- Email: [email protected]
A Simple Method for Use in QC of CT for Radiation Therapy Treatment Planning
$S.OOt
0739-0211/92 .CC Copyright 0 1992 American Assoaatmn of Medical Dostmetrists
Medrcal Dosmefn~. Vol. 17. pp. 73-16 Printed in the U.S.A. All rights reserved.
A SIMPLE
METHOD FOR USE IN QC OF CT FOR RADIATION THERAPY TREATMENT PLANNING*
R. L. STERN, PH.D., L. T. NIIUASON, PH.D.,’ and R. K. TEN HAKEN, PH.D. Departments
of Radiation
Oncology and ‘Radiology, University of Michigan Medical Center, Ann Arbor, MI 48109-0010, U.S.A.
Abstract-A simple procedure for monitoring constancy of spatial measurement and CT number determination from CT images used in radiation therapy treatment planning is described. The procedure uses low-Z material rods glued to the underside of the CT table insert and does not require a special phantom. Measurements are made on the same patient images used for treatment planning. Deviations from predetermined baseline values outside quality control limits of k2 mm in spatial resolution and k20 CT numbers in density can be detected with a confidence level of 97% or better. Key Words:
Quality control, CT treatmentplanning.
INTRODUCI-ION
tions for CT scanners used for treatment planning.’ Those specifications have been adopted here as the QC limits; however, +20 CT numbers, which is equivalent to 2% uncertainty in the water calibration point, is used for monitoring CT number constancy. These limits fall well within the ~3% uncertainty in the measurement of the nature and extent of the involved anatomy and the treatment plan computation of dose recommended by the Committee on Radiation Oncology Studies.‘j For those aspects of a QC program described here, no special phantom is required. Instead, an idea was derived from quantitative CT studies,’ and three rods of different relative electron densities (Table 1) were glued to the underside of the CT table insert (Fig. 1). The rods appear in almost every patient scan with large enough field-of-view (Fig. 2). CT number constancy is monitored by measuring the CT number of any one or all three of the rods, and spatial accuracy is monitored by measuring the separation between the outermost rods. With this method, not only is there no need to take additional scans of a separate QC phantom, the QC measurements are made on the same patient images used in treatment planning. In practice, the images are first entered directly into the RTTP system. The separation between the centers of the outermost rods is then measured. Next, regions of interest (0.8 x 0.8 cm* regions were used here) are drawn about the center of each rod, and the average CT number within each region is determined. Electron density relative to water is then derived by the RTTP system from the average CT number. Monitoring this quantity checks the constancy of the conversion software. These procedures should be per-
Due to the widespread use of CT-derived information in radiation therapy treatment planning (RTTP), it is important to implement a routine quality control (QC) program to insure the accuracy of that information.’ This program should be a supplement to, not a substitute for, established QC programs designed for diagnostic purposes. *s3The program should include checks on the constancy of the spatial resolution, CT number determination, and conversion of CT number to tissue properties. For greater effectiveness the procedures should be simple and easy to perform. For QC, the constancy of measurements is more important than their absolute values, as the absolute accuracy should be determined before initiating the QC program (in CT-based RTTP, for example, during acceptance testing of the system4). Therefore, QC procedures need only detect changes in the quantities of interest beyond some predetermined limits. Measurements should be made by radiation therapy personnel after the CT data have been entered into the RTTP system, in order to test not only the CT images themselves but also the information transfer and image handling routines. MATERIALS AND METHODS Limits of &2 mm in spatial resolution and -~2% in CT number have been recommended as specifica-
* This work was supported in part by NC1 grant ROl-CA43200.
13
74
Medical Dosimetry Table 1. Physical properties of rods glued to underside of CT table insert Rod material
z
Acetal resin* Acrylic Polypropylene
6.95 6.47 5.52
Volume 17, Number 2, 1992
Acetal Resin
4-1
Polypropylene
(p
PJP:
1.37 1.16 0.91
Acrylic
* Trade name @DELRIN.
Fig. 1. Schematic of the underside of the CT table insert showing the three QC rods.
formed periodically (once a week for the results shown here) and the data graphed to display variations with time. Tests were also carried out using a diagnostic CT phantom,* in place of a patient, to determine the effects of various imaging conditions on the measured parameters for the rods underneath the table insert. CT images through a section of the phantom containing five different low-Z insert materials were used. The phantom was scanned in different positions on the CT table insert, in different orientations, and with different scanning techniques and fields-of-view typical of those used for patient scans. RESULTS
seven-month time span show that it is not necessary to correct for any of them in order to detect variations from the normal values outside of the established QC limits (Figs. 4 and 5). The relative electron density results were identical in trend to the average CT number data (Fig. 4) and are not shown here. The baselines used for determination of the QC limits for the CT numbers were the averages of the first 18 weeks (4 months) of measurements, while the true physical separation between the outermost rods was used for the distance measurements. For all cases, the confidence level of detecting variations in the measured values beyond the QC limits was derived by assuming a Gaussian distribution of the individual measure-
CT number was shown (Fig. 3) to vary linearly with the relative electron density for the low-Z inserts of the diagnostic test phantom. Measured CT numbers for the three QC rods fall along the same fit line, as expected. This linearity allows use of any convenient low-Z materials, not just those used here, for this type of QC procedure. Imaging the diagnostic phantom on top of the tabletop containing the QC rods under different conditions resulted in variations of only a few CT numbers for the phantom inserts and the rods. A larger variation (up to ten CT numbers for the acetal rod) was caused by changing the table height. Even larger variations were produced when bags of saline solution were placed on the table insert to simulate the effects of different body sizes. For distance measurements, the resolution is limited by the CT image pixel size, which can be as large as 1 mm for the large fields-ofview required for body scans. Although all of the factors mentioned above can affect the values obtained in the QC procedure, the results from measurements on patient images over a
* CatPhan Mark II, Alderson Research Laboratories, Inc., Stamford, CT.
Fig. 2. Typical patient CT image showing the QC rods and the regions of interest used to determine CT number.
QC of CT for RTTP 0 R. L. STERNet al.
1
PHANTOM INSERTS
0.0
0.5
1.0
RELATIVE
1.5
ELECTRON
2.0
DENSITY
Fig. 3. Measured data (solid squares) and linear fit for CT number vs. relative electron density for the inserts of a standard CT phantom. Measured values for the QC rods (open circles) fall along the fit line.
ments
about
the baseline
and determining
the num-
ber of standard deviations corresponding to the QC limits. Measured values outside of the QC limits would be detectable with a confidence level of 99% for the CT number and 97% for the rod separation. DISCUSSION 11oi
_--___--__-
108 90 80
70
__________-___
I_______________
3-J -60 -701 ___________________________-_____________---__
-1204
0
10
20
30
WEEK Fig. 4. Variation of measured average CT number with time for all three QC rods. The error bars represent one standard deviation in the distribution of CT numbers within the region of interest. The solid line is the baseline value, and the dashed lines represent the QC limits.
The procedures described here provide a simple method for monitoring the constancy of the determination of some physical properties derived from CT images that are used in RTTP. No special phantom or additional imaging is needed. Implementation of these procedures in a QC program will aid in detecting changes in CT scanner output and RTTP system data manipulation that might otherwise go unnoticed. The simple checks above are admittedly inadequate for completely mapping (or monitoring) the linearity and distortion of the whole CT image, as the spatial measurements take place over only one part of the image. They were not designed to do so; but, rather to indicate the utility of at least monitoring some fixed geometric arrangement on a routine basis, within the radiotherapy department, without the necessity for special, extra scans. As mentioned earlier, the checks above, or other measurements of objects that routinely show up on RTTP CT scans,’ should be
only part of or supplement a more complete program of CT QC.2*3
76
Medical Dosimetry
Volume 17, Number 2, 1992
87
85-0
0 0
0
84-
0
0
0 0
83-o
0 00
0
0 00
0 0.0
??0.e.
00
0
??
0 82-,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,~~~~~~~__~_, 81 0
I 10
I 20
1 30
WEEK Fig. 5. Variation of measured separation between the outermost rods with time. For the largest field-of-view, the uncertainty due to pixel size and measurement precision is 0.86 mm. The solid line is the true average separation of 83.8 mm, and the dashed lines represent the QC limits.
REFERENCES 1. Ten Haken, R.K.; Kessler, M.L.; Stem, R.L.; Ellis, J.H.; Niklason, L.T. Quality assurance of CT and MRI for radiation therapy treatment planning. In: Starkschall, G.; Horton, J.L., editors. Quality assurance in radiotherapy physics. Madison: Medical Physics Publishing; 199 1:73- 104. 2. National Council on Radiation Protection and Measurements. Quality assurance for diagnostic imaging equipment. NCRP report no. 99. Bethesda: NCRP; 1988. 3. Hendee, W.R., editor. The selection and performance ofradiologic equipment. Baltimore: Williams and Wilkins; 1985.
4. McCullough, E.C.; Holmes, T.W. Acceptance testing computerized radiation therapy treatment planning systems: direct utilization of CT scan data. Med. Phys. 12~237-242; 1985. 5. Hogstrom, K.R. Implementation of CT treatment planning. In: Wright, A.E.; Boyer, A.L., editors. Radiation therapy treatment planning. Medical physics monograph no. 9. New York AAPMJAIP; 1983:269-28 I. 6. Stewart, J.R.; Hicks, J.A.; Boon, M.L.M., Simpson, L.D. Computed tomography in radiation therapy. ht. J. Radiat. Oncol. Biol. Phys. 4:3 13-324; 1978. 7. Cann, C.E.; Genant, H.K. Precise measurement of vertebral mineral content using CT. J. Comput. Assist. Tomogr. 4:493500; 1980.
0739-0211/92 .CC Copyright 0 1992 American Assoaatmn of Medical Dostmetrists
Medrcal Dosmefn~. Vol. 17. pp. 73-16 Printed in the U.S.A. All rights reserved.
A SIMPLE
METHOD FOR USE IN QC OF CT FOR RADIATION THERAPY TREATMENT PLANNING*
R. L. STERN, PH.D., L. T. NIIUASON, PH.D.,’ and R. K. TEN HAKEN, PH.D. Departments
of Radiation
Oncology and ‘Radiology, University of Michigan Medical Center, Ann Arbor, MI 48109-0010, U.S.A.
Abstract-A simple procedure for monitoring constancy of spatial measurement and CT number determination from CT images used in radiation therapy treatment planning is described. The procedure uses low-Z material rods glued to the underside of the CT table insert and does not require a special phantom. Measurements are made on the same patient images used for treatment planning. Deviations from predetermined baseline values outside quality control limits of k2 mm in spatial resolution and k20 CT numbers in density can be detected with a confidence level of 97% or better. Key Words:
Quality control, CT treatmentplanning.
INTRODUCI-ION
tions for CT scanners used for treatment planning.’ Those specifications have been adopted here as the QC limits; however, +20 CT numbers, which is equivalent to 2% uncertainty in the water calibration point, is used for monitoring CT number constancy. These limits fall well within the ~3% uncertainty in the measurement of the nature and extent of the involved anatomy and the treatment plan computation of dose recommended by the Committee on Radiation Oncology Studies.‘j For those aspects of a QC program described here, no special phantom is required. Instead, an idea was derived from quantitative CT studies,’ and three rods of different relative electron densities (Table 1) were glued to the underside of the CT table insert (Fig. 1). The rods appear in almost every patient scan with large enough field-of-view (Fig. 2). CT number constancy is monitored by measuring the CT number of any one or all three of the rods, and spatial accuracy is monitored by measuring the separation between the outermost rods. With this method, not only is there no need to take additional scans of a separate QC phantom, the QC measurements are made on the same patient images used in treatment planning. In practice, the images are first entered directly into the RTTP system. The separation between the centers of the outermost rods is then measured. Next, regions of interest (0.8 x 0.8 cm* regions were used here) are drawn about the center of each rod, and the average CT number within each region is determined. Electron density relative to water is then derived by the RTTP system from the average CT number. Monitoring this quantity checks the constancy of the conversion software. These procedures should be per-
Due to the widespread use of CT-derived information in radiation therapy treatment planning (RTTP), it is important to implement a routine quality control (QC) program to insure the accuracy of that information.’ This program should be a supplement to, not a substitute for, established QC programs designed for diagnostic purposes. *s3The program should include checks on the constancy of the spatial resolution, CT number determination, and conversion of CT number to tissue properties. For greater effectiveness the procedures should be simple and easy to perform. For QC, the constancy of measurements is more important than their absolute values, as the absolute accuracy should be determined before initiating the QC program (in CT-based RTTP, for example, during acceptance testing of the system4). Therefore, QC procedures need only detect changes in the quantities of interest beyond some predetermined limits. Measurements should be made by radiation therapy personnel after the CT data have been entered into the RTTP system, in order to test not only the CT images themselves but also the information transfer and image handling routines. MATERIALS AND METHODS Limits of &2 mm in spatial resolution and -~2% in CT number have been recommended as specifica-
* This work was supported in part by NC1 grant ROl-CA43200.
13
74
Medical Dosimetry Table 1. Physical properties of rods glued to underside of CT table insert Rod material
z
Acetal resin* Acrylic Polypropylene
6.95 6.47 5.52
Volume 17, Number 2, 1992
Acetal Resin
4-1
Polypropylene
(p
PJP:
1.37 1.16 0.91
Acrylic
* Trade name @DELRIN.
Fig. 1. Schematic of the underside of the CT table insert showing the three QC rods.
formed periodically (once a week for the results shown here) and the data graphed to display variations with time. Tests were also carried out using a diagnostic CT phantom,* in place of a patient, to determine the effects of various imaging conditions on the measured parameters for the rods underneath the table insert. CT images through a section of the phantom containing five different low-Z insert materials were used. The phantom was scanned in different positions on the CT table insert, in different orientations, and with different scanning techniques and fields-of-view typical of those used for patient scans. RESULTS
seven-month time span show that it is not necessary to correct for any of them in order to detect variations from the normal values outside of the established QC limits (Figs. 4 and 5). The relative electron density results were identical in trend to the average CT number data (Fig. 4) and are not shown here. The baselines used for determination of the QC limits for the CT numbers were the averages of the first 18 weeks (4 months) of measurements, while the true physical separation between the outermost rods was used for the distance measurements. For all cases, the confidence level of detecting variations in the measured values beyond the QC limits was derived by assuming a Gaussian distribution of the individual measure-
CT number was shown (Fig. 3) to vary linearly with the relative electron density for the low-Z inserts of the diagnostic test phantom. Measured CT numbers for the three QC rods fall along the same fit line, as expected. This linearity allows use of any convenient low-Z materials, not just those used here, for this type of QC procedure. Imaging the diagnostic phantom on top of the tabletop containing the QC rods under different conditions resulted in variations of only a few CT numbers for the phantom inserts and the rods. A larger variation (up to ten CT numbers for the acetal rod) was caused by changing the table height. Even larger variations were produced when bags of saline solution were placed on the table insert to simulate the effects of different body sizes. For distance measurements, the resolution is limited by the CT image pixel size, which can be as large as 1 mm for the large fields-ofview required for body scans. Although all of the factors mentioned above can affect the values obtained in the QC procedure, the results from measurements on patient images over a
* CatPhan Mark II, Alderson Research Laboratories, Inc., Stamford, CT.
Fig. 2. Typical patient CT image showing the QC rods and the regions of interest used to determine CT number.
QC of CT for RTTP 0 R. L. STERNet al.
1
PHANTOM INSERTS
0.0
0.5
1.0
RELATIVE
1.5
ELECTRON
2.0
DENSITY
Fig. 3. Measured data (solid squares) and linear fit for CT number vs. relative electron density for the inserts of a standard CT phantom. Measured values for the QC rods (open circles) fall along the fit line.
ments
about
the baseline
and determining
the num-
ber of standard deviations corresponding to the QC limits. Measured values outside of the QC limits would be detectable with a confidence level of 99% for the CT number and 97% for the rod separation. DISCUSSION 11oi
_--___--__-
108 90 80
70
__________-___
I_______________
3-J -60 -701 ___________________________-_____________---__
-1204
0
10
20
30
WEEK Fig. 4. Variation of measured average CT number with time for all three QC rods. The error bars represent one standard deviation in the distribution of CT numbers within the region of interest. The solid line is the baseline value, and the dashed lines represent the QC limits.
The procedures described here provide a simple method for monitoring the constancy of the determination of some physical properties derived from CT images that are used in RTTP. No special phantom or additional imaging is needed. Implementation of these procedures in a QC program will aid in detecting changes in CT scanner output and RTTP system data manipulation that might otherwise go unnoticed. The simple checks above are admittedly inadequate for completely mapping (or monitoring) the linearity and distortion of the whole CT image, as the spatial measurements take place over only one part of the image. They were not designed to do so; but, rather to indicate the utility of at least monitoring some fixed geometric arrangement on a routine basis, within the radiotherapy department, without the necessity for special, extra scans. As mentioned earlier, the checks above, or other measurements of objects that routinely show up on RTTP CT scans,’ should be
only part of or supplement a more complete program of CT QC.2*3
76
Medical Dosimetry
Volume 17, Number 2, 1992
87
85-0
0 0
0
84-
0
0
0 0
83-o
0 00
0
0 00
0 0.0
??0.e.
00
0
??
0 82-,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,~~~~~~~__~_, 81 0
I 10
I 20
1 30
WEEK Fig. 5. Variation of measured separation between the outermost rods with time. For the largest field-of-view, the uncertainty due to pixel size and measurement precision is 0.86 mm. The solid line is the true average separation of 83.8 mm, and the dashed lines represent the QC limits.
REFERENCES 1. Ten Haken, R.K.; Kessler, M.L.; Stem, R.L.; Ellis, J.H.; Niklason, L.T. Quality assurance of CT and MRI for radiation therapy treatment planning. In: Starkschall, G.; Horton, J.L., editors. Quality assurance in radiotherapy physics. Madison: Medical Physics Publishing; 199 1:73- 104. 2. National Council on Radiation Protection and Measurements. Quality assurance for diagnostic imaging equipment. NCRP report no. 99. Bethesda: NCRP; 1988. 3. Hendee, W.R., editor. The selection and performance ofradiologic equipment. Baltimore: Williams and Wilkins; 1985.
4. McCullough, E.C.; Holmes, T.W. Acceptance testing computerized radiation therapy treatment planning systems: direct utilization of CT scan data. Med. Phys. 12~237-242; 1985. 5. Hogstrom, K.R. Implementation of CT treatment planning. In: Wright, A.E.; Boyer, A.L., editors. Radiation therapy treatment planning. Medical physics monograph no. 9. New York AAPMJAIP; 1983:269-28 I. 6. Stewart, J.R.; Hicks, J.A.; Boon, M.L.M., Simpson, L.D. Computed tomography in radiation therapy. ht. J. Radiat. Oncol. Biol. Phys. 4:3 13-324; 1978. 7. Cann, C.E.; Genant, H.K. Precise measurement of vertebral mineral content using CT. J. Comput. Assist. Tomogr. 4:493500; 1980.