- Email: [email protected]
Three dimensional conformal radiation therapy treatment planning and delivery verification for mulit-institutional dose escalation trials using an anthropomorphic phantom
356
1. J. Radiation
Oncology
l Biology
0 Physics
Volume 48, Number 3. Supplement.
2000
Materials & Methods: NTCPs for the rectum and small intestine were calculated for 24 bladder cancer patients referred to radical four-field conformal radiotherapy using the patients’ actual dose-fractionation regime. Radiation tolerance doses from several clinical studies were applied, such as the comprehensive data published by Emami et al. (IJROBP 1991;21:109-122), Letchert et al.‘s small intestine tolerance data (Radiother Oncol 1990; I X:307-320) and the rectum data recently presented by Boersma and co-workers (IJROBP 1998;41:83-92). The models were also applied to determine the prescribed dose level in individual patients using both several absolute complication risk thresholds and a combination of an NTCP model with a model for tumour control. For six selected patients we determined the theoretical benefit of conformal therapy as compared to a simple two-field AP/PA pelvic technique according to the different models/parameters. Results: Even when the same clinical radiation tolerance input doses were used for model fitting. were different risk estimates predicted from the two NTCP models. The demonstrated variability between the complication models/parameters translated into signiticant discrepancies (as large as 7-10 Cy) in the recommended prescription doses. However, it was possible to discriminate between a two-field and the four-held conformal treatment plan if the same complication model and set of tolerance parameters were applied. Yet the estimated benetit of the conformal treatment in terms of permitted dose escalation varied with as much as IO-12 Gy between the different NTCP models/parameters. Conclusions: Great caution should be undertaken when using dose response models to guide clinical decisions in radiotherapy treatment of bladder cancer since different models and tolerance parameters might propose different answers to important clinical questions, such as dose prescription and scoring of rival treatment plans. Firm knowledge of the absolute dose response relations of the organs at risk is required for reliable risk predictions of normal tissue side effects.
2187
I mage feature segmentation
using model-based
multi-scale
FFT-correlation
A. L. Boyer, Y. Yan, K. Montgomery Stanford University
School of Medicinr,
Stmjbrd,
CA
Purpose/Objective: Automated segmentation of large data sets acquired by fast CT and MRI scanners is a significant problem for radiotherapy treatment planning. Automated segmentation is possible if there are image features within the data set that can be recognized by the image analysis process. This investigation tested a correlation of unclassified gray scale data in CT scans using simple mathematical correlation implemented with the fast Fourier transform (FFT). The objectives were to determine the precision of the method on test cases and to determine the feasibility of using the FFT. Materials & Methods: A model-based approach was chosen to test one-dimensional and two-dimensional cases. The tests were run on CT data sets that had been acquired for radiotherapy treatment planning. The basic approach was to use data from one patient as a reference to find a feature in test data sets. Image features used in the investigation were the edge of the trochanter in the acetabulum, the anterior wall of the bladder, and a region of the anterior surface of the kidney. For the one-dimensional cases, strips of pixels were manually selected and extracted from the reference data set and the target data set. The gray scale values within the strips were normalized to values between 0 and I Correlations were performed at scales of 64 pixels, 32 pixels and 16 pixels. After each correlation, the region of the correlation in the test data at the next lower scale was centered around the pixel containing the maximum of the previous correlation. The pixel address of the correlation maximum after the last correlation was the algorithm’s choice of the feature location. The two-dimensional cases used square regions of pixels having sides the same length as the strips. The one-dimensional and two-dimensional correlations were implemented using the FFT. Results: The multi-scale approach allowed the algorithms to find features in a second data set in spite of anatomical distortions (such as changes in organ size and rotations between the reference and the test organs). and changes in gray scale value. For the cases encountered in these tests. the large scale correlations successfully adjusted the correlation region for the next smaller scale to compensate for anatomical variations between the reference data set and the test data set. For the one-dimensional cases, the features were matched to within l-2 pixels. For the two-dimensional cases, the features were matched to within I-3 pixels. Both one-dimensional and two-dimensional tests using the FFT correlation required less than I second to execute. Conclusion: Correlation of raw, unclassihed gray scale data after manual selection of test areas is a feasible candidate for an automated segmentation algorithm. It is feasible to use the FFT so that multi-scale correlation over many feature regions can be carried our within a short period of time. This work was supported in part by grant CA43840 from the National Cancer Institute.
2 188
Three dimensional conformal radiation therapy treatment planning and delivery verificaton institutional dose escalation trials using an anthropomorphic phantom
W. B. Harms, St-..’ D. S. Davis,* R. L. Gerber,’ B. R. Paliwal.' W. Hanson,” J. A. Purdy’
J. A. Antolak,’
L. J. Verhey,’
for mulit-
J. M. Baiter,” G. C. Field,’ W. A. Tome,”
Purpose: To evaluate the veracity of three-dimensional (3D) treatment planning data used for quality assurance of multiinstitutional dose escalation trials and ultimately the appropriateness of the use of the 3D dose data in outcome assessment. Materials & Methods: A pelvic anthropomorphic phantom with geometrically well defined internal dose measurement points for both thermoluminescent dosimeters and ionization chambers was distributed to each participating institution. A total of 6 institutions were involved with this study which was performed in support of the 3DOG/RTOG Prostate 9406 3DCRT dose escalation study. Each institution acquired a CT scan of the phantom and a 3D treatment plan was generated using the CT data and a predefined beam arrangement and isocenter dose. The phantom was then set up on the treatment machine and the treatment delivered as specitied by the treatment plan. Two sets of TLDa were irradiated and one set of ionization chamber measurements was obtained. The TLD sets and the ionization data recorded from these irradiations were sent to the Radiological
Proceedings
351
of the 42nd Annual ASTRO Meeting
Physics Center in Houston, TX for analysis. The digital treatment planning data were sent to the 3DQA Center in St. Louis. MO for similar analysis and for subsequent comparison to the measured doses. Results: The deviations in doses calculated in each institution’s treatment plan and within +/-3% using the intended isocenter dose of 300 cGy and the normalization occasionally a problem between the two TLD irradiations. The variations between and those measurements for a ‘perfectly repositionable’ patient reinforces the employed in conformal planning.
the measured doses delivered were generally dose value. Institutional reproducibility was the calculated doses near the edges of a field need of the GTV/CTV/PTV concept to be
Conclusion: To the best of our knowledge, this is the first experiment of its kind, which has attempted to evaluate volumetric treatment planning data with subsequent treatment delivery in a multi-institutional setting. It provides valuable insight into the value of volumetric treatment planning data, particularly the tumor/target and normal structure contours and the corresponding volumetric dose distributions, and the appropriateness of using these data for quantitative outcome assessment. The design of the phantoms for similar studies in the future should more tightly constrain the physical attributes of the phantom to more closely match those of the protocol dose delivery requirements (e.g. with or without heterogeneities built into the phantom depending upon protocol requirements for accounting for such). This work was supported in part by grant numbers 5UOlCA60267. U24CA81647 and UlOCAl0953 from the National Cancer Institute, DHHS.
2189
D ose isotope selection matter for permanent radiobiological effectiveness
C. D. Lin.’ D. B. Leeper.” F. M. Waterman,’ ‘Jr$%r.wtt PA
Medical
College,
Philadelphia,
prostate implant an evaluation
of Pd-103 vs. I-125 based on
A. P. Dicker’
PA, ZThomc~.sJq~ersm
Hospitul
Radiatim
Onc~~log~ Dq~~wttnrr~t. Philr~drlphitr.
Purpose: Transperineal interstitial permanent prostate brachytherapy (TIPPB) has become an increasingly popular treatment for early-stage/favorable-risk adenocarcinoma of prostate. Within TIPPB, permanent implants often employ either Pd- 103 (T,,, effectiveness of Pd- 103 and I- I25 implant\ by = 17 day\) or I-125 (T,,, = 60 days). This study compared the radiobiological using the linear-quadratic model with recently published data regarding: prostate tumor cell doubling times, T,,,,,, cy and o/p ratio. Methods and Materials: The tumor potential doubling times (T,,,,) were determined based on the proliferation constants (K,) found by Berges et al. (1995) and those derived directly by Haustermana et al. (1997). The initial slope of the cell radiation dose survival curve, 01,the terminal slope B, and the (YIPratio were taken from recent published clinical and cellular results. The total dose delivered from each isotope was the dose used clinically, i.e. 120 Gy for Pd.103 and 145 Gy for I-l 25. Dale’s modilied linear-quadratic equation was used to estimate the biological effective dose (BED), the cell surviving fraction. the effective treatment time and the wasted radiation dose for different values of T,,,,. Results: The T,,,, reported for organ-confined prostate carcinomas varied from 16 to 67 days. The best estimate of 01was 0.095 Gym’ and the culp ratio was 15 Gy. T,,(,, and isotope half-life had the greatest effect on BED, surviving fraction. effective treatment time and wasted radiation dose. At short T,,, both isotopes were less effective, but Pd- 103 had much less dependence both isotopes produced similar effects. The minimum surviving fraction for exposure on T,,,,, than I-125. However, at long T to Pd.103 decreased from 140 X IO-$“;0 131 X lo-’ as the T po, increased from 16 to 67 days. By contrast for exposure to I-125, the minimum surviving fraction decreased from 3.98 X IO-’ to 198 X IO-’ over the same range of T,,,,. The range in minimum surviving fraction with change in T,,, was a factor of 200 for I-l 25 compared to a factor of I I for Pd-103. Conclusion: The theoretical prediction of effectiveness using the linear quadratic equation for the common clinically prescribed total radiation doses indicated that Pd.103 should be more effective than I- 125 since it had less dependence on T,,,,. The greatest benefit of Pd- 103 was shown to be with tumors with a short T,,,,. Although there wa\ no cross-over between Pd- IO.?and I- I25 at any particular Trot with respect to efficacy, at long Tpc,r both isotopes produced similar results. Although the regrowth delay would be longer with I- 125, the benefit was inconsequential compared to the very slow doublin, 0 times of localized prostate cancer. These findings may explain why clinically there seemed to be no clear difference in treatment outcome with either isotope. Based on these predictions, we recommend a clinical trial to compare the efficacy of the two isotopes. Maximum BED for
T,x>,(Day\)
Pd.103
(Gy)
Maximum
BED
for I-125
16
93
5x
23
102
76
2X
106
32
II3
IO0
61
I17
I12
67
11x
I I4
2 190
Monte Carlo dose calculation
J. S. Li, D. Findley, T. Pawlicki, Staford
University
for intracavitary
X5
brachytherapy
S. B. Jiang, J. Deng, D. S. Kapp, M. Marircal.
School of medcine,
Startford,
(Gy)
C. M. Ma
CA
Pm-pose/Objective: The radioactive sources together with Fletcher-Suit type afterloading applicators are widely u\ed in brachytherapy for uterine cervical cancer. The purpose of this work is to investigate the effect of the applicator geometry and patient heterogeneous anatomy on the dose distributions for brachytherapy treatment planing using the Monte Carlo method.
1. J. Radiation
Oncology
l Biology
0 Physics
Volume 48, Number 3. Supplement.
2000
Materials & Methods: NTCPs for the rectum and small intestine were calculated for 24 bladder cancer patients referred to radical four-field conformal radiotherapy using the patients’ actual dose-fractionation regime. Radiation tolerance doses from several clinical studies were applied, such as the comprehensive data published by Emami et al. (IJROBP 1991;21:109-122), Letchert et al.‘s small intestine tolerance data (Radiother Oncol 1990; I X:307-320) and the rectum data recently presented by Boersma and co-workers (IJROBP 1998;41:83-92). The models were also applied to determine the prescribed dose level in individual patients using both several absolute complication risk thresholds and a combination of an NTCP model with a model for tumour control. For six selected patients we determined the theoretical benefit of conformal therapy as compared to a simple two-field AP/PA pelvic technique according to the different models/parameters. Results: Even when the same clinical radiation tolerance input doses were used for model fitting. were different risk estimates predicted from the two NTCP models. The demonstrated variability between the complication models/parameters translated into signiticant discrepancies (as large as 7-10 Cy) in the recommended prescription doses. However, it was possible to discriminate between a two-field and the four-held conformal treatment plan if the same complication model and set of tolerance parameters were applied. Yet the estimated benetit of the conformal treatment in terms of permitted dose escalation varied with as much as IO-12 Gy between the different NTCP models/parameters. Conclusions: Great caution should be undertaken when using dose response models to guide clinical decisions in radiotherapy treatment of bladder cancer since different models and tolerance parameters might propose different answers to important clinical questions, such as dose prescription and scoring of rival treatment plans. Firm knowledge of the absolute dose response relations of the organs at risk is required for reliable risk predictions of normal tissue side effects.
2187
I mage feature segmentation
using model-based
multi-scale
FFT-correlation
A. L. Boyer, Y. Yan, K. Montgomery Stanford University
School of Medicinr,
Stmjbrd,
CA
Purpose/Objective: Automated segmentation of large data sets acquired by fast CT and MRI scanners is a significant problem for radiotherapy treatment planning. Automated segmentation is possible if there are image features within the data set that can be recognized by the image analysis process. This investigation tested a correlation of unclassified gray scale data in CT scans using simple mathematical correlation implemented with the fast Fourier transform (FFT). The objectives were to determine the precision of the method on test cases and to determine the feasibility of using the FFT. Materials & Methods: A model-based approach was chosen to test one-dimensional and two-dimensional cases. The tests were run on CT data sets that had been acquired for radiotherapy treatment planning. The basic approach was to use data from one patient as a reference to find a feature in test data sets. Image features used in the investigation were the edge of the trochanter in the acetabulum, the anterior wall of the bladder, and a region of the anterior surface of the kidney. For the one-dimensional cases, strips of pixels were manually selected and extracted from the reference data set and the target data set. The gray scale values within the strips were normalized to values between 0 and I Correlations were performed at scales of 64 pixels, 32 pixels and 16 pixels. After each correlation, the region of the correlation in the test data at the next lower scale was centered around the pixel containing the maximum of the previous correlation. The pixel address of the correlation maximum after the last correlation was the algorithm’s choice of the feature location. The two-dimensional cases used square regions of pixels having sides the same length as the strips. The one-dimensional and two-dimensional correlations were implemented using the FFT. Results: The multi-scale approach allowed the algorithms to find features in a second data set in spite of anatomical distortions (such as changes in organ size and rotations between the reference and the test organs). and changes in gray scale value. For the cases encountered in these tests. the large scale correlations successfully adjusted the correlation region for the next smaller scale to compensate for anatomical variations between the reference data set and the test data set. For the one-dimensional cases, the features were matched to within l-2 pixels. For the two-dimensional cases, the features were matched to within I-3 pixels. Both one-dimensional and two-dimensional tests using the FFT correlation required less than I second to execute. Conclusion: Correlation of raw, unclassihed gray scale data after manual selection of test areas is a feasible candidate for an automated segmentation algorithm. It is feasible to use the FFT so that multi-scale correlation over many feature regions can be carried our within a short period of time. This work was supported in part by grant CA43840 from the National Cancer Institute.
2 188
Three dimensional conformal radiation therapy treatment planning and delivery verificaton institutional dose escalation trials using an anthropomorphic phantom
W. B. Harms, St-..’ D. S. Davis,* R. L. Gerber,’ B. R. Paliwal.' W. Hanson,” J. A. Purdy’
J. A. Antolak,’
L. J. Verhey,’
for mulit-
J. M. Baiter,” G. C. Field,’ W. A. Tome,”
Purpose: To evaluate the veracity of three-dimensional (3D) treatment planning data used for quality assurance of multiinstitutional dose escalation trials and ultimately the appropriateness of the use of the 3D dose data in outcome assessment. Materials & Methods: A pelvic anthropomorphic phantom with geometrically well defined internal dose measurement points for both thermoluminescent dosimeters and ionization chambers was distributed to each participating institution. A total of 6 institutions were involved with this study which was performed in support of the 3DOG/RTOG Prostate 9406 3DCRT dose escalation study. Each institution acquired a CT scan of the phantom and a 3D treatment plan was generated using the CT data and a predefined beam arrangement and isocenter dose. The phantom was then set up on the treatment machine and the treatment delivered as specitied by the treatment plan. Two sets of TLDa were irradiated and one set of ionization chamber measurements was obtained. The TLD sets and the ionization data recorded from these irradiations were sent to the Radiological
Proceedings
351
of the 42nd Annual ASTRO Meeting
Physics Center in Houston, TX for analysis. The digital treatment planning data were sent to the 3DQA Center in St. Louis. MO for similar analysis and for subsequent comparison to the measured doses. Results: The deviations in doses calculated in each institution’s treatment plan and within +/-3% using the intended isocenter dose of 300 cGy and the normalization occasionally a problem between the two TLD irradiations. The variations between and those measurements for a ‘perfectly repositionable’ patient reinforces the employed in conformal planning.
the measured doses delivered were generally dose value. Institutional reproducibility was the calculated doses near the edges of a field need of the GTV/CTV/PTV concept to be
Conclusion: To the best of our knowledge, this is the first experiment of its kind, which has attempted to evaluate volumetric treatment planning data with subsequent treatment delivery in a multi-institutional setting. It provides valuable insight into the value of volumetric treatment planning data, particularly the tumor/target and normal structure contours and the corresponding volumetric dose distributions, and the appropriateness of using these data for quantitative outcome assessment. The design of the phantoms for similar studies in the future should more tightly constrain the physical attributes of the phantom to more closely match those of the protocol dose delivery requirements (e.g. with or without heterogeneities built into the phantom depending upon protocol requirements for accounting for such). This work was supported in part by grant numbers 5UOlCA60267. U24CA81647 and UlOCAl0953 from the National Cancer Institute, DHHS.
2189
D ose isotope selection matter for permanent radiobiological effectiveness
C. D. Lin.’ D. B. Leeper.” F. M. Waterman,’ ‘Jr$%r.wtt PA
Medical
College,
Philadelphia,
prostate implant an evaluation
of Pd-103 vs. I-125 based on
A. P. Dicker’
PA, ZThomc~.sJq~ersm
Hospitul
Radiatim
Onc~~log~ Dq~~wttnrr~t. Philr~drlphitr.
Purpose: Transperineal interstitial permanent prostate brachytherapy (TIPPB) has become an increasingly popular treatment for early-stage/favorable-risk adenocarcinoma of prostate. Within TIPPB, permanent implants often employ either Pd- 103 (T,,, effectiveness of Pd- 103 and I- I25 implant\ by = 17 day\) or I-125 (T,,, = 60 days). This study compared the radiobiological using the linear-quadratic model with recently published data regarding: prostate tumor cell doubling times, T,,,,,, cy and o/p ratio. Methods and Materials: The tumor potential doubling times (T,,,,) were determined based on the proliferation constants (K,) found by Berges et al. (1995) and those derived directly by Haustermana et al. (1997). The initial slope of the cell radiation dose survival curve, 01,the terminal slope B, and the (YIPratio were taken from recent published clinical and cellular results. The total dose delivered from each isotope was the dose used clinically, i.e. 120 Gy for Pd.103 and 145 Gy for I-l 25. Dale’s modilied linear-quadratic equation was used to estimate the biological effective dose (BED), the cell surviving fraction. the effective treatment time and the wasted radiation dose for different values of T,,,,. Results: The T,,,, reported for organ-confined prostate carcinomas varied from 16 to 67 days. The best estimate of 01was 0.095 Gym’ and the culp ratio was 15 Gy. T,,(,, and isotope half-life had the greatest effect on BED, surviving fraction. effective treatment time and wasted radiation dose. At short T,,, both isotopes were less effective, but Pd- 103 had much less dependence both isotopes produced similar effects. The minimum surviving fraction for exposure on T,,,,, than I-125. However, at long T to Pd.103 decreased from 140 X IO-$“;0 131 X lo-’ as the T po, increased from 16 to 67 days. By contrast for exposure to I-125, the minimum surviving fraction decreased from 3.98 X IO-’ to 198 X IO-’ over the same range of T,,,,. The range in minimum surviving fraction with change in T,,, was a factor of 200 for I-l 25 compared to a factor of I I for Pd-103. Conclusion: The theoretical prediction of effectiveness using the linear quadratic equation for the common clinically prescribed total radiation doses indicated that Pd.103 should be more effective than I- 125 since it had less dependence on T,,,,. The greatest benefit of Pd- 103 was shown to be with tumors with a short T,,,,. Although there wa\ no cross-over between Pd- IO.?and I- I25 at any particular Trot with respect to efficacy, at long Tpc,r both isotopes produced similar results. Although the regrowth delay would be longer with I- 125, the benefit was inconsequential compared to the very slow doublin, 0 times of localized prostate cancer. These findings may explain why clinically there seemed to be no clear difference in treatment outcome with either isotope. Based on these predictions, we recommend a clinical trial to compare the efficacy of the two isotopes. Maximum BED for
T,x>,(Day\)
Pd.103
(Gy)
Maximum
BED
for I-125
16
93
5x
23
102
76
2X
106
32
II3
IO0
61
I17
I12
67
11x
I I4
2 190
Monte Carlo dose calculation
J. S. Li, D. Findley, T. Pawlicki, Staford
University
for intracavitary
X5
brachytherapy
S. B. Jiang, J. Deng, D. S. Kapp, M. Marircal.
School of medcine,
Startford,
(Gy)
C. M. Ma
CA
Pm-pose/Objective: The radioactive sources together with Fletcher-Suit type afterloading applicators are widely u\ed in brachytherapy for uterine cervical cancer. The purpose of this work is to investigate the effect of the applicator geometry and patient heterogeneous anatomy on the dose distributions for brachytherapy treatment planing using the Monte Carlo method.