Radiotherapy Elsevier RADION

and Oncology,

19 (1990) 73-87

73

00759

Quality assurance in radiotherapy by in vivo dosimetry. 2. Determination of the target absorbed dose G. Leunens,

J. Van Dam, A. Dutreix and E. van der Schueren

Department of Radiotherapy,

(Received

19 December

University Hospital St. Rafael, Leuven, Belgium

1989, revision received 9 April 1990, accepted 9 April 1990)

Key words: In vivo dosimetry;

Quality assurance;

Silicon diode, Exit dose

Summary Combined entrance and exit dose measurements were performed with semiconductor detectors on patients, treated for neck and oral cavity malignancies. Transmission measurements showed the important influence of contour inaccuracies and tissue inhomogeneities. In 39.6% (21/53) of the checked contours, the discrepancy between the contour diameter used for routine treatment planning and the actual patient diameter was 1 cm or more, and in this group a systematic tendency for patient diameter underestimation due to the procedure was detected. When the X-ray beam passed through important bone structures such as the mandibular bones or the vertebral body, large discrepancies of 10% and more between the measured and the expected transmission were found. The target absorbed dose was determined from the transmission and entrance dose measurement. A systematic underdosage of about 2% at midline level was found to be due to an inaccuracy in the algorithms of the treatment planning system. Underdosages of 5% or more at midline were detected in more than 20% (47/230) of the measurements. In all cases, the reason for erroneous dose delivery was identified. Entrance dose measurements were previously demonstrated to be useful for the assessment of uncertainties related to treatment machine, patient set-up and treatment planning system (part 1). Transmission measurements (the ratio of the exit to the entrance dose measurement) are shown to be very useful to evaluate uncertainties related to patient data such as contour errors and tissue inhomogeneities as well as to the algorithms of the planning system. The influence of these errors on the target absorbed dose can be estimated and corrections can be applied for each individual patient.

Address for correspondence:

G. Leunens,

Department

of Radiotherapy,

University

Leuven, Belgium. 0167~8140/90/%03.50 0 1990 Elsevier Science Publishers

B.V. (Biomedical

Division)

Hospital

St. Rafael, Capucijnenvoer

33,

74 Introduction The administration of a course of radiotherapy requires a whole series of steps going from basic dosimetry over tumor localisation, treatment planning, dose calculation, to patient immobilisation and multiple daily irradiations. The precision of a radiation treatment is essentially related to the exactitude of the daily delivered dose, given over the predetermined volume. Each step involved in the planning or the accomplishment of a treatment is subject to a certain degree of uncertainty leading to a cumulative discrepancy between “prescribed” and “delivered” dose [ 1,4,7,19]. In addition to these inherent uncertainties, there are errors which can occur either on a systematic basis or on a random basis. These errors and uncertainties can be the direct cause of complications or treatment failures because there exists a narrow relationship between probability of local tumor control or normal tissue injury and absorbed dose [5,6,11,18]. Therefore, in vivo dosimetly is a vital part of a departmental quality assurance program. In vivo dosimetry was already performed in 1932 by Sievert with small ionisation chambers and thermoluminescent dosemeters have been used in the sixties and seventies [3,10,17]. At present, semiconductor detectors are commercially available as radiation dosemeters and provide a good alternative for thermoluminescent dosemeters [8,9,13-151. The main advantage of silicon diodes, is that there is no time delay between measurements and results [ 121. When a discrepancy is found between the actually delivered dose and the expected dose, the immediate availability of the results of the in vivo dose measurements makes it possible to check the actually applied patient set-up to detect the sources of the error enabling their correction. In vivo dose measurements can consist of entrance dose, exit dose, intracavitary dose measurements and the determination of the dose delivered to critical organs such as the eyes or the gonads [ 171. In part 1 [9], we demonstrated that entrance

dose measurements are useful: firstly, as a check of the dosimetric aspects of the whole treatment chain, secondly, to evaluate the quality of specific treatment techniques and finally, to control the accuracy of dose delivery for each individual patient, mainly aimed at screening for random and systematic errors. In part 2, our interest will be focused on exit dose measurements. Only very few publications refer to this subject. This is probably due to several factors. Firstly, the calibration and calculation methods needed to convert the dosemeter reading in absorbed dose are not clearly established. Secondly, the estimation of the tumor absorbed dose from the combined entrance and exit dose measurements requires a well defined methodology for the different treatment situations met in clinical practice. The goal of the study was to investigate the feasibility and usefulness of measurements of the exit dose with semiconductor detectors in daily clinical practice. The calculation procedure and the algorithms needed to convert the dosemeter reading in absorbed dose are described. The methodology to assess the tumor absorbed dose from combined entrance and transmission measurements is explained. The results of combined entrance and exit dose measurements, performed on patients treated for neck and oral cavity malignancies are discussed. The absorbed tumor dose is estimated and the influence of different kinds of parameters (algorithms, contours, inhomogeneities) is investigated.

Material and methods I. Equipment The same detector system has been used as in part 1, i.e. EDP-10 semiconductor detectors connected to a DPD-6 electrometer (Scanditronix) and a 0.6 cm3 thimble chamber (NE 2505/3B) connected to a 2500/3 Ionex electrometer (Nuclear Enterprises) for calibration.

75

All the measurements have been performed in a 6 MV X-ray beam produced by a Mevatron 6700 Siemens linear accelerator. (The Quality Index [ 20]- defined by the ratio 120/I 10 - of the 6 MV X-ray beam is 0.67.) The treatment unit is supplied with an automatic verification system (Mevamatic - Siemens) to check treatment set-up parameters such as field size, wedges, gantry and collimator rotation. The treatment couch parameters are not automatically checked. II. Absorbed dose measurement 1I.a. Calibration procedure The procedure used for entrance dose calibration is described in part 1 [9]. The calibration was performed with the diodes positioned on the surface of a polystyrene phantom with a water equivalent thickness of 15 cm at the center of a 10 x 10 cm* field at SSD 100 cm. The entrance dose calibration factor (F,,,,,,) was then determined as the ratio of the absorbed dose measured with the ionisation chamber, D,, at the depth of dose maximum (1.5 cm, Fig. 1) and the semiconductor signal, M,,, in reference conditions.

Fcal,en

=

The exit dose, D,, is here defined as the dose at - 1.5 cm from the exit surface (Fig. 1). In a similar way the exit dose calibration factor was determined, i.e. Fcal,ex = (D2/MJcal. The semiconductor detector was now placed with an 180’ angle at the exit surface. This methodology presents several advantages. Firstly, the geometrical symmetry of the entrance dose and the exit dose measurement points simplifies the procedure in the frequent case of two opposed coaxial beams. The use of symbols: The authors use, in this present paper, the symbols in agreement with general practice and as recommended by the ICRU. In the first part of the paper [9] some of the symbols were different from the present ones: CF instead of C, MD instead of DMM,en and ED instead of L&.

OD,

(Z-3)cm

(3D2

J.

2

1,5cm

; Fig. 1. Determination of the transmission. The exit transmission is defined as the ratio of the dose measured at - 1.5 cm from the exit surface (D, = exit dose) and the dose measured at 1.5 cm from the entrance surface (0, = entrance dose or maximum dose). The total water equivalent thickness of the phantom is z and the two different depths are separated by (z-3) cm of water.

Secondly, the depth between the exit reference point and the exit surface remains constant (1.5 cm). This method is justified because a depth of 1.5 cm is largely sufficient to assure conditions of electronic equilibrium in the backward direction. However, as this depth is insufficient to provide complete photon backscatter at 6 MV, an additional correction factor has to be taken into account (see Section V). The ratio between the exit dose calibration factor and the entrance dose calibration factor F c4e.x lLQ_n 3 remains constant in time (monthly check within 1%). However, this ratio for diode number 1 (1.068)

76 shows a significant difference as compared to the ratio’s for diodes number 2, 3 and 4 (1.10-l. 12) (Table I). We have no definite explanation for this feature but we suspect that this is due to the manufacturing of the diodes: diode no. 1 being from a different manufacturing series from diode no. 2, 3 and 4. This is certainly an indication that calibration factors and correction factors have to be determined for each individual diode. The weekly exit dose calibration factor (Fca,& can then be derived from the weekly entrance dose calibration factor (F_i,&, when the initial determination of the ratio has been performed. 1I.b. Correctionfactors Correction factors have to be determined when the irradiation geometry differs from the reference geometry used in the calibration. The exit correction factor C,, is defined by

Here D, and MS, are measured in the geometry of interest. The factor C,, was measured as a function of the water equivalent thickness (z), the collimator opening, and the presence of wedge and tray for each individual diode. For all four diodes, the exit correction factor, C,,, for the use of wedge and tray was found to be equal to 1 and was therefore neglected. The variation of the correction factors as a function of the water equivalent phantom thickness (C,._)

TABLE

I

Calibration

factor ratio for the different diodes.

Diode identification

Fca,.exlF=a,,en

1 2 3 4

1.06, 1.11, 1.11, 1.10

The ratio of the exit dose calibration factor (Fcal,ex) and the entrance dose calibration factor (Fed,,,) is determined for the four different diodes F_,,ex/Fca,,en.

and collimator opening (Ce.,_,,) was less than 1% but the corresponding correction factors have been applied. After determination of C,,, the diode signal can be converted in absorbed dose by the following equation :

D2= Kc x Fd,

ex

x wex,z x cx,col).

The influence of temperature and dose rate on the diode signal have been studied and the relative correction factors have been found to be negligible for the treatment conditions in this study. The directional dependence of the diode signal has been neglected because the silicon diodes were always used on a surface perpendicular to the beam axis. III. Measurement procedure Measurements have been performed on a fixed day once a week and on each treatment set-up both entrance and exit doses have been measured for each treatment field. The entrance diode was positioned in the center of the irradiation field on the skin of the patient. The exit diode was not positioned exactly along the beam axis passing through the entrance diode but was 1 cm displaced from the beam axis (in a direction perpendicular to the wedge gradient if any) in order to avoid the attenuation by the entrance diode. The attenuation effect has been measured and found to be about 4% at a depth of 10 cm. This is similar to what was observed by Nilsson et al. [ 121 who found the same attenuation effect of 5 % over an area of 1 cm2 at 5 cm depth in 4 MV X-rays (diode type EDP-8) and 21 MV X-rays (diode type EDP-20). The patient diameter on the beam axis was checked with a caliper for each measured treatment set-up. When large deviations were encountered between measured and expected dose, all the treatment parameters were checked with the patient still on the treatment couch and the simulation film was checked to estimate the possible importance of tissue inhomogeneities.

77 IV. Prescribed and expected midline dose In daily clinical practice, the number of monitor units required to deliver the prescribed dose (Dp,mid) is calculated with the algorithms of the treatment planning system. Our computer software, which is 10 years old, does not take into account scatter defects due to the use of shielding blocks or missing tissue when part of the field is in air (cf. part 1). This means that the monitor units are correct only in conditions without scatter defects. In most of the head-and-neck cancer treatments either shielding blocks are used or part of the field is in air, introducing a discrepancy between the dose delivered and the prescribed dose for the previous reason. Phantom measurements were performed to study the importance of the scatter defects in the 6 MV X-ray beam. The overestimation of the dose at the field center varied between 0.5 and 1% at the entrance point and was about 2% at midline depth, depending on the extent of shielding blocks and missing tissues. Therefore, the target dose was manually recalculated from the monitor units with new algorithms taking into account scatter defects [2]. The corrected target dose is called the expected dose (DE,+J.

V. Calculation of the expected entrance and exit dose

For each patient, expected entrance (D& and exit doses (DE,_) were calculated from the monitor units, taking into account the scatter defects as explained before. The exit dose, D,, is not measured in complete photon backscatter conditions. Therefore, a backscatter correction factor (B’) was determined as the ratio of the ionisation chamber reading in full backscatter conditions (RFB) and the ionisation chamber reading in exit dose measurement conditions (R,,) for different collimator opening (Fig. 2). B’=?.

R R MC

1.0&L

10

5

15

20

25

3ocmA

Collimator opening "C"

Fig. 2. Backscatter correction factor. The backscatter correction factor (B’) is plotted as a function of the collimator opening (c) in cm. B’ is the ratio of the ionisation chamber reading with full backscatter (I&) and the ionisation chamber reading in exit dose measurement conditions (R,,).

*‘=+. MC

The expected exit dose, DE_, is then determined as the ratio of the calculated exit dose and the correction factor for full backscatter loss (actual field size). D E/S

=

calculated

exit dose

B’



VI. Determination of the transmission From the combined entrance and exit dose measurements, the transmission was calculated. The transmission is here defined as the ratio of the dose at the exit and entrance depths, which are separated by a thickness of (z-3) cm of water equivalent tissues, z being the total water equivalent thickness (Fig. 1). When calculating the dose distribution, the patient tissues are assumed to be

78 water equivalent since inhomogeneities are not taken into account for dose calculation and planning in patients treated for neck and oral cavity malignancies. The expected transmission (T,$ was determined as the ratio of the expected exit dose and the expected entrance dose, i.e. TE = D, JDE en. The measured transmission (TM) was deiermiied as the ratio of the measured exit dose and the measured entrance dose, i.e. TM = D, .JDM en. The transmission is a factor which’ depeids only slightly on the patient set-up and the accelerator parameters. It depends strongly on patient thickness, tissue inhomogeneities and the actual field size.

VII. Estimation of the target absorbed dose VI1.a. Target absorbed dose determination The target absorbed dose can not be directly measured, because it is difficult to position accurately diodes in the oral cavity or the larynx region of the patient. Nevertheless, it is possible to estimate the target absorbed dose from combined entrance and exit dose measurements. The method used to estimate the target absorbed dose from in vivo entrance and exit dose measurements is similar to the method described by Rizzotti et al. [ 161 for estimation of the midline absorbed dose. In most of the patients treated for neck and oral cavity cancer, the dose is prescribed

100 ‘;i

2

80

._S : 60 ._ E 2 50 m e

40

5

10 Water

15 equivalent

20

25

15

cm Z/2

30

cm Z

thickness

Fig. 3. Transmission curves. The transmission is plotted as a function of water equivalent thickness. Two groups of curves are given: the exit transmission curves (T,,) as a function of the total water equivalent thickness (z) in cm. Secondly, the midline transmission curves (Tmid) as a function of the water equivalent midline depth (z/2) in cm. Both groups consist of curves for different equivalent square fields (5 x 5 to 20 x 20 cm*). T,, and r,, have been calculated from tissue phantom ratios taking into account the actual position of the calibration points and the lack of backscatter.

79 in the midline where the target volume is located. Therefore, the target absorbed dose is equal to the midline absorbed dose (Amid) for these tumor localisations. The method for the estimation of the target absorbed dose consists of: - firstly, the determination of the measured exit transmission (TM) for a given patient; - secondly, the midline transmission ( Tmid) for the given patient is then derived from the measured exit transmission with the help of two groups of curves (Fig. 3). - finally, the absorbed midline dose (Dmid) can be estimated from the product of the measured entrance dose (DM,en) and the midline transmission ( Tmid). VI1.b. Calculation of transmission curves The exit and midline transmission curves needed for the derivation of the midline transmission for a given patient from the measured exit transmission (TM) are plotted in Fig. 3. The exit transmission (T,,) and the midline transmission ( Tmid) curves are theoretical curves calculated from tissue phantom ratio’s (TPR, and TPR,,) for square fields (A, A’) varying from 5 x 5 cm2 to 20 x 20 cm2 as a function of either the patient thickness (z) or the midline depth (z/2).

T,, =

TPR,,,,

TPR,,,,

z-1.5) X 1.5)

Tmid= TPR,A, 0) TPR,,,

X-

1.5)

x

1

100 - z/2 + 1.5 2 x 100 + z/2 - 1.5

1

100 - z/2 + 1.5 2 x 100

BA

B A0

The TPR ratio’s are corrected for inverse square law taking into account the isocentric technique with a SAD of 100 cm. B,, B,, and B,, are the backscatter factors for the field sizes A, A’ and Ao, respectively at the SAD (100 cm), at the

exit dose point (z - 1.5 cm) and the entrance dose point (1.5 cm). A’ = A x 100 + z/2 - 1.5 100

Ao=Ax

100 - z/2 + 1.5 100

B’ is the correction factor for backscatter loss. The curves are used as follows: for a transmission measured on a given patient (TM) the corresponding total water equivalent thickness of the patient (z) can be read out from the T,, curve for the appropriate equivalent square field A’ (e.g. Fig. 3). As an example, when TM is 50% and the actual field size A at 100 cm is 10 x 10 cm2: for a measured patient thickness of 14 cm, the field size A ’ = 10.5 x 10.5 cm2 and the equivalent total thickness z is equal to 16 cm (Fig. 3: arrow 1 and the abscissa of arrow 2). The water equivalent thickness of the patient is then known and the midline transmission is determined as the read-out on the y-axis of the Tmid curve for the 10 x 10 cm* field size (A) (e.g. z is 16 cm, T,, is 72% -Fig. 3: HOW 2 and HOW 3). The absorbed midline dose (Amid) in the patient can then be estimated from the product of the measured entrance dose &,,) and the determined midline transmission ( Tmid).

In this study, the midline dose (Dmid) delivered to the patient is estimated from in vivo measurements as explained and compared: - firstly, to the expected midline dose (DE,&, calculated as explained in Section IV, - secondly, to the tumor dose prescribed at midline (DP,mid)* The method described could be used to estimate the absorbed dose at any given depth. The given depth has to be converted in water equivalent

80 depth (w). A depth correction factor (FJ is determined as the ratio of the water equivalent patient thickness (z) estimated from the measured transmission and the patient thickness estimated from the body contour (d). F, = ‘. d

The water equivalent depth (w) of any given depth (c) is then equal to the product of the given depth and the depth correction factor (FJ. w = c x Fd.

The transmission at depth w can be read out of the Tmid curves for the equivalent square field at depth c, as the intersection of arrow 2 (Fig. 3) and the X-axis corresponding to the water equivalent depth w. There is one important restriction inherent to this method: tissue inhomogeneities should be symmetrical and equally distributed for reliable absorbed dose determination. Therefore, we consider this method not suitable for absorbed dose determination in the thorax and facies. We have shown that it may be useful for absorbed dose determinations in the breast, brain, neck and oral cavity.

Results Combined entrance and exit dose measurements have been performed on 34 patients treated for neck and oral cavity malignancies. A total number of 230 treatment set-ups (83 treatment fields) have been measured. All these patients were treated isocentrically on the 6 MV linear accelerator. The treatment technique was extensively described in part 1 [ 91. The in vivo measurements discussed in this study are entrance and exit dose measurements performed on the lateral treatment fields. Combined entrance and exit dose measurements were not performed on cervical anterior fields. The results of the entrance dose measurements,

transmission measurements and midline dose determinations are plotted in histograms as their respective ratios: DM,en/DE,en, TM/T,, Dmid/ DP,mid9 Dmid/DE,mid * The expected doses (DE,en, TE, Dmid) have been manually calculated with algorithms taking into account scatter defects (Material and Methods, Section IV, V and VI). The systematic error due to the algorithms of the treatment planning system, will thus be reflected in the results of the Dmid/DP,mid ratios but not in the histograms showing the results of the TM/T, and Dmid/DE,mid ratios. All &,J&Il, the ratio’s have been calculated in percentage. N is the number of treatment set-ups measured, the mean value (X) and one relative standard deviation (SD) have been calculated. I. Entrance dose measurements The results of the DM,en/DE,en ratio’s of the 230 treatment set-ups measured, show a narrow Gaussian distribution with a mean value of 100.3% and one relative standard deviation of 2.5% (Fig. 4). A large error has been defined as a discrepancy between measured and expected dose larger than 5 % (2 SD). Such errors have been detected in 3 % (7/230) of the measured treatment set-ups. The underdosages of 7 and 6% have been due to inaccuracies in the planning contour: the diameter used for calculation being an overestimation of the real patient diameter. One underdosage of 6% has been due to unsteady dose rate of the linear accelerator. The 6 and 7% overdosages have been measured for lateral fields with treatment couch rotation (see part 1). II. Transmission measurements The results of the ratio TM/T, as percentage are plotted in Fig. 5. A broad distribution is found with a mean value of 97.4% and one relative standard deviation of 6.5% (N = 230). The discrepancy between the measured

and the expected

81 25

r

LO-

2 t

1

% 35z k 305

mn

!

25TM/T,

20-

Fig. 5. Results of transmission measurements. The histogram shows the broad frequency distribution of TM/T, (N = 230) with a relative mean value (X) of 97.4% and one standard deviation of 6.5%. The mean values of the three subgroups show the influence of contour errors on the TM/T, ratios. When the contour is correct, the mean value of TM/TE is 98.5% and the standard deviation is 3.3% (white group). On the other hand, when the patient diameter is underestimated (black group) or overestimated (grey group), the mean values are 91.2 and 107.7%, respectively.

9'0

i5

IOil

105

liO%

'M.dDE.en

Fig. 4. Results of entrance dose measurements. The histogram shows the frequency distribution of DM,en/DE,en (N = 230) with a mean value (X) of 100.3% and one relative standard deviation of 2.5%. The three subgroups (black, white and grey) correspond to different inaccuracies in contours: black values are used for patient diameter underestimations (2 1 cm): N = 75; white values for correct contours: N = 132; grey values for patient diameter overestimations (2 1 cm): N = 33.

mean value is about 2.5 % . Very large deviations of the TM from the TE have been detected going from

-22% to + 16%. Large deviations, defined as a deviation larger than 5% from the relative mean value, occurred in 35.7% (82/230) of the treatment set-ups measured. In 17 of the measured treatment setups, the beam axis passed through important bone structures such as the vertebral body or both the mandibular bones.

III. Estimated midline dose In Fig. 6, the frequency distribution of the percentage ratios of the absorbed midline dose (Amid) estimated from in vivo measurements and the expected midline dose (DE,mid) calculated manually from monitor units are shown. The mean value is 99.1 y0 and one standard deviation is 3%. Discrepancies larger than - 5% from the expected midline dose (100 %) have been detected in 7 y0 (16/230) of the treatment set-ups measured. The ratio’s of the absorbed midline dose (Amid) and the prescribed midline dose (DP,mid) in percentage are plotted in Fig. 7. A Gaussian distribution is found with a relative mean value of 97.2% (versus 99.1 y0 for ratio’s) and one relative standard Dmid lDE,rnid deviation of 3% (N = 230). The systematic error on the mean value is 2.8 %. Deviations larger than 5% from the relative mean value (X = 97.2%) have been detected in only 5.7% (13/230) of the treatment set-ups measured.

82

23s FJ z ," I-L3c

25

x : z

35-

:

30-

N = 230 x = 91.2% so = 3%

L-l

25-

2c

15

10

5

90

95

1100

105

110%

Dmid’DE,mid

Fig. 6. Absorbed midline dose (Dmid) as percentage value of the expected midline dose (DE,,&. The expected midline dose has been calculated manually with algorithms taking into account scatter defects due to the use of shielding blocks or missing tissues when part of the field is in air. The histogram shows the frequency distribution of Dmid/DE,mid: the mean value (1) is 99.1% and the standard deviation is 2.9% (N = 230). The mean values (X) of the black, white and grey subgroups are 97.1, 99.6 and 102%, respectively.

But underdosages at midline of 5% or more from the prescribed dose (equal to 100%) have been detected in as many as 20.4% (47/230) of the measurements.

9’5 A

Id0

Id5

11'0%

‘tnid”P,mid

Fig. 7. Absorbed midline dose (Dmid) as percentage value of the prescribed midline dose (DP,,&. The histogram shows the frequency distribution ofD,iJDp,mid: the mean value (X) is 97.2% and the standard deviation is 3% (IV = 230). The mean values (X) of the black, white and grey subgroups are 95.4, 97.8 and 99x, respectively.

tion than the DMM,JDE,_ percentage values (Fig. 4) (SD = 6.5% versus SD = 2.48%). This was expected because the transmission measurement (TM) depends strongly on patientrelated uncertainties such as contour errors (error on thickness) and tissue inhomogeneities (type of tissues) which have no or only a small influence on the delivered entrance dose (&&.

I. Transmission measurements

1.a. The reproducibility of transmission measurements The very large spread of the TM/TE values raises

The histogram of the TM/TE percentage values (Fig. 5) shows a considerably broader distribu-

the question of the reproducibility of the measurements. As a check, the transmission measurements performed 6 times or more on the same

Discussion

83 patient have been investigated: 71 transmission measurements performed on 10 patients. For each of the 10 patients, the mean value of the TM/T, values has been calculated and the TM/TE values have then been normalized. The normalized results of the 71 TM/TE values show a narrow Gaussian distribution with one relative standard deviation of 2.4%. It was concluded that the reproducibility of the measurements, performed on the same patient was good (SD = 2.4%) and therefore could not explain the broad distribution of the overall TM/T, results (SD = 6.5%). 1.b. Errors on contours The external contour of the patient is determined during the treatment simulation. In routine, one contour is taken through the centers of the opposed fields with a simple mechanical contour device. The simulation information and the external contour provide the data used for the dose distribution calculation with the treatment planning system. At the time of irradiation and in vivo dosimetry, the patient diameter has been checked along the beam axis with a caliper after completion of the treatment set-up. The measured patient diameter was then compared to the diameter reported on the planning contour. Only deviations of 1 cm or more, between the diameter on the planning contour and the measured patient diameter were considered as errors. 1.b. 1. Results ofcontour checks. A total number of 53 different planning contours have been checked. The patient diameter has been measured at least two times on two different treatment weeks. Discrepancies of 1 cm or more between the planning contour diameter and the real patient diameter (both measured along the beam axis) have been detected in 39.6% (21/53) of the cases. An underestimation (2 1 cm) of the real patient diameter has been detected in 26.4% (14/53) of the planning contours. The investigation of the methodology used for obtaining the external contour of the patient at the time of treatment simulation, has shown that there was a tendency

to a systematic underestimationof the real patient diameter due to the poor transfer on paper of the external contour, determined with the mechanical device. The few overestimations of the real patient diameter were all due to the plastic mask which did not fit to the patient skin at the level of the neck (larynx carcinoma). The detected contour errors have not been corrected on this patient group in order to be able to investigate firstly, the frequency and repetitive character of contour errors, secondly, the causes of the contour errors and finally their influence on the transmission and the tumor absorbed dose. The detected contour errors have simply been noted to be corrected subsequently. I.b.2. Influence ofcontour errors. The results of the TM/T, percentages (Fig. 5) were separated in groups according to the detected contour errors at the time of the transmission measurement. The first group (white distribution) are the TM/T, values obtained on treatment set-ups with contour errors less than 1 cm. The total number of treatment set-ups measured is 122, the mean value is 98.5% and the standard deviation is 3.3%. The TM/TE values obtained on patients with diameter underestimations are plotted on histogram 5 in black, the number of treatment set-ups measured being 75, the mean value being 91.2%. The grey frequency distribution on Fig. 5 shows the TM/T, values obtained on patients with diameter overestimations (N = 33, X = 107.7%). Contour errors thus introduced large discrepancies between the measured and the expected transmission. A bias in the transmission measurement and patient diameter check has been avoided by the fact that the expected exit dose (DE,_) was only calculated after the accomplishment of the whole radiation treatment making it impossible to correlate TM/TE values with patient diameter at the time of the measurement. The expected transmission (TE) is the ratio of the expected entrance dose (&J and the expected exit dose @&J. An error of a few centimeters on

84 the patient thickness leads only to a small error on entrance dose due to the error on SSD (isocentric technique). However, for the exit dose the error on X-ray transmission is large since it varies exponentially with the patient thickness. The importance of a correct planning contour is demonstrated by the comparison of the white distribution of TM/TE values (correct contour) with the overall distribution of TM/T, values (Fig. 5). The first distribution is considerably narrower (S.D. = 3.25% versus SD. = 6.5%) with a measured mean value showing a discrepancy of only 1.5% from the expected mean value. The mean value of the overall distribution is 1% lower (97.43% versus 98.48%) due to the large number of patient diameter underestimations, twice more than overestimations (75 versus 33). It is obvious that the determination of the external contour of the patient at the time of the treatment simulation is one of the weakest points in the treatment chain of our department. The inaccuracy in the determination of the external contour is related to the mechanical device in use which should be replaced in the near future. I.c. Tissue inhomogeneities When calculating the dose distributions, tissue inhomogeneities are not taken into account for patients treated for neck and oral cavity malignancies in our department: patient tissues are assumed to be water equivalent. Most of the tissues in the neck region and the oral cavity have a relative electron density higher than 1, e.g. muscle, skin, connective tissue, cartilage, bone (ICRU report 44). Only fat and the air-filled cavities have a low relative electron density. In the neck region, the beam attenuation characteristics of the different tissues compensate each other as demonstrated by the very small discrepancy (1.5 %) between the measured and the expected mean value of the TM/TE distribution for correct contours (white, Fig. 5). In 17 of the measured treatment set-ups, the beam axis passed through important bone structures such as the vertebral body or both the mandibular bones. In all these

treatment set-ups, the percentage ratio of the measured and the expected transmission was lower than 90 % . It was concluded that in addition to the systematic tendency to underestimate the patient diameter (except when the plastic mask did not fit the patient’s skin in the neck region which led to a patient diameter overestimation), important bone structures were the cause of the relatively low values at the left side of the overall distribution of TM/TE values (Fig. 5). II. Estimated midline dose On

Fig. 7,

the frequency distribution of DrnidlDP,mid values are shown. A remarkably narrow Gaussian distribution is found (S.D. only 3%), as compared to the distribution of the TM/TE values (S.D. 6.5%) with a mean value of 97.2%. The discrepancy between the absorbed midline dose and the prescribed midline dose is due to contour errors, to tissue inhomogeneities and to the systematic inaccuracy in the calculation algorithms of the treatment planning system

[91. I1.a. Influence of the inaccuracy in the algorithms of the treatment planning system The number of monitor units to deliver the prescribed dose is calculated with the software of the treatment planning system. A systematic error has been detected, due to the scatter defects not being taken into account. This means that the target dose is systematically overestimated when shielding blocks are used or when part of the field is in air (missing tissue). The comparison of the two histograms, 7 (ratio of the absorbed midline dose to the prescribed dose), and 6 (ratio of the absorbed midline dose to the expected dose) shows that the two standard deviations are similar (S.D. 3% and S.D. 2.9x), but the mean values are different (X = 97.2% and X = 99.1%). This demonstrates the systematic underdosage of 2% at midline due to the systematic error in the algorithms of the treatment planning system. This error explains why an

8.5 underdosage of more than 5% at midline depth occurred in 1 I.5 of the measured treatment set-ups. 1I.b. Influence of contour errors and tissue inhomogeneities

The results of the Dmid/DE,mid percentage values are plotted in Fig. 6 and show a narrow Gaussian distribution with a mean value of 99.1% and a standard deviation of 3%. The transmission as expected from

the patient

contour (TE) is overestimated

due to the underestimation of the patient diameter. The entrance dose (DE,_,) is underestimated due to the overestimation of the SSD because the contour error is in general symmetrical with respect to the midline point. Therefore, the error on the transmission is partially compensated by the error on the entrance dose. This is demonstrated by the inverse position of the black and grey distribution with respect to the mean value of the overall DM,en/DE,en and TM/TE distributions, respectively (Figs. 4 and 5). The three values of 108% were due to a systematic positioning error resulting in an entrance dose and transmission error in the same direction. In 7 % (16/230) of the measured treatment set-ups a discrepancy larger than minus 5% is detected between the absorbed midline dose and the expected midline dose. Of these large errors, 75 y0 (12/16) were due to important bone structures along the beam axis.

Conclusions

In vivo measurements of the entrance dose [9] have been shown to be very useful firstly as a check of the quality of the whole treatment chain, secondly to evaluate the quality of several treatment techniques and finally as a check of the dose delivered to a given patient. The sources of the inaccuracies detected by entrance dose measurements are related to errors in the algorithms for dose calculation, errors related to the treatment unit and finally set-up errors or human mistakes.

In this paper, it has been proved that exit dose measurements with semiconductor detectors can be performed and are reliable when due correction factors are determined and applied for exit dose measurement conditions, e.g. correction for lack of backscatter. The exit dose is not only dependent on patient set-up and accelerator parameters (cfr entrance dose) but is also dependent of the patient data such as patient thickness and tissue inhomogeneities. The combination of both entrance and exit dose measured at the same time and on the same patient are very useful to separate uncertainties related to set-up and accelerator parameters on one hand and uncertainties related to patient data on the other hand. The importance of the second kind of uncertainties is estimated from the measured transmission, that is to say, the ratio of the measured exit dose and the measured entrance dose, and when the sources of the error have been detected, corrective actions can be undertaken. However, the radiation therapist is more interested in the target absorbed dose or in the dose delivered to critical organs than in the entrance or exit dose. The target absorbed dose can be estimated from the measured entrance dose and the measured transmission. An important restriction of this method is that inhomogeneities should be symmetrical and equally distributed for reliable absorbed dose determination. In the present study performed on patients treated for head and oral cavity malignancies, three important causes leading to erroneous dose delivery have been detected: (1) Contour errors which occurred frequently and which led to an underdosage because a systematic tendency for patient contour underestimation has been detected. Therefore, a new contour device is being set up. (2) Tissue inhomogeneities such as the mandibular bones or the vertebral body are not taken into account when calculating the monitor units with the treatment planning system.

86 (3) An error in the algorithms of the treatment planning system due to the fact that scatter defects are not taken into account. A more sophisticated treatment planning system will be used in the near future. Each of these three inaccuracies led to an underdosage at the target volume. As a result, underdosages of 5% and more at midline have been detected in l/5 of the measured treatment set-ups. These errors could only be detected by the combi-

nation of both entrance and exit dose measurements. Moreover, it has been shown that on this patient group and on this treatment unit, uncertainties related to patient data have been found to be both more frequent and more important than uncertainties on set-up and accelerator parameters. In addition, the measurements of the transmission for a given patient are reproducible during the total treatment time. This offers possibilities for individual corrections of the target dose. As a conclusion it can be stated that patient dose control by in vivo measurements of the entrance dose is not sufhcient to estimate the overall uncertainty in dose delivery. The combination of entrance and exit dose measurements on the same patient and at the same time are therefore required.

Acknowledgements We wish to thank J. Verstraete for his enthusiastic cooperation in this study. The authors greatly appreciate the secretarial work by Mrs. L. Minnen and the assistance in computer programming by Mr. J. Van Kelecom. We acknowledge Hans Svensson for his very useful indications and criticism. This work is supported by grants of the “Belgisch Werk tegen Kanker”.

References 1 Brahme, A. Dosimetric precision requirements in radiation therapy. Acta Radiol. Oncol. 23: 379-391, 1984.

2 Bridier, A. and Dutreix, A. Method of dose calculation to any point in water for high energy photon beams based on the new dosimetry protocols. Proceeding of XXVIIIiBme Congres SFPH, Lyon, 1989. 3 Chavaudra, J., Marinello, G., BrfilC, A. M. and Nguyen, J. Utilisation pratique du borate de lithium en dosimetrie par thermoluminescence. J. Radiol. Electrol. 57: 435-445, 1976. 4 Dutreix, A. When and how can we improve precision in radiotherapy? Radiother. Oncol. 2: 275-292, 1984. 5 Goitein, M. Calculation of the uncertainty in the dose delivered during radiation therapy. Med. Phys. 12: 608-612, 1985. 6 Herring, D. F. and Compton, D. M. J. The degree of precision required in the radiation dose delivered in cancer therapy. Proceedings of the Third International Conference on Computers in Radiotherapy, Special Report, number 5, 1971. 7 Johansson, K. A. Studies of different methods of absorbed dose determination and a dosimetric intercomparison at the Nordic radiotherapy centres. Thesis, University of GBteborg, Sweden, 1982. 8 Lagrange, J. L., Marcie, S. and Costa, A. Contribution a l’optimisation des mesures in vivo par dttecteurs semiconducteurs. Dosimetry in Radiotherapy. IAEA, Vienna, 1988. 9 Leunens, G., Van Dam, J., Dutreix, A. and van der Schueren, E. Quality assurance in radiotherapy by in vivo dosimetry. 1. Entrance dose measurements, a reliable procedure. Radiother. Oncol. 17: 141-151, 1990. 10 Lindskoug, B. Development and use of a radiothermoluminescent dosimetry system. Automatisation of equipment and procedures. Thesis, University of Gbteborg, Sweden, 1974. 11 Mijnheer, B. J., Battermann, J. J. and Wambersie, A. What degree of accuracy is required and can be achieved in photon and neutron therapy? Radiother. Oncol. 8: 237-252, 1987. 12 Nilsson, B., RudCn, B. I. and Sorcini, B. Characteristics of silicon diodes as patient dosemeters in external beam therapy. Radiother. Oncol. 11: 279-288, 1988. 13 Noel, A. and Aletti, P. Les mesures in vivo systematiques. A propos de 800 controles. Proceeding of XXVIieme CongrQ CFPH Vittel: 124-136, 1987. 14 Rikner, G. Silicon diodes as detectors in relative dosimetry ofphoton, electron and proton radiation fields. Thesis, Uppsala University, Sweden, 1983. 15 Rikner, G. and Grusell, E. General specifications for silicon semiconductors for use in radiation dosimetry. Phys. Med. Biol. 32: 1109-l 117, 1987. 16 Rizzotti, A., Compri, C. and Garusi, G. F. Dose evaluation to patients irradiated by 6oCo beams, by means of direct measurement on the incident and on the exit surfaces. Radiother. Oncol. 3: 279-283, 1985. 17 Rudtn, B. I. Evaluation of the clinical use of TLD. Acta Radiol. Ther. Phys. Biol. 15: 447-464, 1976.

87 18 Shukovsky, L. J. Dose, time, volume relationships in squamous cell carcinoma of the supraglottic larynx. Am. J. Roentgenol. 108: 27-29, 1970. 19 Svensson, H. Quality assurance in radiation therapy: physical aspects. Int. J. Radiat. Oncol. Biol. Phys. 10: 59-65, 1984.

20 Task group 21, Radiation Therapy Committee, American Association ofphysicists in Medicine. A protocol for the determination of absorbed dose from high energy photon and electron beams. Med. Phys. 10: 741-771, 1983.

Quality assurance in radiotherapy by in vivo dosimetry. 2. Determination of the target absorbed dose.

Combined entrance and exit dose measurements were performed with semiconductor detectors on patients, treated for neck and oral cavity malignancies. T...
1MB Sizes 0 Downloads 0 Views