| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Full Paper |
1 Università degli Studi di Milano, Scuola di Specializzazione in Fisica Sanitaria, Milano, and 2 Servizio di Fisica Sanitaria, H. San Raffaele, Via Olgettina 60, 20132 Milano, Italy
Correspondence: Sara Broggi, Servizio di Fisica Sanitaria, H.S. Raffaele, Via Olgettina 60, 20132 Milano, Italy
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and methodsResultsDiscussion and conclusionsReferences |
|---|
1.7%) for both Batho and ETAR algorithms; for bone structures, mean deviations were approximately -0.6% (1 SD2.7%) for both algorithms. The worst agreement was found in the anterior neck where the mean deviation between measured and calculated midline dose was +3.1% (1 SD=1.4%) and +3.4% (1 SD= 2%) using Batho and ETAR, respectively. Sufficiently accurate 2D midplane dose maps may be simply obtained in vivo in the irradiation of head/neck cancer by using a portal detector in combination with a pair of diodes, in order to verify the dose actually delivered during treatment. |
Introduction |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussion and conclusionsReferences |
|---|
Use of a portal imaging device is very suitable in many situations because it permits checking of the actual dose delivered in the whole irradiation field. The opportunity to determine dose in an entire plane may be very useful in complex situations, such as intensity modulated techniques, where large dose gradients are present or when using individual compensators both to verify effective dose homogeneity throughout the irradiated volume and to derive the entrance beam intensity map to achieve expected compensation [28].
In addition, portal dosimetry can be used to investigate variations in dose distribution linked to variations in attenuation by different anatomical structures, and so act as a quality control tool for treatment planning systems (TPSs) [22, 23].
Some studies investigated direct use of transmission dose data to predict exit dose distributions [11, 12, 14, 19, 20, 21]. It has been demonstrated that transmission dose images obtained with portal films correspond very well with relative exit dose maps when small air gaps between phantom/patient and detector are used.
The possibility of using a portal imaging device in conjunction with absolute dosemeters (TLDs or semiconductors) is very attractive, permitting the measurement of midline dose distribution, which often corresponds to the dose actually delivered to the target volume.
Recently the validity of a simple and robust method for estimating midline dose maps, without applying any correction factors, in several phantoms simulating relevant clinical situations, including head and neck treatment, was tested [13]. Results demonstrated the possibility of estimating in vivo midline distributions without any additional patient information concerning external three-dimensional (3D) body contour, generally within 3% in many clinical situations.
In this study we investigated the practical application of our proposed method in assessing the midline dose distribution in patients irradiated for head/neck tumours. Midline dose values were estimated in different anatomical positions and compared with midplane doses calculated with our 3D TPS.
|
Materials and methods |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussion and conclusionsReferences |
|---|
With this approximation, the patient is considered to be equivalent to a parallelepiped phantom with a thickness, z, equal to the real patient's thickness on the field axis, and with a variable electronic density, 
(y), for off-axis points, depending on the water equivalent thickness crossed by the beam.
Based on this assumption, the exit dose map can be determined from relative dose maps measured at the portal film detector level by geometrical back projection to the corresponding exit points followed by normalization with the on-axis absolute exit dose value measured by the diode. In the same way, the relative entrance dose map may be derived by correcting the absolute dose value measured with the diode at the entrance surface by only the off-axis ratios (OARs).
Based on these considerations, for each position we may assess the absolute entrance dose, the corresponding exit dose and estimate the corresponding midline dose by applying Rizzotti-Leunens or Huyskens algorithms, as reported in the literature [7, 18, 31].
The proposed method, using portal films, could be effective even if a generic portal detector as an EPID were used, provided that the gap between detector and exit surface is kept as small as possible.
Patient population and treatment technique
10 patients treated with curative intent for head/neck cancer were considered for this study. Patients were treated in the supine position with a fixed individual thermoplastic mask for immobilization. Markers for patient set-up, field centre and fields edges were indicated on this mask.
Lateral opposed 6 MV X-ray beams were used (Clinac 600; Varian, Zug, Switzerland). The first phase of treatment was given with a large field, encompassing tumour and adjacent lymph node regions. A photon boost was then delivered to the anterior neck by blocking the posterior region, and an electron boost was delivered to the posterior neck nodes. Most of the treatment was delivered without a wedge, however, a wedge was sometimes added in the boost fields to compensate for curvature of the neck.
1–3 factions per patient were monitored; only unwedged "large" fields were considered in this investigation.
Patients characteristics are shown in Table 1
.
|
In the treatment plans studied, the isocentric prescribed daily dose (International Commission for Radiation Units and Measurements dose) ranged from 1.8–1.9 Gy.
Set-up of measurements
After patient set-up, a pair of diodes was placed on the beam axis at the entrance and exit surface. In order to avoid diodes shielding each other, the exit diode was exactly placed on the beam axis, and the entrance diode slightly moved. The correct position of the diodes with respect to the central beam axis and patient anatomy was checked using the portal image. We used p-type silicon diodes (EDP 10; Scanditronix, Uppsala, Sweden) connected to a multichannel electrometer (DPD 510; Scanditronix, Uppsala, Sweden), calibrated in the different irradiation geometries to convert the diode signal to entrance and exit absorbed doses [6, 7, 10]. The use of correction and calibration factors defined for each detector permitted the measurement of absolute entrance and exit doses with an accuracy of 1–1.5%.
After the set up of the diodes, a portal film was positioned at the exit surface, keeping the air gap between patient and portal detector as small as possible. Generally the air gap was approximately 5–10 cm. Previous papers demonstrated that "small" air gaps may be satisfactory with respect to the agreement between relative exit doses and relative portal dose [14, 16, 21].
We used X-Omat films (Kodak) in localization cassettes (Kodak) positioned in a movable holder, which enabled us to check that they were orthogonal to the beam axis.
Portal films were only left in the beam for a limited number of monitor units prescribed for the treatment, keeping the maximum film optical density within 1–1.2 in order to stay in the quasi linear region of the sensitometric curve.
The film was set only for a single lateral beam. Previous papers [13, 24] demonstrated that, when taking the sum of two opposing fields, one film was sufficient to derive transmission portal information. The diodes were left in the beam during the whole treatment in order to estimate the dose actually delivered to the midplane during a complete fraction.
Estimation of the midplane dose
Irradiated films were developed in an automatic processing unit and then read by the VXR 12 film digitizer (Vidar System Corporation, VA). Optical densities were automatically converted with Rfa-Plus 5.3 software (Scanditronix, Uppsala, Sweden) by the appropriate calibration curve, previously measured with the same geometrical distribution of radiation and clinical technique. In this way, optical density distributions read at the level of the portal film at the exit surface of a homogeneous phantom can be correlated with relative dose values measured with an ionization chamber set at the maximum depth from the exit surface of the phantom among the same radiation technique. Previous investigations [14, 15, 17, 20, 21] showed a dosimetric reproducibility of approximately 1–1.5% in assessing relative optical density values. There are possible variations in beam homogeneity, film processing and densitometer reproducibility.
The relative exit doses read at the portal film level were then combined with the on-axis absolute entrance dose value measured with the entrance diode to estimate the corresponding midplane dose by applying the Huyskens algorithm [18].
As reported in our previous paper [17], if we define the entrance transmission (Tin), exit transmission (Tout) and midplane transmission (Tmid) in an inhomogeneous phantom with a real thickness d, and the corresponding values (T'in, T'out and T'mid) in a homogeneous phantom with an equivalent water thickness equal to z, the midline dose can be estimated by:
|
|
where SAD is the source–midplane distance, m the depth of maximum dose and d is the real thickness of an inhomogeneous phantom. From this it was concluded that
|
|
In fact we can plot T'mid and Tout, derived from theoretical tissue:phantom ratios (TPRs) taken at the depth of the entrance, exit and midplane of the beam, against the water equivalent thickness z. Moreover, by plotting
|
|
against z, if d is known and Tout is measured, z can be derived. Then, T'mid corresponding to z can be assigned and the midplane dose in the inhomogeneous phantom estimated.
We compared the estimated midplane doses with the calculated doses for some relevant selected points within bony structures (vertebrae), air cavities such as trachea, and soft tissues. Besides the isocentre point, seven off-axis points, where possible, were defined; two points in air cavities (points 1 and 6), three points in bony structures (points 3, 4 and 5) and two points in homogeneous tissue (points 2 and 7), of which one (point 7) was in the anterior neck region (Figure 1).
|
The exact position of these measurement points on films and TPS was found by the distances from well defined and visible bony landmarks. In order to compare midplane dose values estimated by transmission data with midplane dose distributions calculated by treatment planning, the possible differences in set up of the patients during CT simulation (treatment planning slices) and therapy (portal film) had to be taken into account. For this purpose, the projection at the isocentre plane of the distance between beam axis and a number of easily visible bony landmarks on the portal film and CT slices was measured. In this way the possible change between CT set-up and that of the portal film could be assessed and the optical density profiles for the value at the point corresponding to the beam axis in the CT plan normalized. Inaccuracy in evaluation of relative off-axis doses owing to a wrong assessment of optical density at the normalization point is limited since the optical density gradient in the region around the beam axis, the normalization point, for this radiation treatment is not very high.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussion and conclusionsReferences |
|---|
shows the percentage deviations between the measured midplane dose and the dose calculated with the Batho algorithm for all patients and all anatomical positions considered. It was found that for 89 of 116 measurement points, estimated midplane dose derived from transmission dose data agreed to within 3.5%, and only in 4 of 116 points was agreement outside±5%. Total mean percentage deviation was estimated to be +0.6% (1 standard deviation (SD)=2.8%). By subdividing on-axis (isocentre point) and off-axis (points 1–7) measurements, mean percentage deviations of +0.3% (1 SD=2.2%) and +1% (1 SD
2.6%), respectively, both with the Batho and the ETAR algorithm were found. Detailed results are shown in Table 2.
|
|
In inhomogeneous off-axis points, different results were found according to anatomical location.
In air cavities points 1 and 6, 19 of 22 deviations were positive, with a mean percentage deviation of +2.2% (1 SD=1.7%) for both dose calculation algorithms. In contrast, in the bony structures, points 3, 4 and 5, the estimated deviations tended to be negative, with a mean percentage deviation of approximately –0.6% (1 SD=2.5%).
The least agreement was found at the anterior neck region measurement point, where mean percentage deviations were +3.1% (1 SD=1.4%) and +3.4% (1 SD=2%), respectively, for Batho and ETAR algorithms.
|
Discussion and conclusions |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussion and conclusionsReferences |
|---|
The main advantage of this method is its ability to derive midplane dose maps without any additional patient information and without applying correction factors that take the different portal film–patient distances and source-to-skin distances at the entrance point into account.
This method does not depend on CT data used in the treatment planning process, therefore the accuracy of the measured midplane dose maps is not affected by anatomical changes, but estimates the dose actually delivered to the target volume.
In other to take into account the patient's geometry during treatment, two diodes were set on the beam axis.
In general, midplane doses determined from portal dose measurements agreed to within ±3.5%, compared with values obtained with the TPS. Estimated total mean percentage deviation was +0.6%, in agreement with deviations found using absolute point dosemeters [1–10]. Different allowances have to be made according to the anatomical measurement points.
At the isocentre point, results are in agreement with deviations found in routine entrance in vivo dosimetry. Large deviation, >2%, were found mostly when the beam axis passed through air cavities or when the on-axis point was near bony structures. In these situations, to avoid large deviations, relative dose values were normalized to the mean optical density value around the diode, allowing also for the shadow of the exit diode. This is important if the exit diode is near an interface between two structures with different densities, e.g. air, soft tissues [23].
Inhomogeneous structures caused two contrasting trends. Estimated midplane dose from transmission dose data seemed to underestimate the calculated dose for bony structures but overestimate it for air cavities. These results were congruent with the fact that when the air gap between patient and portal detector increases, the relative importance of primary radiation is greater, because scattered radiation is less. This fact explains the increase in portal off-axis values for the low density inhomogeneity structures and the decrease for high density inhomogeneity tissues. As well as the limit of the applied method that cannot allow for the loss of scattered radiation, the deviations found in these inhomogeneous structures could also be linked to the validity of the algorithms used for dose distribution calculation.
The same variations were also found in studies reported by the Leuven group [12, 22], who compared the exit dose distributions obtained from transmission data with the exit dose profile measured with an ionization chamber in homogeneous and inhomogeneous phantoms, simulating clinical conditions [12], and with exit dose values calculated with the TPS in patients irradiated for head and neck tumours [22]. They found an underdosage behind aluminium inhomogeneities and high density structures such as vertebrae, and an overdosage behind air cavities such as the trachea. The authors proposed a simple method, applicable also in the case of an oblique exit surface, to derive exit dose distributions from transit dose profiles. The relative doses on the films were found to be proportional to the exit doses, without the application of inverse square law correction if the film was placed at a large distance from the patient, because at large focus-to-film distances, (e.g. 130–140 cm) the deviation caused by the decrease of scatter photons seemed to counterbalance the corresponding inverse square correction. We think that this proposed rule may not be effective in all situations owing to the variety of patient shapes, field sizes and thicknesses that can be encountered in clinical practice. In particular, the method proposed by the authors could be inaccurate for inhomogeneous structures where the proportion of the loss of scattered radiation is larger than in soft tissues.
As shown in Table 2
, in accordance with the results found on phantoms [13], the poorest agreement was found in the anterior neck region measurement point, where the midplane dose estimated from portal dosimetry was greater than the calculated dose. In this anatomical region, where the patient's separation narrows, the contribution of scattered photons at the exit surface should be much lower than the corresponding contribution in the beam axis. However, at the level of the portal detector the relative contribution of the scattered component could be more pronounced than at the exit level owing to the increase in the scattered component at this off-axis position.
In addition to the studies reported by Weltens et al [22, 23], a limited number of groups investigated the possibility of using a portal detector in clinical practice for dosimetry. The Amsterdam group [34] proposed the use of an EPID by comparing transmission dose values with calculated values in patients irradiated for lung cancer, in order to determine the accuracy of dose calculation algorithms based on CT densities. In particular, the exit off-axis ratio measured with a diode was compared with the transmission OAR estimated with the EPID. Like the method applied in this paper, no correction was applied for the loss of scattered photons, but transmission data obtained with the EPID were corrected for patient contour, from the CT image. To minimize the inaccuracy in exit dose measurement, the EPID was positioned as close as possible to the exit side of the patient. The agreement between transmission and exit dose values was found, on average, to be 1.1%, comparable with the deviations estimated in this study.
In other papers [24, 25], the same group developed a theoretical model to calculate exit and midplane doses from transit dose measurements. From the primary transit dose component, measured directly by an EPID positioned at a large distance (40 cm or more) from the patient, the primary and scatter exit dose components can be derived by backprojection and convolution algorithms, respectively. The difference in contribution of primary and scatter components between exit and midline positions are then corrected for separately, yielding the midplane dose. It is emphasised that, when the air gap between exit patient's surface and portal detector is not kept small, sophisticated algorithms should be applied to take into account the loss of scattered radiation. As in our technique, the main advantage of the method tested by Boellaard et al [24] is that it did not require CT data and operates independently from the TPS, as the radiological thickness of the patient is determined from the ratio of a portal dose acquired with and without the patient in the beam.
However, detailed anatomical information is not included if, the applicability of the midplane dose measurements is restricted to symmetrical situations, when inhomogeneities are symmetrically distributed around the midplane or when the patient is irradiated with two opposing fields, as for the method proposed [13] and applied in this investigation.
Although patient CT data are not used and dose measurements not affected by anatomical changes, midplane doses may be estimated using only an EPID. In contrast, the method investigated in this paper used a portal detector in conjunction with absolute dosemeters in order to take into account the actual patient position during treatment.
The method, tested first in homogeneous and inhomogeneous phantoms, was applied in a number of relevant clinical cases, such as larynx, breast, lung and prostate cancers, to estimate midplane dose distributions. The estimated deviations between measured and calculated midline distributions were similar to these reported in this paper.
A different approach was proposed by Pasma [30] to compare measured portal dose images with corresponding portal values predicted from the planning CT scan of the patient to reveal possible internal organ motion, frequently caused by gas pockets inside the rectum, during treatment of prostate cancer or during acquisition of the planning CT scan. Transmission dose measurements obtained with an EPID were compared with a predicted transmission dose distribution, based on patient's CT data and treatment parameters. This method is not independent of the CT data used in the treatment planning process. Furthermore, only when CT data represent actual patient geometry during treatment can a difference in transmission be used to evaluate the difference in dose delivery in the target volume; otherwise, dose reconstruction based on measured transmission dose could be inaccurate.
Besides the above considerations, use of a portal film is limited compared with an EPID as proposed by Boellaard et al [25, 26] and Pasma [30], which provides off-line transmitted information. Conversely, an EPID does not always allow the air gap between detector and exit surface to be kept as small as possible. In this case, transmission distribution can be equal to that of the exit only if scatter correction factors are applied.
The use of a portal film, however, demonstrates that portal dosimetry may also be used in radiotherapy departments in which EPIDs are not used, or in irradiation techniques where use of an EPID is not possible, (e.g. total body irradiation).
As reported in Table 2, midplane dose distributions were calculated both with Batho and ETAR inhomogeneity correction algorithms. Very similar mean deviations were found, which suggested that the use of 3D inhomogeneity correction algorithms has not had a significant impact on the accuracy of dose calculations in this situation.
These first clinical tests demonstrate that transit dosimetry can be implemented in practice without an excessive increase in workload. Accurate positioning of two diodes and portal film during treatment takes approximately 2–3 min, and film processing, reading and estimation of the agreement between estimated and calculated midplane doses requires less than 1 h. These times could be significantly decreased by using software tools for automatic matching of calculated midplane dose maps and in vivo transmission data, opportunely backprojected to the midplane.
The results and considerations reported here confirmed that sufficiently accurate in vivo 2D midplane dose maps can be obtained easily by using transmission dosimetry data.
|
Footnotes |
|---|
Received for publication December 11, 2001. Revision received May 2, 2002. Accepted for publication May 28, 2002.
|
References |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussion and conclusionsReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  P C Williams IMRT: delivery techniques and quality assurance Br. J. Radiol., November 1, 2003; 76(911): 766 - 776. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| BJR | DMFR | IMAGING | ALL BIR JOURNALS |
