5
R ADIDTHERAPY aO~cD~DDY ELSEVIER SCIENCE IRELAND Radiotherapy and Oncology 31 (1994) 176-180 Technical note Transfer errors of planning CT to simulator: a possible source of setup inaccuracies? A. Bel*, H. Bartelink, R.E. Vijlbrief, J.V. Lebesque Radiotherapy Department, The Netherlands Cancer Institute (Antoni van Leeuwenhoek Huis), Plesmanlaan l.?I. 1066 CX Amsterdam. The Netherlands (Received 30 July 1993; revision received 17 November 1993;accepted 8 December 1993) Abstract The purpose of this study was to analyse whether the intended patient setup, based on a CT scan, was different from the setup at the simulator. Furthermore, we investigated how these possible transfer errors between the planned patient setup and the actual simulator setup affected the resulting overall treatment setup accuracy. Two groups, of 15 prostate patients each, were studied. For one group (group II), the simulation time was about twice as large as for the other (group I), since digitally reconstructed radiographs (DRRs) were used to get a good visual agreement between the intended and the simulator setup. For the purpose of this study DRRs were also calculated for the patients in group I, and for both groups DRRs were matched with the simulator images to obtain quantitative data of the transfer errors. The resulting overall treatment setup accuracy was determined by compar- ing the DRRs with portal images. For group I, the standard deviations (SD) of the differences between the DRRs and the simulator images (‘transfer errors’) were 1.5 mm and 4.5 mm in the lateral (x) and crania-caudal 0) direction, respectively. For group II the SDS were smaller: 1.4 mm and 1.5 mm in the x- and y-direction, respectively. For both groups, the magnitude of the overall mean was less than 1.3 mm. For group I, the SDS of the resulting overall setup deviations during treatment were 1.6 mm and 4.1 mm in the x- and y-direction, respectively. For group II, these figures were 2.4 mm and 2.6 mm, respectively. For both groups, the magnitude of the overall mean was less than 1.0 mm. It can be concluded that transfer errors can be the predominant factor in the treatment accuracy since the transfer errors from CT to simulator can be larger than differences between the simulator and accelerator. By a careful simulation, including the use of a DRR, the amount of transfer errors, and consequently the treatment inaccuracy, can be reduced considerably. Key wora!x Portal imaging; DRR; Patient setup accuracy 1. lntraduction Patient setup during radiotherapeutic treatment should be in agreement with the prescribed setup. By making portal im- ages, setup deviations with respect to a reference setup can be detected and consequently corrected [4,9,10,14]. When a con- ventional simulator is used to establish the intended setup, the simulator image will by definition be the proper reference for portal images. Use of a CT simulator also results in a proper * Corresponding author. reference setup when the patient is kept fixed in the intended setup during the planning [15,16]. Ambiguities of the proper reference image can arise when a CT scan for treatment planning is performed outside the radiotherapy department and the definitive beam setup is established on a simulator. The simulator image can subse- quently be used as a reference for portal images. By following this procedure, the treatment setup on the simulator is im- plicitly assumed to be equal to the intended setup based on CT-scan data. This assumption, however, should be verified. Deviations between the intended setup, based on CT data, and the simulator setup (‘transfer errors’) can be important, 0167-8140/94/$07.00 0 1994 Elsevier Science Ireland Ltd. All rights reserved. SSDI 0167-8140(93)01337-O

Transfer errors of planning CT to simulator: a possible source of setup inaccuracies?

  • Upload
    jv

  • View
    213

  • Download
    0

Embed Size (px)

Citation preview

R ADIDTHERAPY aO~cD~DDY

ELSEVIER SCIENCE IRELAND Radiotherapy and Oncology 31 (1994) 176-180

Technical note

Transfer errors of planning CT to simulator: a possible source of setup inaccuracies?

A. Bel*, H. Bartelink, R.E. Vijlbrief, J.V. Lebesque Radiotherapy Department, The Netherlands Cancer Institute (Antoni van Leeuwenhoek Huis), Plesmanlaan l.?I. 1066 CX Amsterdam. The

Netherlands

(Received 30 July 1993; revision received 17 November 1993; accepted 8 December 1993)

Abstract

The purpose of this study was to analyse whether the intended patient setup, based on a CT scan, was different from the setup at the simulator. Furthermore, we investigated how these possible transfer errors between the planned patient setup and the actual simulator setup affected the resulting overall treatment setup accuracy. Two groups, of 15 prostate patients each, were studied. For one group (group II), the simulation time was about twice as large as for the other (group I), since digitally reconstructed radiographs (DRRs) were used to get a good visual agreement between the intended and the simulator setup. For the purpose of this study DRRs were also calculated for the patients in group I, and for both groups DRRs were matched with the simulator images to obtain quantitative data of the transfer errors. The resulting overall treatment setup accuracy was determined by compar- ing the DRRs with portal images. For group I, the standard deviations (SD) of the differences between the DRRs and the simulator images (‘transfer errors’) were 1.5 mm and 4.5 mm in the lateral (x) and crania-caudal 0) direction, respectively. For group II the SDS were smaller: 1.4 mm and 1.5 mm in the x- and y-direction, respectively. For both groups, the magnitude of the overall mean was less than 1.3 mm. For group I, the SDS of the resulting overall setup deviations during treatment were 1.6 mm and 4.1 mm in the x- and y-direction, respectively. For group II, these figures were 2.4 mm and 2.6 mm, respectively. For both groups, the magnitude of the overall mean was less than 1.0 mm. It can be concluded that transfer errors can be the predominant factor in the treatment accuracy since the transfer errors from CT to simulator can be larger than differences between the simulator and accelerator. By a careful simulation, including the use of a DRR, the amount of transfer errors, and consequently the treatment inaccuracy, can be reduced considerably.

Key wora!x Portal imaging; DRR; Patient setup accuracy

1. lntraduction

Patient setup during radiotherapeutic treatment should be in agreement with the prescribed setup. By making portal im- ages, setup deviations with respect to a reference setup can be detected and consequently corrected [4,9,10,14]. When a con- ventional simulator is used to establish the intended setup, the simulator image will by definition be the proper reference for portal images. Use of a CT simulator also results in a proper

* Corresponding author.

reference setup when the patient is kept fixed in the intended setup during the planning [15,16].

Ambiguities of the proper reference image can arise when a CT scan for treatment planning is performed outside the radiotherapy department and the definitive beam setup is established on a simulator. The simulator image can subse- quently be used as a reference for portal images. By following this procedure, the treatment setup on the simulator is im- plicitly assumed to be equal to the intended setup based on CT-scan data. This assumption, however, should be verified. Deviations between the intended setup, based on CT data, and the simulator setup (‘transfer errors’) can be important,

0167-8140/94/$07.00 0 1994 Elsevier Science Ireland Ltd. All rights reserved. SSDI 0167-8140(93)01337-O

A. Bet et al. / Radiother. Oncol. 31 (1994) 176-180 177

especially when extreme accuracy is required with high-dose, high-precision treatment techniques [ 11,13,17]. Moreover, the use of decision rules for correcting patient setups [ 1) or on-line control of the setup [5,6] only makes sense when the reference image is in accordance with the intended setup.

Transfer errors can be traced by comparing a simulator image with a digitally reconstructed radiograph (DRR), representing a simulator image for the intended beam direc- tion, calculated from the CT data set. The aim of this study is to investigate the occurrence of transfer errors with or without use of the DRRs for visual comparison with the simulator image. As a second aim of the study we analysed the impact of the transfer errors on the resulting overall treatment setup accuracy.

2. Patients, materials and methods

We studied two groups, each of 15 patients, treated for pros- tate carcinoma, stage Tz,,, T, and Te Both groups of patients were treated with a 3-field technique, consisting of an anterior- posterior (AP) and a left and right lateral field.

2.1. Treatment planning and simulation

Treatment planning of both patient groups started by mak- ing a CT scan of the pelvic region with the patients lying in the supine position. The patients were positioned by means of lat- eral longitudinal, midline longitudinal and cross-table hori- zontal laser beams. The laser beams were used to determine the position of radio-opaque catheters and skin markers, which were placed on the patient, to indicate the tentative position of the isocentre. The same lasers were used to establish the exact location of the y = 0 slice of the CT data set, the y-axis corresponding to the crania-caudal direction.

CT information was entered into the three-dimensional (3D) planning system (Scandiplan) and the definitive isocentre was defined relative to the tentative isocentre in the y = 0 plane. After the planning procedure, a simulation of the intended treatment plan was performed. The position of the definitive isocentre was found with respect to the temporary position of the isocentre during CT scanning, using translations of the table in 3 directions which were determined during the treat- ment planning process.

2.2. Portal images

Portal images were obtained with an electronic portal imag- ing device (EPID), developed in our institution [12,18,19]. Portal images of the 3 fields were made about twice a week during the course of the treatment.

For this study, we only considered anterior-posterior (AP) field images. No on-line control of the patient setup during the irradiation was performed; all portal images were analysed after the fraction was delivered.

2.3. Match procedure

The simulator images were digitized and read into a compu- ter. With a drawing editor, a drawing of bony structures and the outline of the field was made on the simulator image [2].

In the portal image, the field edges were detected automatical- 1Y 131.

The field outlines of the simulator and the portal images were aligned. The setup deviation of the patient with respect to the field edge was determined by matching automatically the simulator drawing to the bony structures on the portal image [7]. When necessary, the automatic anatomy match was corrected manually.

The measured setup deviation consisted of a 2-D translation and a rotation. We defined an x- and a y-coordinate, cor- responding to the lateral and the crania-caudal directions, respectively. We did not study rotations, since they were small (the average deviation was 0” with an SD of 1” for both groups, measured with the portal-simulator image match). These small rotations do not result in significant deviations at the border of the simultaneous conformal AP boost fields, which are approximately circularly shaped. Therefore, these small rotations were not considered to be important for the conformal simultaneous boost region, although they may lead to significant deviations at the border of large rectangular fields.

The DRR was matched to the simulator image, using the procedure described for a portal image; this match was com- pletely performed manually. By the subsequent match of the anatomy drawing to the bony anatomy in the DRR, transfer errors between the intended and the actual simulator setup could be determined.

The resulting overall treatment setup accuracy, defined by the difference between the intended setup and the patient’s mean setup during treatment, was determined by adding the transfer errors to the deviations found by the portal to simulator image match.

2.4. Treatment procedures for patient groups I and II

For patient group I, our conventional treatment technique was applied. These patients received 20 fractions of 2 Gy with rectangular pelvic fields and an additional 15 fractions of 2 Gy with boost fields.

During CT scanning, the slice thickness was 5 mm and the spacing was 10 mm. The DRR was calculated retrospectively for this study. The total available simulation time was 45 min. During the simulation, CT-scan information could be used to check if the resulting position of the isocentre was in accord- ance with the intended setup.

For the portal image analysis, only the images of the AP rec- tangular pelvic fields were used. We obtained approximately 8 AP portal images for each patient in this group and the total number of portal images analysed in this study was 118.

The other patient group (group II) was treated with the simultaneous boost technique [ 111, where the boost fields were given simultaneously with the large fields.

During CT scanning, the CT-slice thickness was 3 mm and the spacing was 5 mm. The DRR was calculated prior to the simulation. The available simulation time was 90 min in order to enable a careful check of the setup with the DRR. The simulator image and the DRR, having the intended field outline and central axis position, were compared visually and the position of the isocentre with respect to the bony anatomy was adapted, if judged necessary.

178

15 A 15 B

t t Y Y

bm) km)

I.... I I I.... I. -15 x (mm)+ 15 -15 x (mm)+ 15

Fig. I. Transfer errors, determined by matching the digitally recon- structed radiograph to the simulator image for patient group I (A) and patient group II (B). The x- and Y-directions correspond to the lateral and craniotaudal directions, respectively. The shaded region indicates

the 95% confidence region.

For the portal image analysis, the outline of the AP rec- tangular pelvic field was used. About 11 portal images of each patient were analysed, with a total number of 171.

2.5. Accuracy of the match procedure

The precision of the DRR calculation was checked by means of a phantom. The phantom was CT-scanned, utilizing the CT-scan spacing and thickness of either patient group. DRRs were calculated and matched with the phantom’s simulator image. No systematic errors were found in the DRRs.

The accuracy of the match of a DRR to a simulator image was determined by repeating the match for one patient from each group ten times. The visual quality of the DRRs was about equal in each group. Therefore, the match accuracy determined for one DRR was representative for the whole group. The distribution of transfer errors can be corrected for this match accuracy by using the relation ezlO, = e*,,, + cZle for the x- and y-directions separately. Here, a,,, is the total stan- dard deviation, u,,, is the standard deviation of the match accuracy and a,, is the resulting standard deviation of the transfer error after correcting for the match accuracy.

The accuracy of the procedure of matching simulator images to portal images for AP pelvic fields was analysed by Gilhuijs

A. Bel et al. /Radiother. Oncol. 31 (1994) 176-180

et al. [8]. Assuming that the out-of-plane rotations are smaller than 1.2”, the translation vector can be determined without a systematic error [2]. The accuracy (1 SD) of the measurement of each component of the translation vector was determined to be 0.3 mm in the x-direction and 0.5 mm in the y-direction [8].

3. Results

3. I. Transfer errors

The inaccuracy of the match procedure of the DRR to the simulator image for group I was found to be larger in the y- direction (0.7 mm, 1 SD) than in the x-direction (0.2 mm, 1 SD). For the other group, the precision of the match for both directions was about equal: 0.3 mm and 0.4 mm (1 SD) in the x- and y-directions, respectively.

The DRRs were matched to the simulator images (Fig. 1 and Table 1). For both groups, no significant correlation be- tween the deviations in the x- and y-directions (which would result in a tilted ellipse) was found. The overall means of the deviations of both groups were not significantly different from zero (Table 1).

The distribution of the transfer errors of group I showed a smaller dispersion in the x-direction (1.5 mm, 1 SD) than in the y-direction (4.5 mm, 1 SD). For group II, the standard devia- tions were approximately equal for both directions, 1.4 mm and 1.5 (1 SD), respectively.

We recalculated the standard deviation of both distributions by taking into account the match accuracy. These corrections were, however, so small that it did not alter the numerical results.

3.2. Overall accuracy of the treatment

The portal images were matched to the simulator images. An approximation of the systematic deviation between simula- tion and treatment setup during the whole treatment of each patient was obtained by calculating the patient’s average deviation from the daily deviations (Fig. 2) 131. One patient setup from group II was corrected after two fractions. For this patient, the overall average during the whole treatment was used. There was no signilicant correlation between the devia- tions in both directions for both groups. The overall means of

Table I The setup accuracy for patient groups I and II

Group I Group II

Transfer error Systematic Overall Transfer error Systematic Overall (CT-sim) deviation (CT-treatment) (CT-sim) deviation (CT-treatment)

(sim-treatment) (sim-treatment)

I SD (mm) x I.5 1.8 I.6 1.4 1.8 2.4 Y 4.5 I.4 4.1 I.5 2.6 2.6

Mean (mm) X 1.3 -0.2 I.1 1.0 -0.8 0.2 Y -0.4 -0.4 -0.8 -0.3 0.7 0.4

x and y correspond to the lateral and crania-caudal directions, respectively.

A. Bel et al. / Radiother. Oncol. 31 (1994) 176-180 179

15 A 15 n

t t Y

(mm) I-+ Y

(mm)

L....I..... I... I...., -15 x (mm)+ 15 -15 x (nn)+ 15

Fig. 2. The overall accuracy determined by matching the portal images to the simulator images for group I (A) and group II (B). Each dot indicates the average (systematic) deviation during the treatment. The

shaded region indicates the 95% confidence region.

both groups were not significantly different from zero (Table 1). For group I, the standard deviations for the x- and y- directions were 1.8 mm and 1.4 mm, respectively. For group II, the standard deviations were 1.8 mm and 2.6 mm, respec- tively.

By adding the transfer errors to the setup deviations as ob- tained from the simulator-portal comparison, the deviation re- lative to the intended setup was obtained (Fig. 3, Table 1).

Again, for both groups, there was no significant correlation between the deviations in the x- and y-directions and the overall means of the deviations for both directions were not significantly different from zero. For group I, the standard deviation in the x-direction was considerably smaller than in y-direction (1.6 mm and 4.1 mm, respectively). For group II, the standard deviations were about equal for both directions (2.4 mm and 2.6 mm, respectively).

4. Discussion

By comparing the DRR with the simulator image, we in- vestigated whether the setup on the simulator was equal to the intended setup in two groups of prostate patients: conven- tionally treated (I) and treated with the simultaneous boost technique (II). For the latter group, the DRR was used during the simulation to get a good visual agreement between setup

15 15 n

t t Y Y

(nn) (nn)

Fig. 3. The resulting overall treatment setup accuracy (the difference between intended and actual setup) after adding the transfer errors (Fig. I) to the results of the simulator to portal image match (Fig. 2) for group I (A) and group II (B). The shaded region indicates the 95%

confidence region.

on the simulator and the intended setup. For both groups, we found patients with differences (‘transfer errors’) between the setup on the simulator and on the intended setup.

A possible drawback of this retrospective study was the dif- ferent treatment protocol in both groups, especially the dif- ferent CT-slice thickness and spacing for the calculation of DRRs. The accuracy in the position of the y = 0 CT plane was not different for the two groups, since we used laser beams at CT scanning to localize this plane. The match accuracy in the y-direction, however, depended on the CT-slice spacing and thickness. As expected, this accuracy was about two times bet- ter for group II (0.4 mm) than for group I (0.7 mm). For both groups, this match accuracy was better than the corresponding pixel resolution, since all visible bony structures in the DRR were used in the matching procedure.

For group I, the standard deviation in the transfer errors in the cranio-caudal (y) direction was about three times larger than in the lateral (x) direction. An explanation can be found in the simulation procedure for this patient group. The defini- tive isocentre could be checked and eventually corrected by inspecting CT slices. During this procedure, the middle of the pubic symphysis could be used as a marker, resulting in accurate positioning in the x-direction. There was no such easy anatomical marker for the y-direction.

For group II, the standard deviation of the transfer errors in the y-direction was smaller than for group I. We tested if this observation was solely due to the use of the DRR during the simulation or if other factors played a role as well. Possible corrections of the (definitive) isocentre were performed after a visual comparison of the simulator image with the DRR. We determined the distribution of transfer errors before these cor- rections were performed, resulting in average deviations of -0.5 mm and -0.4 mm with standard deviations of 1.5 mm and 1.9 mm in the x- and y-directions, respectively. Comparison of these figures with the results in Table 1 shows that the use of the DRR resulted in a small decrease of the transfer errors in the y-direction but the transfer errors were not as large as for group I. Thus, for group II, the main cause of the reduction in the transfer errors was probably the more careful position- ing of the patients by the radiographers before the setup was checked with the DRR. The more careful patient positioning was enabled by the longer available simulation time and, moreover, the radiographers were aware of the ensuing check of the setup. Currently, the simulation procedure for group II is performed more routinely and the additional required simu- lation time is decreasing. However, it is obvious that, if for group I a DRR would have been used during the simulation, large transfer errors could have been detected, without additional simulation time.

As could be expected, the overall setup accuracy during the treatment turned out to be worse after adding the transfer er- rors to the deviations found by comparing simulator with por- tal images (Table 1). For group I, transfer errors from CT to simulator in the y-direction appeared to be the predominant factor in overall treatment setup inaccuracies, since the stan- dard deviation increased from 1.4 to 4.1 mm after adding the transfer errors. DRRs should be used as a reference for check- ing portal images rather than simulator images, which may in- clude some transfer errors. If required, the setup during treatment can be adjusted further, using a setup verification procedure [ 1).

180

5. Conclusions

Transfer errors relative to the intended setup, based on CT scan data, at the simulator setup can affect the resulting overall treatment accuracy considerably. These transfer errors can contribute more to treatment inaccuracies than do devia- tions from the simulator to the accelerator setups. A more careful simulation procedure, consisting of a prolonged simu- lation time, more attention by the personnel and a visual check on the simulator by comparing the setup with a digitally reconstructed radiograph (DRR), increases the setup accu- racy. The best agreement between the intended setup, based on CT data, and the actual treatment can be obtained by using the DRR directly as a reference for portal images.

6. Acknowledgements

We would like to thank Albert Blom and Ton Minderhoud for their assistance during the use of the planning system. We acknowledge Dr Marcel van Herk and Kenneth Gilhuijs for assistance with hardware and software problems and Dr Ben Mijnheer for useful suggestions during the preparation of the manuscript. This study was supported by the Dutch Cancer Society, NKB Grant NKI 91-01.

References

Bel, A. van Herk, M. Bartelink, H. and Lebesque, J.V. A veriti- cation procedure to improve patient setup accuracy using portal images. Radiother. Oncol. 29: 253-260, 1993. Bijhold, J. van Herk, M. Vijlbrief, R. and Lebesque, J.V. Fast evaluation of patient setup during radiotherapy by aligning fea- tures in portal and simulator images. Phys. Med. Biol. 36: 1665-1679, 1991. Bijhold, J. Gilhuijs, K.G.A. van Herk, M. and Meertens, H. Ra- diation field edge detection in portal images. Phys. Med. Biol. 36: 1705-1710, 1991. Bijhold, J. Lebesque, J.V. Hart, A.A.M. and Vijlbrief, R.E. Max- imizing setup accuracy using portal images as applied to a con- formal boost technique for prostatic cancer. Radiother. Oncol. 24: 261-271, 1992. De Neve, W. Van den Heuvel, F. De Beukeleer, M. Coghe. M. Thon, L. De Roover, P. Van Lancker, M. and Storme, G. Routine clinical on-line portal imaging followed by immediate

6

7

.8

9

IO

II

I2

I3

I4

IS

I6

I7

I8

I9

A. Bel er al. / Radiother. Oncol. 31 (1994) 176-180

field adjustment using a tele-controlled patient couch. Radiother. Oncol. 24: 45-54, 1992. Ezz A. Munro, P. Porter, A.T. Battista, J. Jaffray, D.A. Fenster, A. and Osborne, S. Daily monitoring and correction of radiation field placement using a video-based portal imaging system: A pilot study. Int. J. Radiat. Oncol., Biol. Phys. 22: 159-165, 1992. Gilhuijs, K.G.A. and van Herk, M.B. Automatic on-line inspec- tion of patient setup from digital portal images. Med. Phys. 20: 667-677, 1993. Gilhuijs, K.G.A. El-Gayed, A.H. van Herk, M.B. and Vijlbrief, R.E. An algorithm for automatic analysis of portal images: clini- cal evaluation for prostate treatments. Radiother. Oncol. 29: 261-268, 1993. Grifliths, SE. Khoury, G.G. and Eddy, A. Quality control of radiotherapy during pelvic irradiation. Radiother. Oncol. 20: 203-209, 1991. Huizenga, H. Levendag, C. De Porre M.Z.R. and Visser, A. G. Accuracy in radiation field alignment in head and neck cancer: a prospective study. Radiother. Oncol. II: 181-188, 1988. Lebesque, J.V. and Keus, R.B. The simultaneous boost tech- nique: the concept of relative normalized total dose. Radiother. Oncol. 22: 45-55, 1991. Meertens, H. van Herk M. Bijhold, J. and Bartelink, H. First clinical experience with a newly developed electronic portal im- aging device. Int. J. Radiat. Oncol., Biol. Phys. 18: 1173-1181, 1990. Mohan, R. Mageras, G.S. Baldwin, B. Brewster, L.J. Kutcher, G.J. Leibel, S. Burman, C.M. Ling, CC. and Fuks Z. Clinically relevant optimization of 3-D conformal treatments. Med. Phys. 19: 933-944, 1992. Rabinowitz, I. Broomberg, J. Goitein, M. McCarthy, K. and Leong, J. Accuracy of radiation field alignment in clinical prac- tice. Int. J. Radiat. Oncol., Biol. Phys. II: 1857-67, 1985. Ragan, P.R. He, T. and Mesina, C.F. Ratanatharathorn, V. CT- based simulation with laser patient marking. Med. Phys. 20: 379-380, 1993. Sherouse, W.G. and Chaney, E.L. The portable virtual simulator. Int. J. Radiat. Oncol., Biol. Phys. 21: 475-482, 1991. Ten Haken, R.K. Perez-Tamayo, C. Tesser, R.J. McShan, D.L. Fraas, B.A. and Lichter, A.S. Boost treatment of the prostate using shaped, fixed fields. lnt. J. Radiat. Oncol., Biol. Phys. 16: 193-200, 1989. Van Herk, M. Physical aspects of a liquid-tilled ionization chamber with pulsed polarizing voltage. Med. Phys. 18: 692-702, 1991. Van Herk M. and Meertens, H. A matrix ionization chamber imaging device for on-line patient setup verification during radiotherapy. Radiother. Oncol. I I: 369-378, 1988.