each other. If the length of the
PTVcns was greater than 80 cm, B and C were set to
make A, B, and C 30 cm apart to each other. This was
to ensure sufficient beam overlap among the fields with
different isocenters. The collimator angle was set at 0
for all the field sets. The fields in every field set of an
isocenter had the same field size. With this set-up
scheme, two beam overlap regions with lengths of 15 cm
or 10 cm were formed between the three isocenters. The
field set with isocenter A (IsoA) contained seven fields,
and the gantry angles were 0, 65, 100, 123, 230, 257
and 290, respectively. The field sets with IsoB and IsoC
had three radiation fields each, and the gantry angles
were 145, 180 and 215, respectively. As the distance
from spine to treatment machine head for these
posterior and posterior oblique fields in the field sets for
spinal cord treatment is shorter than 100 cm, the leaf
width of MLC projected in the spinal cord region is
smaller than its nominal width, this could theoretically
provide better homogenous dose distribution in the
volumes of field junctions.
The constrains set for the inverse optimizations were as
following, 99% of the PTV (for both PTV brain and PTVspinal) is to
receive 95% of the prescribed dose, the doses to the lens and
eyes are to be kept to the lowest achievable, and there are no
dose constrains for other OARs.
TIJJ IMRT plan
The method first proposed by Cao et al.  is shown in
Figure 1b. Plans covered the cranial and spinal PTVs
with the use of three isocenters. Care was taken to ensure
Figure 1 Beam arrangements for TIOJ plan and TIJJ plan. Sagittal view shows beam arrangements for both cranial and spinal regions for
TIOJ plan (a) and TIJJ plan (b). The conventional plan figure shows the placement of isocenters in a TIOJ plan has beams with shorter source to
surface distances (SSDs) which make the distances between the MLC and the spinal cord shorter.
that no beam entered through the patients shoulders. One
isocenter (Iso1) was placed in the cranial PTV, and two
isocenters (Iso2 and Iso3) were placed in the spinal PTV,
with Iso2 located superior to Iso3. The three isocenters
were collinear, and were placed near the patients midline.
For the patients scanned in the supine position, the Iso1
field set consisted of seven fields with gantry angles of 0,
65, 100, 123, 230, 257, and 290. Both field sets from
Iso2 and Iso3 consisted of three beams with gantry angles
of 145, 180, and 215. The collimator angles were set at
0 for all fields. Adjacent field sets were intentionally
overlapped to treat a common region of the spine. Field edges
were staggered in 1.1 cm steps. The lateral or nearly lateral
field cannot treat through the shoulders.
The two plans were compared, mainly using HI and CI of
the target areas. Currently there are multiple definitions
for HI and CI. The definitions used by Cao et al.  were
adopted in order to compare the two methods. HI is
where DMedian, D2% and D98% are doses received by 50%,
2% and 98% the PTV volume, respectively. CI is defined as
V T pres
V T pres
where VT pres is the targe volume covered by 95%
isodose surface. Vpres is the treated volume covered by 95%
isodose surface and VT is the volume of target. The
value of CI is in the range of 01, where a value closer
to 1 indicates better conformity. The value of HI is in
the range of 10, where a value closer to 0 indicates
Plan quality assurance for the cranial-spinal and
In complex plans, multiple isocentes can overlap.
Homogeneity of dose distribution in overlapping radiation
fields of different isocenters must be achieved. The
influence of variation in treatment table positioning on dose
distribution must be carefully evaluated. In this study,
the plan quality assurance (QA) was conducted with an
IBA matrix (IBA Dosimetry Germany) The QA plans for
the IBA matrix were generated in Eclipse TPS, and the
measurements of dose distributions in the overlapped
areas of field junctions were performed afterwards. Due
to limitations in the size of the IBA matrix test model
(27 cm 27 cm), only dose distribution and pass rate in
the overlapping areas were measured in each plan.
Statistical analysis was performed using the SPSS
statistical analysis software package, Version 18.0 (SPSS Inc.,
Chicago, IL). A nonparametric related-samples Wilcoxon
signed-ranks test was chosen because the sample sizes were
small and not of a normalized distribution, P values <0.05
were considered statistically significant.
For the nine patients included in this study, both TIOJ
and TIJJ IMRT reach the goal of the 95% isodose curve
covering at least 99% of the PTV. Figure 2 shows the
PTVcns coverage for patient 5 using a dose volume
histogram (DVH). Both plans meet the initial goal for
the target volume coverage.
PTVcns coverage, HI, CI and other results in the two
plans are listed in Table 2. HI and CI obtained with TIOJ
are 0.071 0.003 (Mean variance) and 0.80 0.012,
respectively. HI and CI obtained with TIJJ are 0.077
0.002 and 0.80 0.009, respectively. Both results meet
the PTV dose coverage requirements as published in the
International Commission on Radiation Units (ICRU)
guidelines [10,11]. There are no significant differences
for CI and HI (P > 0.05).
Figure 3a and b shows the dose distribution in the two
plans in patient 5. No cold dosing spots or hot
dosing spots are found in the radiation beam overlapping
regions between isocenters. The two treatment planning
techniques provide similar plans in this regard.
Table 2 Evaluation parameters for TIOJ and TIJJ plans
TIOJ TIJJ TIOJ TIJJ
1 0.083 0.086 0.83 0.83
2 0.072 0.072 0.74 0.78
3 0.070 0.070 0.81 0.80
4 0.083 0.085 0.86 0.84
5 0.061 0.063 0.80 0.78
6 0.070 0.070 0.79 0.79
7 0.061 0.060 0.81 0.81
8 0.070 0.070 0.83 0.72
9 0.069 0.081 0.76 0.69
Mean variance 0.071 0.003 0.077 0.002 0.80 0.012 0.80 0.009
Abbrevations: CI conformity index, HI homogeneity index.
The total monitor units (MUs) needed for delivering the
fraction dose of 1.8 Gy are 1907.4 60.5 (ranging from
1575 to 2104) with TIOJ plan and 1903.3 34.8 (ranging
from 1575 to 2253) with TIJJ plan, respectively, there is no
significant difference (P > 0.05). The average doses to the
torso in the two plans are 8.75 0.59 Gy with TIOJ and
8.26 0.53 Gy with TIJJ, respectively, there is no
significant difference (P > 0.05). No correlation between total
MU and torso dose in either plan was found.
Figure 4 shows the doses to different OARs in the two
plans. There are no significant differences in the doses
of different OARs with the two plans (P > 0.05).
The comparison of the dose distribution in spinal-spinal
beam overlapping regions using an IBA matrix for patient
5 is shown in Figure 5. QA evaluations of TIOJ plans
using Gamma index (3% for absolute dose and 3 mm for
relative dose evaluation) were performed. All plans has a
pass rate above 90%. The mean pass rate is 94.6% (ranging
Figure 2 Dose volume histogram (DVH) comparison of target coverage. DVH of a representative plan comparing target coverage between
the TIOJ plan and the TIJJ plan. The target coverage of both plans meet the initial goal for the target volume coverage.
Figure 3 Isodose distribution of Patient 5 for TIOJ plan and TIJJ plan. Isodose distribution on a midline sagittal CT slice of Patient 5 for TIOJ
plan (a) and TIJJ plan (b). Color legend indicates isodose lines, in centigrays. Planning target volume is shown in pink.
from 92.5% to 97.7%). The measured and planned dose
profiles agree well and there are no cold or hot spots.
Simulation, planning, QA and treatment delivery
processes for CSI techniques require great care. Patient
position, patient immobilization, target delineation,
protection of OARs, HI, and the field junctions can have
serious effects on the treatment outcome. Compared to
traditional 2D radiation therapies, 3D radiation
technologies such as IMRT and VMAT [4,5] provide much
Figure 4 Median and variation. Median and variation (shown as
bars) of mean doses to OARs in TIOG (shade) and TIJJ (white) plans.
L: Left; R: right.
better target coverage and protection of OARs. Although
there have been previous reports of successful
applications of IMRT technology in CSI [6,7,12,13], most of
these only partially modified or optimized the traditional
CSI plan by adding an IMRT component. The technical
advantages of IMRT were not fully utilized. Recently,
Cao et al.  simplified the radiation field matching
issue involved in traditional planning by using a single
IMRT plan with three isocenters. In practice, this
technique requires large amounts of manual adjustment of
radiation fields to generate beam overlapping regions
which is time consuming.
The TIOJ IMRT technique presented in this
manuscript is similar to the ones proposed by Seppala et al.
 and Cao et al. . But with TIOJ, there is no need
for manual adjustment of radiation fields to generate
beam overlapping regions and the difficulties in the
planning process is greatly reduced while maintaining
similar and satisfactory plan results.
As compared to the plan by Cao et al., the placement
of isocenters in a TIOJ plan has beams with shorter
source to surface distances (SSDs) which make the
distances between the MLC and the spinal cord shorter.
This makes the projected MLC width smaller in the
spinal cord, which is equivalent to having thinner MLC
leafs, which helps maximize the potential of the inverse
treatment planning system and avoiding hot or cold
Figure 5 QA results in the overlapping spinal cord region with TIOJ plan for patient 5. (a) and (c) show the coronal plane dose
distributions and profiles for the spinal-spinal junction. The black lines indicate the junction regions. (b) shows the gamma index (3% for absolute
dose and 3 mm) for the relative dose evaluation.
spots in beam overlapping areas. And with the new TIOJ
technique, the lengths of all radiation fields have the
maximum length set by MLC (40 cm). In TIJJ, field lengths
are manually set. In contrast, larger beam overlap regions
can be achieved in the target area with TIOJ, and the
planning process involves far less manual setting and
adjustment which facilitates dose optimization in the
inverse planning system and helps avoid dose cold spots and
hot spots in beam overlap regions.
In the TIJJ plan proposed by Cao et al. , the filed
length in the overlap region between isocenters needs to
be manually adjusted in order to generate jagged
junction areas with 1.1 cm intervals. During this process,
the field width in left-right direction also has to be
manually adjusted to avoid excessive radiation field
width relative to patient lateral size and MLC leakage.
The planning process is complex and time consuming.
The proposed TIOJ method only requires three
isocenters to be manually determined. The field size and
shape are then automatically set by the planning system.
The new method greatly simplifies the planning process
and reduces planning time. The average time needed to
make a TIOJ IMRT plan is about 10 min shorter than
that using TIJJ (30 min vs. 40 min). This simplification
can also facilitate a standardized design procedure,
which minimizes human errors in the process.
VMAT, another emerging radiation therapy
technology, has also been applied in CSI treatment [4,5]. With
its high dose rate and technical advantages similar to
helical tomotherapy, VMAT can satisfactorily solve some
CSI related problems (such as a long target area and
matching between radiation fields). Lee et al.  applied
VMAT technology to treat five patients. PTV lengths
were 48.1-83.7 cm. The prescription dose was 23.4Gy
in13 fractions. The final mean CI and HI were 1.22
(range: 1.09-1.45) and 1.04 (range: 1.03-1.07),
respectively. The formulas Lee et al. used to calculate HI and
CI were different from ours. For comparison purposes,
we used their formulas to re-process the data collected
from our nine patients. The CI and HI found with the
TIOJ method were 1.23 (range: 1.17-1.34), and 1.08
(range: 1.07-1.11), respectively. Although we could not
perform statistical analyses to compare their data with
ours, it can be seen that the CI and HI obtained using
TIOJ for CSI were very similar to that obtained by Lee
et al. using the VMAT technique.
Most patients that receive CSI treatment are teenagers.
Before treatment, evaluation of OAR dose needs to be
performed to minimize the incidence of
radiotherapyrelated complications. In the current study, dose limitation
requirements were defined for the lens and eyes, similar to
the studies by Cao et al.  and Lee et al. . Figure 4
shows the OAR doses received with each of the two plans,
OAR doses were at a satisfactory level and basically the
same with TIOJ and TIJJ. The mean dose of left lens and
right lens was 6.89 and 6.91Gy, respectively, using TIOJ.
The mean dose of left lens and right lens was 6.91 and
6.78GY, respectively, using TIJJ. These doses were
significantly lower than the 9.1Gy, as reported by Cao et al. ,
and significantly lower than the 20Gy and 18Gy reported
by Lee et al.  with VMAT. These differences may be
due to the more stringent limit conditions we set for the
lens and eye.
The total MU and the mean dose received by the torso
are the most important risk factors for secondary tumors
[14-16]. We found that the total MU and dose in the
torso using TIOJ were similar to that found using TIJJ.
Total MU using TIOJ was 1907.4 60.6 (range: 1575
2104). Total MU using TIJJ was 1903.3 34.8 (range:
15752253). So when use the same dose rate, the deliver
(beam on) time for both technique (TIOJ and TIJJ) will
be very similar. We also found that the total MU were
substantially higher than that reported by Lee et al. with
VMAT. This suggests that, VMAT technology may be
more advantageous than IMRT in reducing the
incidence of secondary tumors. We believe that as IMRT
treatment involves a large amount of MLC occlusion,
total MU may not fully reflect dose received by the
whole body. In the current study, mean dose received by
the torso was calculated as 8.75 0.59 with TIOJ and
8.26 0.53 with TIJJ. There was no significant difference
between the two plans. As Lee et al. did not calculate
the torso dose with VMAT, we were not able to make
The TIOJ IMRT method for CSI treatment outlined in
this article can creat plans with satisfactory CI and HI.
The use of three isocenters and beam overlap regions
between the isocenters helps avoid typical CSI problems,
such as over-long radiation fields and matching between
the fields. As compared to previously reported methods,
TIOJ greatly simplified the steps required to manually
set field widths and boundaries and improved efficiency.
Only one treatment plan and a simple bed capable of
moving in one direction are needed to complete the
entire treatment. TIOJ IMRT provides a simple and
efficient choice for CSI treatment.
CSI: Craniospinal irradiation; TIOJ: Three-isocenter overlap-junction;
TIJJ: Three-isocenter jagged-junction; HI: Heterogeneity index; CI: Conformity
index; OARs: Organs at risk; 3DCRT: Three-dimensional conformal
radiotherapy; IMRT: Intensity modulated radiation therapy; VMAT: Volumetric
modulated arc therapy; PTV: Planning target volume; MLC: Multileaf
collimator; SAD: Source to axis; DVH: Dose volume histogram;
ICRU: International commission on radiation units; MUs: Monitor units;
SSD: Source to surface distances.
WZ, JW and FYM participated in the study design, contributed equally in
analysis of data and drafting the manuscript, GY, CZ and SB participated in
the treatment planning. GY made important contributions in collecting and
analyzing data, and in revising the content. All authors read and approved
the final manuscript.
1. Bauman G , Yartsev S , Coad T , Fisher B , Kron T : Helical tomotherapy for craniospinal radiation . Br J Radiol 2005 , 78 : 548 - 552 . PMID: 15900062 .
2. Parker W , Brodeur M , Roberge D , Freeman C : Standard and nonstandard craniospinal radiotherapy using helical tomoTherapy . Int J Radiat Oncol Biol Phys 2010 , 77 : 926 - 931 . PMID: 20231076 .
3. St . Clair WH , Adams JA , Bues M , Fullerton BC , La Shell S , Kooy HM , Loeffler JS , Tarbell NJ : Advantage of protons compared to conventional X-ray or IMRT in the treament of a pediatric patient with medulloblastoma . Int J Radiat Oncol Biol Phys 2004 , 58 : 727 - 734 . PMID: 14987427 .
4. Fogliata A , Bergstrom S , Cafaro I , Clivio A , Cozzi L , Dipasquale G , Hllstrm P , Mancosu P , Navarria P , Nicolini G , Parietti E , Pesce GA , Richetti A , Scorsetti M , Vanetti E , Weber DC : Cranio-spinal irradiation with volumetric modulated arc therapy: a multi-institutional treatment experience . Radiother Oncol 2011 , 99 : 79 - 85 . PMID: 21421273 .
5. Lee YK , Brooks CJ , Bedford JL , Warrington AP , Saran FH : Development and evaluation of multiple isocentric volumetric modulated arc therapy technique for craniospinal axis radiotherapy planning . Int J Radiat Oncol Biol Phys 2012 , 82 : 1006 - 1012 . PMID: 21345612 .
6. Seppala J , Kulmala J , Lindholm P , Minn H : A method to improve target dose homogeneity of craniospinal irradiation using dynamic split field IMRT . Radiother Oncol 2010 , 96 : 209 - 215 . PMID: 20581696 .
7. Pai Panandiker A , Ning H , Likhacheva A , Ullman K , Arora B , Ondos J , Karimpour S , Packer R , Miller R , Citrin D : Craniospinal irradiation with spinal IMRT to improve target homogeneity . Int J Radiat Oncol Biol Phys 2007 , 68 : 1402 - 1409 . PMID: 17467921 .
8. Parker WA , Freeman CR : A simple technique for craniospinal radiotherapy in the supine position . Radiother Oncol 2006 , 78 : 217 - 222 . PMID: 16330119 .
9. Cao F , Ramaseshan R , Corns R , Harrop S , Nuraney N , Steiner P , Aldridge S , Liu M , Carolan H , Agranovich A , Karvat A : A three-isocenter jaggedjunction IMRT approach for craniospinal irradiation without beam edge matching for field junctions . Int J Radiat Oncol Biol Phys 2012 , 84 : 648 - 654 . PMID: 22436794 .
10. International Commission of Radiation Units and Measurements: ICRU Report 50 : Prescribing, recording, and reporting photon beam therapy . Bethesda: International Commission on Radiation Units and Measurements ; 1993 .
11. International Commission of Radiation Units and Measurements: ICRU Report 62 : Prescribing, recording, and reporting photon beam therapy (supplement to ICRU Report 50) . Bethesda: International Commission on Radiation Units and Measurements ; 1999 .
12. Kusters JM , Louwe RJ , van Kollenburg PG , Kunze-Busch MC , Gidding CE , van Lindert EJ , Kaanders JH , Janssens GO : Optimal normal tissue sparing in craniospinal axis irradiation using IMRT with daily intrafractionally modulated junction . Int J Radiat Oncol Biol Phys 2011 , 81 : 1405 - 1414 . PMID: 21300472 .
13. Parker W , Filion E , Roberge D , Freeman CR : Intensity-modulated radiotherapy for craniospinal irradiation: target volume considerations, dose constraints, and competing risks . Int J Radiat Oncol Biol Phys 2007 , 69 : 251 - 257 . PMID: 17707279 .
14. Followill D , Geis P , Boyer A : Estimates of whole-body dose equivalent produced by beam intensitymodulated conformal therapy . Int J Radiat Oncol Biol Phys 1997 , 38 : 667 - 672 . PMID: 9231693 .
15. Hall EJ : Intensity-modulated radiation therapy, protons, and the risk of second cancers . Int J Radiat Oncol Biol Phys 2006 , 65 : 1 - 7 . PMID:16618572.
16. Kry SF , Salehpour M , Followill DS , Stovall M , Kuban DA , White RA , Rosen II : The calculated risk of fatal secondary malignancies from intensity modulated radiation therapy . Int J Radiat Oncol Biol Phys 2005 , 62 : 1195 - 1203 . PMID: 15990025 .