- Email: [email protected]
Accuracy of a Relocatable Stereotactic Radiotherapy Head Frame Evaluated by use of a Depth Helmet
Clinical Oncology (2002) 14: 31–39 doi:10.1053/clon.2001.0001, available online at http://www.idealibrary.com on
Brain Tumours Accuracy of a Relocatable Stereotactic Radiotherapy Head Frame Evaluated by use of a Depth Helmet K. E. BURTON*†, S. J. THOMAS‡, D. WHITNEY‡, D. S. ROUTSIS†, R. J. BENSON†, N. G. BURNET† Departments of †Oncology and ‡Medical Physics, Addenbrooke’s Hospital, Cambridge CB2 2QQ, U.K. Received: 12 February 2001
Revised: 19 July 2001
Accepted: 23 July 2001
ABSTRACT: In high precision radiotherapy, the more accurately the patient can be relocated, the smaller the clinical to planning target volume margin can be, with reduction in the volume of normal tissue irradiated. The Gill–Thomas–Cosman (GTC) relocatable stereotactic head frame provides immobilization of the patient which is highly reproducible. A depth helmet and measuring probe were used to confirm the accuracy of relocation of 31 patients treated in the GTC frame. The measurements were processed in a spreadsheet developed to calculate the size of the patient’s displacement as a vector. Twenty-seven patients received fractionated stereotactically-guided conformal radiotherapy, and 4 single fraction stereotactic radiosurgery, amounting to 564 measurement episodes. The accuracy was extremely good, and considerably more accurate than standard thermoplastic head shells. Ninety-two percent of the displacement vectors were less than 2 mm, and 97% less than 2.5 mm. Considering each dimension separately, the largest mean displacement was 0.4 mm in the superior–inferior direction. Accuracy was constant through a fractionated course for most patients, but prediction based on measurements from the first few fractions was not reliable. Results were dependent on patient selection, with worse reproducibility in patients with neurological deficits, or difficulty cooperating. The depth helmet measurements detected a loosened mouth bite in one patient and allowed repositioning to be verified without the need for the simulator. Total treatment time, including use of the depth helmet to verify treatment position, is quicker (mean 15.7 min) than using portal films. The depth helmet, used in conjunction with the vector displacement spreadsheet, provides a simple way to define the CTV–PTV margin. For fractionated stereotactic radiotherapy we use a 3 mm CTV–PTV margin. This system could assist technology transfer to centres starting stereotactic radiotherapy using the GTC frame. Burton, K. E. et al. (2002). Clinical Oncology 14, 31–39. 2002 The Royal College of Radiologists Key words: Relocation accuracy, stereotactic radiotherapy head frame
INTRODUCTION
In high precision radiotherapy, accuracy of relocation of the patient is an important element in minimizing the dose to surrounding normal tissue. The more precise the relocation, the smaller the safety margin required and the smaller the volume of normal tissue receiving unnecessary radiation. Reduction of the volume of normal tissue within the planning target volume (PTV) Author for correspondence: K. E. Burton, Department of Oncology, Addenbrooke’s Hospital, Cambridge CB2 2QQ, U.K. Tel: +44 (0)1223 274312; Fax: +44 (0)1223 763120; E-mail: [email protected] 0936–6555/02/010031+09 $35.00/0
should reduce normal tissue complications, or allow dose escalation with potential increase in local control and cure, either strategy improving the therapeutic ratio [1]. For irradiation of small intracranial lesions, the use of a Gill–Thomas–Cosman (GTC) relocatable stereotactic head frame provides immobilization of the patient which is highly reproducible from day to day. This allows the use of a narrow margin between the clinical target volume (CTV) and the PTV, minimizing the volume of normal tissue irradiated around the target. In some situations, it may allow complete avoidance of some critical normal structures. 2002 The Royal College of Radiologists
32 Table 1 – Conditions treated for the 31 patients in the study Diagnosis
Number of patients
Meningioma
9
Chondrosarcoma Chordoma
2 1
Low grade glioma Primary neuroectodermal tumour
2 1
High grade glioma – retreatment High grade glioma – localized boost
2 7
Metastasis
7*
Thirteen patients were treated using multiple arc stereotactic radiotherapy, 15 with fixed beam stereotactically-guided conformal radiotherapy, and three using both techniques. *Of these, four patients had single fraction stereotactic radiosurgery.
To make an informed decision on the size of the margin when expanding the CTV to the PTV, it is necessary to measure the accuracy of relocation. Portal imaging and computer image analysis tools can be used to compare planning or simulator verification films with portal films or electronic images. However, they do not provide information on small movements, can only review two dimensions at a time, and they are labour intensive. Moreover, there are important manpower issues if these techniques are to be used ‘on line’, during a radiotherapy treatment. To define the appropriate CTV to PTV margin we have used a depth helmet and measuring probe to determine the accuracy of relocation of the patient within the GTC frame. These measurements have been processed in a spreadsheet designed to calculate the displacement in each direction, and as a combined displacement vector. This vector gives a total 3 dimensional displacement.
PATIENTS AND METHODS
In the four years up to December 2000, all patients immobilized for radiotherapy using the relocatable GTC stereotactic head frame were entered into the study of set-up accuracy. A total of 31 patients with varying diagnoses were included (Table 1), with a total of 564 measurement episodes. Twenty-seven patients received fractionated stereotactically-guided conformal radiotherapy, and 4 underwent single fraction high dose stereotactic radiosurgery. The basic design for the GTC frame was developed by Gill and colleagues in London (Gill SS 1987, UK Patent Application N 8728150) [2,3]. The depth helmet was an addition, designed and added to the frame in Boston, U.S.A., where it was named the ‘Depth Confirmation Helmet’ [4]. The GTC frame comprises a standard head ring to which an individualized dental impression of
the upper teeth and a moulded occipital support are attached [2,3]. The individualized components are manufactured in the mould room. Once the frame is located on the patient using the mouth bite, it is held firmly in position with a set of straps across the vertex. Once the frame is fitted, the patient lies supine in the treatment position, and the frame is attached to brackets on the treatment couch. The depth helmet attaches securely to the head frame, once the patient is positioned. The helmet is a Perspex hemisphere with a series of 24 evenly spaced measuring portals. Using a metal measuring probe inserted into each portal, a set of measurements is made from the outside edge of the helmet portal to the patient’s scalp. The helmet is removed before the localization box is attached to position the isocentre. All patients had sufficiently good dentition for relocation of the mouth bite. Measurements were taken at each planning, verification and treatment episode. At the first visit, when the mouth bite and occipital plate were made, the frame was fitted once and a set of measurements taken. The frame was removed completely and refitted before another series of measurements were taken. This ensured at the first visit that reproducibility could be achieved. The next visit was for the CT localization scan, and the data set taken at that appointment was used as the baseline for future comparison. At each visit, the frame was refitted until the measurements were in agreement with those taken at the localization scan. Initially, acceptance criteria for relocation were formulated empirically: the frame was considered to have relocated acceptably provided that the differences from the baseline measurement set were no worse than one reading >2 mm, or three readings >1.5 mm. When these tolerances were not met, the frame was refitted and measurements repeated. These acceptance criteria allow rapid assessment of whether the fitting is acceptable. However, they do not describe the true displacement of the patient within the frame since only individual measurements are considered, nor do they account for differences between radiographers. For these reasons, a spreadsheet was designed to calculate the 3-dimensional vector displacement of the patient’s position within the frame compared to the baseline position. This method uses information from every measurement to calculate the patient’s position within the frame, and thus estimates the displacement of the treatment isocentre from the planned isocentre position. For each measurement, the displacement in the x, y and z dimensions, and the length of the displacement vector (given by √[x2 +y2 +z2]) are calculated using simple trigonometry from the position and angle of the measuring portal; see Appendix 1 for details. From a standard template, a separate spreadsheet was established for each patient. This was used to calculate the displacements on a fraction by fraction basis, as well as the mean and standard deviation for the completed course. The individual spreadsheets can be linked to produce
population data. A copy of the spreadsheet is available on request. Measurements were recorded from as many portals in the depth helmet as possible, in order to maximize data used to calculate the displacement. Measurements were not recorded where the pointer failed to reach to the patient’s head or where they fell on the edge of the velcro strap. Measurements onto the occipital plate were recorded, but not used for calculation. Thus, typically four portals were omitted from the calculation. All readings recorded in the baseline set were measured at all subsequent episodes. To minimise inter-observer variation, the radiographer who obtained the baseline readings would remain the measurer for all other episodes when possible, but this was not always practicable for patients having fractionated treatment. A maximum of three radiographers acted as measurer for any one patient in a fractionated course. In order to assess probe measurement uncertainty and inter-observer differences, an experiment was performed using a volunteer immobilized in the frame. Three radiographers each measured 5 complete data sets, making a total of 15 in all, without the frame being repositioned. The anthropomorphic phantom was not considered suitable for this, because its surface does not have the elasticity of the scalp and will not permit multiple measurements. During the early part of the study all measurements were entered into the spreadsheet after completion of the radiotherapy course. Later in the study, for one patient undergoing radiosurgery, the spreadsheet was used to verify the accuracy of relocation after the frame had been fitted in the treatment room, and prior to commencement of the exposure.
RESULTS Individual Patient Results
Individual data for all patients are presented in Table 2, showing the mean displacements in the x, y and z directions, and the vector length displacements. Data for the three directions of movement were analysed separately, and plotted against time. The anterior–posterior (AP) displacements for each treatment for three representative patients are shown in Fig. 1. Displacements in the other two dimensions showed a similar pattern. These graphs were used to evaluate the possibility of increasing or decreasing accuracy of relocation during the course of a fractionated treatment. In all three directions, the accuracy of relocation for an individual patient remained largely constant through the treatment. In a few patients, the accuracy of relocation improved during the course, but in two patients it worsened, particularly in the superior–inferior (SI) direction, indicating that early readings did not predict accuracy later in the course.
33
Table 2 – Data for 564 measurement episodes in all 31 patients, showing the individual mean values of displacements in the x, y and z directions, and the vector length displacements. All measurements are in millimetres Mean values (mm) Patient no 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 Mean SD
No of readings
Ant–Post
Left–Right
Sup–Inf
Vector length
3 5 6 5 6 2 12 27 30 11 30 6 2 2 30 30 30 30 5 30 30 30 2 6 5 39 30 30 30 30 30
0.6 0.4 0.4 0.5 0.8 0.8 0.3 0.7 0.3 0.3 1.0 1.3 0.5 0.3 0.6 0.2 0.3 0.3 0.0 0.3 0.3 0.0 0.0 0.0 0.3 0.2 0.5 0.4 0.0 0.5 0.5
2.3 0.7 0.6 0.0 0.8 0.4 1.4 0.0 0.4 1.2 0.8 0.2 0.8 0.7 0.2 0.2 0.3 0.7 0.3 0.9 0.2 1.3 0.2 0.9 1.5 0.3 0.6 0.8 0.5 0.1 0.3
1.3 0.8 0.1 0.9 0.3 0.6 0.2 0.3 0.0 0.2 0.3 0.7 0.6 0.4 0.3 0.9 1.2 0.8 0.1 0.2 0.7 0.6 0.3 0.2 1.6 0.7 0.9 0.7 0.1 0.4 0.2
2.8 1.2 0.9 1.3 1.2 1.3 1.6 1.2 0.7 1.3 1.5 1.8 1.2 1.0 1.0 1.1 1.3 1.4 0.6 1.1 1.0 1.8 1.2 1.0 2.4 1.0 1.3 1.2 1.0 0.9 1.0
0.1 0.6
0.1 0.9
0.4 0.7
1.2 —
Overall Results
To assess the overall accuracy of relocation within the GTC frame, the 564 data sets were pooled (Table 3 and Figs 2 & 3). Figures 2a–2c show the frequency distribution of displacements in each direction, conforming to a normal distribution. The mean displacement away from the correct isocentre position for the entire data set is extremely small, the greatest mean displacement reaching 0.4 mm in the SI direction. In the x, y and z directions, 97% of the treatment episodes had measured displacements of <2 mm. Although the greatest displacement for any patient was 3.5 mm, less than 1% of all treatments had displacements greater than 2.5 mm. The largest displacements in each direction did not occur in the same patients. Figure 3 shows the frequency distribution of the length of the displacement vector. This distribution is skewed rather than normal, with a mean of 1.2 mm. Ninety-two percent of the displacement vectors were less than 2 mm, and only 3% exceeded 2.5 mm (Table 3).
34 2.50 patient 11 patient 20 patient 26
2.00
Displacement (mm)
1.50 1.00 0.50 0.00 –0.50 –1.00 –1.50 –2.00 –2.50
2
4
6
8
10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 Episode No.
Fig. 1 – Plot of displacements against time in the anterior–posterior direction, for 3 representative patients. Displacements in the lateral and superior–inferior directions (not shown) are similar.
Scrutiny of the raw measurements, using the initial acceptance criteria, successfully excluded major deviations. However, the actual frame displacement, which was calculated from the entire measurement set using the spreadsheet, was always smaller than the variation in individual measurements.
Measurement Uncertaintly and Inter-observer Differences
Verifying Frame Integrity and Position
In one patient, the occipital plate accidentally became detached from the frame. Depth helmet measurements onto the occipital plate were available and this allowed the plate to be repositioned correctly. In another patient the mouth bite plate loosened spontaneously. This was not apparent until depth helmet measurements were carried out, which demonstrated that the patient was incorrectly positioned. The loose mouth bite was tightened and the patient subsequently refitted with acceptTable 3 – Pooled data for the mean and range of set up deviations Percentage of readings within
Direction
Mean*
Standard deviation*
Ant–Post Left–Right Sup–Inf
0.1 0.1 0.4
0.6 0.9 0.7
99.1 92.1 94.6
99.6 97.1 97.8
99.8 98.9 99.2
–
75.1
91.6
97.1
1.5 mm
2 mm
able accuracy. In a third patient, the mouth bite broke, and after repair the position was verified using repeat depth helmet measurements. In all these cases, patient positioning was confirmed without the need for re-verification in the simulator.
The data for each measuring portal in the 15 volunteer measurement sets were used to calculate the mean depth for that portal. These figures were used as the baseline measurement; the difference between this and each individual measurement was then calculated. The five data sets for each observer were considered separately. For all three observers variation between the data sets was less than 0.1 mm. The data sets for each observer were pooled, giving 100 measurement differences for each person. The mean difference for each observer was zero (less than 0.01 mm). Analysis of variance showed no difference between the three observers (F=0.049, P=0.95).
2.5 mm
DISCUSSION
Vector length†
1.2
*All measurements in millimetres. These values are also given in Table 2. †Vector length is given by √(x2 +y2 +z2) where x, y and z are the displacements in each direction (see Appendix 1). The standard deviation for the vector length is not shown, since the distribution is skewed rather than normal.
Quantitative Results
Overall, relocation was extremely accurate. The mean deviation of the group as a whole measured around 0.1 mm in AP and left–right (LR) directions, and 0.4 mm in the SI direction. This degree of accuracy of relocation is of the same magnitude as that described by other groups [2,3,5–9]. Assessment of the reproducibility of the original frame, carried out at St Bartholomew’s, without the
0
0 –3.5 –3.3 –3.1 –2.9 –2.7 –2.5 –2.3 –2.1 –1.9 –1.7 –1.5 –1.3 –1.1 –0.9 –0.7 –0.5 –0.3 –0.1 0.1 0.3 0.5 0.7 0.9 1.1 1.3 1.5 1.7 1.9 2.1 2.3 2.5 2.7 2.9 3.1 3.3 3.5
No. of measurement episodes –3.5 –3.3 –3.1 –2.9 –2.7 –2.5 –2.3 –2.1 –1.9 –1.7 –1.5 –1.3 –1.1 –0.9 –0.7 –0.5 –0.3 –0.1 0.1 0.3 0.5 0.7 0.9 1.1 1.3 1.5 1.7 1.9 2.1 2.3 2.5 2.7 2.9 3.1 3.3 3.5
0
–3.5 –3.3 –3.1 –2.9 –2.7 –2.5 –2.3 –2.1 –1.9 –1.7 –1.5 –1.3 –1.1 –0.9 –0.7 –0.5 –0.3 –0.1 0.1 0.3 0.5 0.7 0.9 1.1 1.3 1.5 1.7 1.9 2.1 2.3 2.5 2.7 2.9 3.1 3.3 3.5
No. of measurement episodes
No. of measurement episodes
35
60 (a)
50
40
30
20
10
Displacement (mm)
60 (b)
50
40
30
20
10
Displacement (mm)
60 (c)
50
40
30
20
10
Displacement (mm)
Fig. 2 – Frequency distribution of displacements of all 564 measurements in 31 patients, in (a) anterior–posterior, (b) left–right, and (c) superior–inferior dimensions.
36
50 40 30 20
4
4.2
3.8
3.6
3.4
3
3.2
2.8
2.6
2.4
2.2
2
1.8
1.6
1.4
1.2
1
0.8
0.6
0.4
0
0.2
10
0
No. of measurement episodes
60
Displacement (mm)
Fig. 3 – Frequency distribution of displacement vectors.
benefit of the depth helmet, suggested an accuracy of 0.3 mm with a standard deviation of 0.15 mm, and the overall accuracy of the system was considered to be 2.4 mm [2]. The Royal Marsden group assessed the accuracy of the GTC frame [3,5,6] using simulator films. The mean displacement was 1.0 mm, with a standard deviation of 0.6 mm, and a maximum of 2.3 mm. Including errors associated with the treatment machine, they estimated the maximum inaccuracy at around 3 mm [7]. In Boston, Kooy et al. [4] used the helmet to evaluate reproducibility, but measured only portals on the helmet which were in the true AP, LR and SI directions. From measurements on 20 patients they estimated a standard deviation of patient repositioning of about 0.4 mm. Other studies using different frames have also reported similar accuracy of relocation [8,9]. Use of our vector displacement spreadsheet has helped to define the set-up accuracy. Since the distances are calculated as displacement vectors, it largely removes variation due to differences in the pressure of the measurement, which may occur on different days or with different radiographers making the measurements. Nevertheless, the inter-observer variation is extremely small, as noted above. This shows that observers are interchangeable provided they are appropriately trained. Although it is our standard policy to keep to a minimum the number of different radiographers carrying out the head frame measurements, it is impossible to restrict this to only one individual for fractionated treatment. However, measurements in the SI dimension suffer from the weakness that there are no measuring portals inferior to the stereotactic frame to balance the superior ones. Thus in this dimension differences in the pressure exerted on the measuring probe by different operators are not compensated for. The straps do allow slight movement superior-inferiorly, and upward twisting of the mouth bite is more likely than AP movement due to the design characteristics of the mouth bite attachment. Our
conclusion is that the GTC frame may be slightly less accurate in the SI dimension, which is consistent with Kooy et al. [4], but the mean error here is still only 0.4 mm. A slightly larger displacement in the superior– inferior dimension is also a regular finding with both rigid and thermoplastic head shells [10–12]. These results have also allowed us to evaluate our original crude acceptance criteria: they successfully exclude major deviations, but do not give a true representation of the actual displacement, which is smaller than the unprocessed measurements suggest. Individual Patient Results and Patient Selection
All the patients in this series had good dentition, which was assessed before considering them for treatment in the stereotactic radiotherapy frame. The results show the average accuracy of relocation to be extremely good. In the x, y and z dimensions, under 3% of set-ups had a displacement of greater than 2 mm. The displacement vectors exceeded 2.5 mm in less than 3% of episodes. One patient treated in the frame had some difficulty cooperating due to anxiety and larger displacements than average were noted. In patients with severe anxiety or motor and sensory deficits, depth helmet measurements showed no value in the use of the stereotactic frame compared to a standard clear thermoplastic shell, illustrating the importance of careful patient selection. Changes Over Time Through a Fractionated Treatment
There is no consistent trend for either improving or worsening set-up as the course of treatment progresses (Fig. 1). This is reassuring because it suggests that the dental impression material does not become loose or damaged, and that other factors also remain unchanged.
A separate question is whether the CTV to PTV margin could be tailored to suit the patient, relying on measurements from the first few fractions. In a few cases displacements increased for a few fractions in the later part of the treatment, so that early measurements cannot be used to individualize the CTV to PTV margin.
Relocation of the Mouth Bite
The accuracy of the GTC frame is based on accurate relocation of the mouth bite. There is information specifically regarding the accuracy of mouth bite relocation, which supports use of this technique, from development work for a repositioning system for thermoplastic shells. Bova et al. [13] reported that a plate could be repositioned with a mean error of 0.48 mm, and a standard deviation of 0.39 mm. These measurements are consistent with our data.
37
siderably greater accuracy than standard head shells, particularly those of the thermoplastic type. Additional Advantages of the Depth Helmet
Readings from the depth helmet have allowed us to refit loosened mouth bite and occipital plates, and to repair a broken dental impression. These are hazards when the same frame is used for several patients, which is realistic in busy departments given the cost of commercial frame systems. Use of the depth helmet, with the vector displacement spreadsheet, makes it clear which patients set up accurately and for which patients the technique confers no advantage. Use of the depth helmet with the spreadsheet is a simple way to measure set-up accuracy to define the CTV to PTV margin, which is needed for implementation of stereotactic radiotherapy. It could therefore permit technology transfer of stereotactic radiotherapy into centres starting this techniques.
Comparison with Head Shell Systems
Time Taken for Depth Helmet Measurements
The Heidelberg group have developed a relocatable stereotactic radiotherapy system using a mask substantially more rigid than conventional thermoplastic, together with a stereotactic frame and localizer [10]. Using this immobilization device, with a photogrammetric apparatus to detect displacements, they report standard deviations of reproducibility in all 3 dimensions of AP 0.9 mm, LR 0.6 mm and SI 1.3 mm. Hamilton et al. [11] assessed the reproducibility of a head shell and frame system, akin to the Heidelberg type, using radio-opaque spheres embedded in a custom mouthpiece which was independent of the mask and frame. The reproducibility was analysed using portal films on which images of the spheres could be seen. In 104 set-ups on 12 patients they found a mean deviation in the position of the isocentre of 1.8 mm, with a standard deviation of 1.4 mm. The maximum displacement seen was 6 mm. Thermoplastic shells immobilize much less accurately. In the hands of the Heidelberg group the largest deviations occurred laterally, with a standard deviation of 4.6 mm, and a maximum of 8 mm [10]. This level of accuracy is consistent with other reports of cranial set-up accuracy [11,14–16]. Byhart et al. [14] found errors greater than 5 mm in 7% of treatments. Verhey et al. [15] reported reproducibility with errors of up to roughly 5 mm using a mask system. Hanna et al. [12] recorded set-up inaccuracy of the same order, with deviations of greater than 5 mm occurring in 1% of measurements in the AP direction but 13% in the SI direction. Gildersleve et al. [16] reported errors greater than 5 mm in 8% of set-ups in the AP and 14% in the SI directions. These published results clearly demonstrate that relocatable stereotactic radiotherapy frames, in a number of manifestations, can reposition patients with con-
A major concern with the system is the additional time required for the depth helmet measurements. The mean total treatment time, including depth helmet measurements, for 39 stereotactically-guided fixed field treatment episodes in 3 patients was 15.7 min (median 15.2, standard deviation 3.2, range 11.1–25 min). This is consistent with timings reported in the literature [6,13]. The equivalent standard treatment using a clear thermoplastic shell, that is a three-field plan with blocks on each field, takes an average of 12.5 min in our department [17]. Since use of the depth helmet includes assessment of accuracy, a more direct comparison should include portal films of each field. This increases the average treatment time to 18.8 min [17]. Thus, use of the stereotactic frame with the depth helmet takes on average 3 min longer than using a shell, but is quicker than taking portal films. Since the depth measurements provide daily verification of patient position for high precision treatments, this represents an excellent balance between speed and high precision. Use of Portal Imaging for Measurement of Positioning Accuracy
It is our standard practice to carry out simulator verification of fixed field radiotherapy treatment plans, prior to starting treatment. In this way, most systematic errors can be accounted for. In addition, portal images are routinely taken on each field, at the start of treatment. However, we have found that, with small field treatments, it is often impossible to find sufficient bony anatomy to compare the treatment set and simulator films. Where blocks are being used, double exposure films can be taken, but the field has to be opened very wide, causing unnecessary irradiation to normal tissue.
38
Beam films taken with the same frequency so depth helmet measurements could add sufficient dose to the lens of the eye, for example, to approach the threshold for cataract formation. After the first treatment, we prefer to rely on the depth helmet measurements, rather than irradiate more normal tissue. For arc treatments, the isocentre position is verified with CT. It is impractical with our equipment to take portal films of arc treatments. Whilst low-weighted orthogonal fixed fields could be added to arc treatment plans, and imaged on set, we prefer to use the depth helmet and spreadsheet to verify patient positioning. Defining the CTV to PTV Margins for Treatment
The discrepancies measured by the depth helmet relate to patient position within the frame. Set-up inaccuracies due to the treatment machine also occur, though they are not measured by this method. For fractionated treatment, we estimate the maximum discrepancy due to machine inaccuracies to be of the order of 1 mm. These should be added in quadrature to the patient set up inaccuracies measured from the head frame. For single fraction stereotactic radiosurgery, which we carry out using multiple arcs, we estimate possible machine inaccuracies to be up to 1.6 mm, and this should be added arithmetically to the measured head frame deviations. In both situations, using the mean measured displacement vector length, the total errors are extremely small. For fixed fields, these would amount to 1.6 mm, and for single fraction multiple arcs to 2.8 mm. Use of the mean displacement, however, would allow some fractions in some patients to under treat the PTV, albeit by a small amount. Ninety-seven percent of vector displacements are encompassed by a 2.5 mm margin. We have elected to routinely use a 2.5 mm margin for vector discrepancies. For fractionated treatment the total PTV margin becomes 2.5 mm plus 1 mm for machine inaccuracy; added in quadrature this gives a PTV margin of 2.7 mm. As a result of these calculations, for fixed field stereotactically-guided conformal radiotherapy our routine practice is to use a 3 mm margin. Use of the Relocatable Frame for Single Fraction Stereotactic Radiosurgery
A margin of 4.1 mm (2.5+1.6 mm added arithmetically), needed to guarantee coverage of the PTV for stereotactic radiosurgery, is larger than clinically acceptable. There is delicate balance between adequate coverage of the PTV and risk of normal tissue toxicity. For this reason the Boston group [4] recommend against the use of the relocatable frame for radiosurgery. Therefore, in patients receiving radiosurgery, we now routinely process the depth helmet measurements to verify that positioning is accurate to better than 1.4 mm vector displacement prior to treatment. This acceptance
criterion allows a maximum error of 3 mm, when added arithmetically to 1.6 mm for machine inaccuracy. Ideally, for radical radiosurgery where coverage of the PTV is critical, use of a skull-fixed frame to reduce the CTV–PTV margin to a minimum is preferable. CONCLUSIONS
Use of the depth helmet with the vector displacement spreadsheet has proved extremely effective as an adjunct to setting up patients in the relocatable stereotactic head frame. The technique is relatively quick and produces excellent day-to-day accuracy. It is valuable in selecting patients and developing stereotactic readiotherapy techniques. It may help centres wishing to start stereotactic radiotherapy. It has allowed us to choose a value for the CTV and PTV margin for stereotactic radiotherapy and radiosurgery in our department. For fractionated treatments we suggest a CTV to PTV margin of 3 mm. For single fraction stereotactic radiosurgery we assess the accuracy of relocation before the start of treatment, enabling us to use a minimum CTV–PTV margin. Acknowledgements. We wish to thank Mrs Carol Moxey for her help in preparing the manuscript, Mrs Kath Walker for her support, the radiographers in the department who carried out the depth helmet measurements, and Miss Eleanor Pinto for statistical advice. REFERENCES 1 Burnet NG, Wurm R, Nyman J, Peacock JH. Normal tissue radiosensitivity – how important is it? Clin Oncol 1996;8:25–34. 2 Thomson ES, Gill SS, Doughty D. Stereotactic multiple arc radiotherapy. Br J Radiol 1990;63(754):745–751. 3 Gill SS, Thomas DG, Warrington AP, Brada M. Relocatable frame for stereotactic external beam radiotherapy. Int J Radiat Oncol Biol Phys 1991;20(3):599–603. 4 Kooy HM, Dunbar SF, Tarbell NJ, et al. Adaptation and verification of the relocatable Gill–Thomas–Cosman frame in stereotactic radiotherapy. Int J Radiat Oncol Biol Phys 1994; 30(3):685–691. 5 Graham JD, Warrington AP, Gill SS, Brada M. A non-invasive, relocatable stereotactic frame for fractionated radiotherapy and multiple imaging. Radiother Oncol 1991;21(1):60–62. 6 Laing RW, Thompson V, Warrington AP, Brada M. Feasibility of patient immobilization for conventional cranial irradiation with a relocatable stereotactic frame. Br J Radiol 1993;66(791):1020–1024. 7 Warrington AP, Laing RW, Brada M. Quality assurance in fractionated stereotactic radiotherapy. Radiother Oncol 1994; 30(3):239–246. 8 Scott TW, Beach JL, Mendiondo OA. A precision repeat localization head frame for fractionated stereotactic radiotherapy. Med Dosim 1997;22(1):5–8. 9 Theodorou K, Kappas C, Tsokas C. A new non-invasive and relocatable immobilization frame for fractionated stereotactic radiotherapy. Radiother Oncol 1998 Jun;47(3):313–317. 10 Schlegel W, Pastyr O, Bortfeld T, Gademann G, Menke M, Maier-Borst W. Stereotactically guided fractionated radiotherapy: technical aspects. Radiother Oncol 1993;29(2):197–204. 11 Hamilton RJ, Kuchnir FT, Pelizzari CA, Sweeney PJ, Rubin SJ. Repositioning accuracy of a noninvasive head fixation system for stereotactic radiotherapy. Med Phys 1996;23(11):1909–1917. 12 Hanna CL, Slade S, Mason MD, Burnet NG. Translating radiotherapy Clinical Target Volumes into Planning Target Volumes for bladder and brain tumour patients. Clin Oncol 1999;11:93–98.
13 Bova FJ, Buatti JM, Friedman WA, Mendenhall WM, Yang CC, Liu C. The University of Florida frameless high-precision stereotactic radiotherapy system. Int J Radiat Oncol Biol Phys 1997; 38(4):875–882. 14 Byhart RW, Cox JD, Hornburg AG, et al. Weekly localisation films and detection of field placement deviations. Int J Radiat Oncol Biol Phys 1978;4:881–887. 15 Verhey LJ, Goitein M, McNulty P, Munzenrider JE, Suit HD. Precise positioning of patients for radiation therapy. Int J Radiat Oncol Biol Phys 1982;8(2):289–294. 16 Gildersleve J, Dearneley DP, Evans PM, et al. Reproducibility of patient positioning during routine radiotherapy, as assessed by an integrated megavoltage imaging system. Radiother Oncol 1995; 35:151–160. 17 Burnet NG, Routsis DS, Murrell P, et al. A tool to measure radiotherapy complexity and workload: derivation from the Basic Treatment Equivalent (BTE) concept. Clin Oncol 2001;13:14–23.
APPENDIX 1
Figure 4 shows horizontal and vertical sections through the helmet, to illustrate the method used to combine the depth helmet readings. The angle applies to all vertical planes spaced around the helmet. For each probe position a unit displacement ri in the direction shown can be divided into its x, y and z components: xi =ri cosi cosi yi =ri sini cosi zi =ri sini For each depth probe position, the vectors can be added, and normalized to give the overall displacement. Horizontal
θ
Vertical
φ
Fig. 4 – Horizontal and vertical sections through the helmet, to illustrate the method used to combine the depth helmet readings
39
Thus:
The effect of the algorithm can be demonstrated by considering first the 2D situation of the horizontal positions around the base of the helmet, for a patient displacement of (2.0 mm, 1.0 mm). For a set of depth differences of (1, 0.71, 2, 2.12, 1, 0.71, 2, 2.12), the algorithm would give: x =(0+0.5+2+1.5+0+0.5+2+1.5)/ (0+0.5+1+0.5+0+0.5+1+0.5) =8/4 =2.0 mm y =(1+0.5+01.51+0.5+01.5)/ (1+0.5+0+0.5+1+0.5+0+0.5) = 4/4 = 1.0 mm. If another operator pressed less hard, and all depths were decremented by 0.5 mm, the algorithm still returns the values x=2.0, y= 1.0. In the 3D case, for all 24 points, if a set of depths that gave (2, 1.0) were all decremented by 0.5 mm, the result would be (2.0, 1.0, 0.66). This is because the z dimension (superior–inferior) is measured in one direction only, whereas the x and y directions have opposing points. If all 24 points are present, an error of 0.5 mm in a single point can change the calculated overall displacement by at most 0.06 mm in any direction. In practice there are always a few points missing, typically 4, and care should be taken that the number of points is roughly balanced between anterior and posterior, and between left and right.
Brain Tumours Accuracy of a Relocatable Stereotactic Radiotherapy Head Frame Evaluated by use of a Depth Helmet K. E. BURTON*†, S. J. THOMAS‡, D. WHITNEY‡, D. S. ROUTSIS†, R. J. BENSON†, N. G. BURNET† Departments of †Oncology and ‡Medical Physics, Addenbrooke’s Hospital, Cambridge CB2 2QQ, U.K. Received: 12 February 2001
Revised: 19 July 2001
Accepted: 23 July 2001
ABSTRACT: In high precision radiotherapy, the more accurately the patient can be relocated, the smaller the clinical to planning target volume margin can be, with reduction in the volume of normal tissue irradiated. The Gill–Thomas–Cosman (GTC) relocatable stereotactic head frame provides immobilization of the patient which is highly reproducible. A depth helmet and measuring probe were used to confirm the accuracy of relocation of 31 patients treated in the GTC frame. The measurements were processed in a spreadsheet developed to calculate the size of the patient’s displacement as a vector. Twenty-seven patients received fractionated stereotactically-guided conformal radiotherapy, and 4 single fraction stereotactic radiosurgery, amounting to 564 measurement episodes. The accuracy was extremely good, and considerably more accurate than standard thermoplastic head shells. Ninety-two percent of the displacement vectors were less than 2 mm, and 97% less than 2.5 mm. Considering each dimension separately, the largest mean displacement was 0.4 mm in the superior–inferior direction. Accuracy was constant through a fractionated course for most patients, but prediction based on measurements from the first few fractions was not reliable. Results were dependent on patient selection, with worse reproducibility in patients with neurological deficits, or difficulty cooperating. The depth helmet measurements detected a loosened mouth bite in one patient and allowed repositioning to be verified without the need for the simulator. Total treatment time, including use of the depth helmet to verify treatment position, is quicker (mean 15.7 min) than using portal films. The depth helmet, used in conjunction with the vector displacement spreadsheet, provides a simple way to define the CTV–PTV margin. For fractionated stereotactic radiotherapy we use a 3 mm CTV–PTV margin. This system could assist technology transfer to centres starting stereotactic radiotherapy using the GTC frame. Burton, K. E. et al. (2002). Clinical Oncology 14, 31–39. 2002 The Royal College of Radiologists Key words: Relocation accuracy, stereotactic radiotherapy head frame
INTRODUCTION
In high precision radiotherapy, accuracy of relocation of the patient is an important element in minimizing the dose to surrounding normal tissue. The more precise the relocation, the smaller the safety margin required and the smaller the volume of normal tissue receiving unnecessary radiation. Reduction of the volume of normal tissue within the planning target volume (PTV) Author for correspondence: K. E. Burton, Department of Oncology, Addenbrooke’s Hospital, Cambridge CB2 2QQ, U.K. Tel: +44 (0)1223 274312; Fax: +44 (0)1223 763120; E-mail: [email protected] 0936–6555/02/010031+09 $35.00/0
should reduce normal tissue complications, or allow dose escalation with potential increase in local control and cure, either strategy improving the therapeutic ratio [1]. For irradiation of small intracranial lesions, the use of a Gill–Thomas–Cosman (GTC) relocatable stereotactic head frame provides immobilization of the patient which is highly reproducible from day to day. This allows the use of a narrow margin between the clinical target volume (CTV) and the PTV, minimizing the volume of normal tissue irradiated around the target. In some situations, it may allow complete avoidance of some critical normal structures. 2002 The Royal College of Radiologists
32 Table 1 – Conditions treated for the 31 patients in the study Diagnosis
Number of patients
Meningioma
9
Chondrosarcoma Chordoma
2 1
Low grade glioma Primary neuroectodermal tumour
2 1
High grade glioma – retreatment High grade glioma – localized boost
2 7
Metastasis
7*
Thirteen patients were treated using multiple arc stereotactic radiotherapy, 15 with fixed beam stereotactically-guided conformal radiotherapy, and three using both techniques. *Of these, four patients had single fraction stereotactic radiosurgery.
To make an informed decision on the size of the margin when expanding the CTV to the PTV, it is necessary to measure the accuracy of relocation. Portal imaging and computer image analysis tools can be used to compare planning or simulator verification films with portal films or electronic images. However, they do not provide information on small movements, can only review two dimensions at a time, and they are labour intensive. Moreover, there are important manpower issues if these techniques are to be used ‘on line’, during a radiotherapy treatment. To define the appropriate CTV to PTV margin we have used a depth helmet and measuring probe to determine the accuracy of relocation of the patient within the GTC frame. These measurements have been processed in a spreadsheet designed to calculate the displacement in each direction, and as a combined displacement vector. This vector gives a total 3 dimensional displacement.
PATIENTS AND METHODS
In the four years up to December 2000, all patients immobilized for radiotherapy using the relocatable GTC stereotactic head frame were entered into the study of set-up accuracy. A total of 31 patients with varying diagnoses were included (Table 1), with a total of 564 measurement episodes. Twenty-seven patients received fractionated stereotactically-guided conformal radiotherapy, and 4 underwent single fraction high dose stereotactic radiosurgery. The basic design for the GTC frame was developed by Gill and colleagues in London (Gill SS 1987, UK Patent Application N 8728150) [2,3]. The depth helmet was an addition, designed and added to the frame in Boston, U.S.A., where it was named the ‘Depth Confirmation Helmet’ [4]. The GTC frame comprises a standard head ring to which an individualized dental impression of
the upper teeth and a moulded occipital support are attached [2,3]. The individualized components are manufactured in the mould room. Once the frame is located on the patient using the mouth bite, it is held firmly in position with a set of straps across the vertex. Once the frame is fitted, the patient lies supine in the treatment position, and the frame is attached to brackets on the treatment couch. The depth helmet attaches securely to the head frame, once the patient is positioned. The helmet is a Perspex hemisphere with a series of 24 evenly spaced measuring portals. Using a metal measuring probe inserted into each portal, a set of measurements is made from the outside edge of the helmet portal to the patient’s scalp. The helmet is removed before the localization box is attached to position the isocentre. All patients had sufficiently good dentition for relocation of the mouth bite. Measurements were taken at each planning, verification and treatment episode. At the first visit, when the mouth bite and occipital plate were made, the frame was fitted once and a set of measurements taken. The frame was removed completely and refitted before another series of measurements were taken. This ensured at the first visit that reproducibility could be achieved. The next visit was for the CT localization scan, and the data set taken at that appointment was used as the baseline for future comparison. At each visit, the frame was refitted until the measurements were in agreement with those taken at the localization scan. Initially, acceptance criteria for relocation were formulated empirically: the frame was considered to have relocated acceptably provided that the differences from the baseline measurement set were no worse than one reading >2 mm, or three readings >1.5 mm. When these tolerances were not met, the frame was refitted and measurements repeated. These acceptance criteria allow rapid assessment of whether the fitting is acceptable. However, they do not describe the true displacement of the patient within the frame since only individual measurements are considered, nor do they account for differences between radiographers. For these reasons, a spreadsheet was designed to calculate the 3-dimensional vector displacement of the patient’s position within the frame compared to the baseline position. This method uses information from every measurement to calculate the patient’s position within the frame, and thus estimates the displacement of the treatment isocentre from the planned isocentre position. For each measurement, the displacement in the x, y and z dimensions, and the length of the displacement vector (given by √[x2 +y2 +z2]) are calculated using simple trigonometry from the position and angle of the measuring portal; see Appendix 1 for details. From a standard template, a separate spreadsheet was established for each patient. This was used to calculate the displacements on a fraction by fraction basis, as well as the mean and standard deviation for the completed course. The individual spreadsheets can be linked to produce
population data. A copy of the spreadsheet is available on request. Measurements were recorded from as many portals in the depth helmet as possible, in order to maximize data used to calculate the displacement. Measurements were not recorded where the pointer failed to reach to the patient’s head or where they fell on the edge of the velcro strap. Measurements onto the occipital plate were recorded, but not used for calculation. Thus, typically four portals were omitted from the calculation. All readings recorded in the baseline set were measured at all subsequent episodes. To minimise inter-observer variation, the radiographer who obtained the baseline readings would remain the measurer for all other episodes when possible, but this was not always practicable for patients having fractionated treatment. A maximum of three radiographers acted as measurer for any one patient in a fractionated course. In order to assess probe measurement uncertainty and inter-observer differences, an experiment was performed using a volunteer immobilized in the frame. Three radiographers each measured 5 complete data sets, making a total of 15 in all, without the frame being repositioned. The anthropomorphic phantom was not considered suitable for this, because its surface does not have the elasticity of the scalp and will not permit multiple measurements. During the early part of the study all measurements were entered into the spreadsheet after completion of the radiotherapy course. Later in the study, for one patient undergoing radiosurgery, the spreadsheet was used to verify the accuracy of relocation after the frame had been fitted in the treatment room, and prior to commencement of the exposure.
RESULTS Individual Patient Results
Individual data for all patients are presented in Table 2, showing the mean displacements in the x, y and z directions, and the vector length displacements. Data for the three directions of movement were analysed separately, and plotted against time. The anterior–posterior (AP) displacements for each treatment for three representative patients are shown in Fig. 1. Displacements in the other two dimensions showed a similar pattern. These graphs were used to evaluate the possibility of increasing or decreasing accuracy of relocation during the course of a fractionated treatment. In all three directions, the accuracy of relocation for an individual patient remained largely constant through the treatment. In a few patients, the accuracy of relocation improved during the course, but in two patients it worsened, particularly in the superior–inferior (SI) direction, indicating that early readings did not predict accuracy later in the course.
33
Table 2 – Data for 564 measurement episodes in all 31 patients, showing the individual mean values of displacements in the x, y and z directions, and the vector length displacements. All measurements are in millimetres Mean values (mm) Patient no 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 Mean SD
No of readings
Ant–Post
Left–Right
Sup–Inf
Vector length
3 5 6 5 6 2 12 27 30 11 30 6 2 2 30 30 30 30 5 30 30 30 2 6 5 39 30 30 30 30 30
0.6 0.4 0.4 0.5 0.8 0.8 0.3 0.7 0.3 0.3 1.0 1.3 0.5 0.3 0.6 0.2 0.3 0.3 0.0 0.3 0.3 0.0 0.0 0.0 0.3 0.2 0.5 0.4 0.0 0.5 0.5
2.3 0.7 0.6 0.0 0.8 0.4 1.4 0.0 0.4 1.2 0.8 0.2 0.8 0.7 0.2 0.2 0.3 0.7 0.3 0.9 0.2 1.3 0.2 0.9 1.5 0.3 0.6 0.8 0.5 0.1 0.3
1.3 0.8 0.1 0.9 0.3 0.6 0.2 0.3 0.0 0.2 0.3 0.7 0.6 0.4 0.3 0.9 1.2 0.8 0.1 0.2 0.7 0.6 0.3 0.2 1.6 0.7 0.9 0.7 0.1 0.4 0.2
2.8 1.2 0.9 1.3 1.2 1.3 1.6 1.2 0.7 1.3 1.5 1.8 1.2 1.0 1.0 1.1 1.3 1.4 0.6 1.1 1.0 1.8 1.2 1.0 2.4 1.0 1.3 1.2 1.0 0.9 1.0
0.1 0.6
0.1 0.9
0.4 0.7
1.2 —
Overall Results
To assess the overall accuracy of relocation within the GTC frame, the 564 data sets were pooled (Table 3 and Figs 2 & 3). Figures 2a–2c show the frequency distribution of displacements in each direction, conforming to a normal distribution. The mean displacement away from the correct isocentre position for the entire data set is extremely small, the greatest mean displacement reaching 0.4 mm in the SI direction. In the x, y and z directions, 97% of the treatment episodes had measured displacements of <2 mm. Although the greatest displacement for any patient was 3.5 mm, less than 1% of all treatments had displacements greater than 2.5 mm. The largest displacements in each direction did not occur in the same patients. Figure 3 shows the frequency distribution of the length of the displacement vector. This distribution is skewed rather than normal, with a mean of 1.2 mm. Ninety-two percent of the displacement vectors were less than 2 mm, and only 3% exceeded 2.5 mm (Table 3).
34 2.50 patient 11 patient 20 patient 26
2.00
Displacement (mm)
1.50 1.00 0.50 0.00 –0.50 –1.00 –1.50 –2.00 –2.50
2
4
6
8
10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 Episode No.
Fig. 1 – Plot of displacements against time in the anterior–posterior direction, for 3 representative patients. Displacements in the lateral and superior–inferior directions (not shown) are similar.
Scrutiny of the raw measurements, using the initial acceptance criteria, successfully excluded major deviations. However, the actual frame displacement, which was calculated from the entire measurement set using the spreadsheet, was always smaller than the variation in individual measurements.
Measurement Uncertaintly and Inter-observer Differences
Verifying Frame Integrity and Position
In one patient, the occipital plate accidentally became detached from the frame. Depth helmet measurements onto the occipital plate were available and this allowed the plate to be repositioned correctly. In another patient the mouth bite plate loosened spontaneously. This was not apparent until depth helmet measurements were carried out, which demonstrated that the patient was incorrectly positioned. The loose mouth bite was tightened and the patient subsequently refitted with acceptTable 3 – Pooled data for the mean and range of set up deviations Percentage of readings within
Direction
Mean*
Standard deviation*
Ant–Post Left–Right Sup–Inf
0.1 0.1 0.4
0.6 0.9 0.7
99.1 92.1 94.6
99.6 97.1 97.8
99.8 98.9 99.2
–
75.1
91.6
97.1
1.5 mm
2 mm
able accuracy. In a third patient, the mouth bite broke, and after repair the position was verified using repeat depth helmet measurements. In all these cases, patient positioning was confirmed without the need for re-verification in the simulator.
The data for each measuring portal in the 15 volunteer measurement sets were used to calculate the mean depth for that portal. These figures were used as the baseline measurement; the difference between this and each individual measurement was then calculated. The five data sets for each observer were considered separately. For all three observers variation between the data sets was less than 0.1 mm. The data sets for each observer were pooled, giving 100 measurement differences for each person. The mean difference for each observer was zero (less than 0.01 mm). Analysis of variance showed no difference between the three observers (F=0.049, P=0.95).
2.5 mm
DISCUSSION
Vector length†
1.2
*All measurements in millimetres. These values are also given in Table 2. †Vector length is given by √(x2 +y2 +z2) where x, y and z are the displacements in each direction (see Appendix 1). The standard deviation for the vector length is not shown, since the distribution is skewed rather than normal.
Quantitative Results
Overall, relocation was extremely accurate. The mean deviation of the group as a whole measured around 0.1 mm in AP and left–right (LR) directions, and 0.4 mm in the SI direction. This degree of accuracy of relocation is of the same magnitude as that described by other groups [2,3,5–9]. Assessment of the reproducibility of the original frame, carried out at St Bartholomew’s, without the
0
0 –3.5 –3.3 –3.1 –2.9 –2.7 –2.5 –2.3 –2.1 –1.9 –1.7 –1.5 –1.3 –1.1 –0.9 –0.7 –0.5 –0.3 –0.1 0.1 0.3 0.5 0.7 0.9 1.1 1.3 1.5 1.7 1.9 2.1 2.3 2.5 2.7 2.9 3.1 3.3 3.5
No. of measurement episodes –3.5 –3.3 –3.1 –2.9 –2.7 –2.5 –2.3 –2.1 –1.9 –1.7 –1.5 –1.3 –1.1 –0.9 –0.7 –0.5 –0.3 –0.1 0.1 0.3 0.5 0.7 0.9 1.1 1.3 1.5 1.7 1.9 2.1 2.3 2.5 2.7 2.9 3.1 3.3 3.5
0
–3.5 –3.3 –3.1 –2.9 –2.7 –2.5 –2.3 –2.1 –1.9 –1.7 –1.5 –1.3 –1.1 –0.9 –0.7 –0.5 –0.3 –0.1 0.1 0.3 0.5 0.7 0.9 1.1 1.3 1.5 1.7 1.9 2.1 2.3 2.5 2.7 2.9 3.1 3.3 3.5
No. of measurement episodes
No. of measurement episodes
35
60 (a)
50
40
30
20
10
Displacement (mm)
60 (b)
50
40
30
20
10
Displacement (mm)
60 (c)
50
40
30
20
10
Displacement (mm)
Fig. 2 – Frequency distribution of displacements of all 564 measurements in 31 patients, in (a) anterior–posterior, (b) left–right, and (c) superior–inferior dimensions.
36
50 40 30 20
4
4.2
3.8
3.6
3.4
3
3.2
2.8
2.6
2.4
2.2
2
1.8
1.6
1.4
1.2
1
0.8
0.6
0.4
0
0.2
10
0
No. of measurement episodes
60
Displacement (mm)
Fig. 3 – Frequency distribution of displacement vectors.
benefit of the depth helmet, suggested an accuracy of 0.3 mm with a standard deviation of 0.15 mm, and the overall accuracy of the system was considered to be 2.4 mm [2]. The Royal Marsden group assessed the accuracy of the GTC frame [3,5,6] using simulator films. The mean displacement was 1.0 mm, with a standard deviation of 0.6 mm, and a maximum of 2.3 mm. Including errors associated with the treatment machine, they estimated the maximum inaccuracy at around 3 mm [7]. In Boston, Kooy et al. [4] used the helmet to evaluate reproducibility, but measured only portals on the helmet which were in the true AP, LR and SI directions. From measurements on 20 patients they estimated a standard deviation of patient repositioning of about 0.4 mm. Other studies using different frames have also reported similar accuracy of relocation [8,9]. Use of our vector displacement spreadsheet has helped to define the set-up accuracy. Since the distances are calculated as displacement vectors, it largely removes variation due to differences in the pressure of the measurement, which may occur on different days or with different radiographers making the measurements. Nevertheless, the inter-observer variation is extremely small, as noted above. This shows that observers are interchangeable provided they are appropriately trained. Although it is our standard policy to keep to a minimum the number of different radiographers carrying out the head frame measurements, it is impossible to restrict this to only one individual for fractionated treatment. However, measurements in the SI dimension suffer from the weakness that there are no measuring portals inferior to the stereotactic frame to balance the superior ones. Thus in this dimension differences in the pressure exerted on the measuring probe by different operators are not compensated for. The straps do allow slight movement superior-inferiorly, and upward twisting of the mouth bite is more likely than AP movement due to the design characteristics of the mouth bite attachment. Our
conclusion is that the GTC frame may be slightly less accurate in the SI dimension, which is consistent with Kooy et al. [4], but the mean error here is still only 0.4 mm. A slightly larger displacement in the superior– inferior dimension is also a regular finding with both rigid and thermoplastic head shells [10–12]. These results have also allowed us to evaluate our original crude acceptance criteria: they successfully exclude major deviations, but do not give a true representation of the actual displacement, which is smaller than the unprocessed measurements suggest. Individual Patient Results and Patient Selection
All the patients in this series had good dentition, which was assessed before considering them for treatment in the stereotactic radiotherapy frame. The results show the average accuracy of relocation to be extremely good. In the x, y and z dimensions, under 3% of set-ups had a displacement of greater than 2 mm. The displacement vectors exceeded 2.5 mm in less than 3% of episodes. One patient treated in the frame had some difficulty cooperating due to anxiety and larger displacements than average were noted. In patients with severe anxiety or motor and sensory deficits, depth helmet measurements showed no value in the use of the stereotactic frame compared to a standard clear thermoplastic shell, illustrating the importance of careful patient selection. Changes Over Time Through a Fractionated Treatment
There is no consistent trend for either improving or worsening set-up as the course of treatment progresses (Fig. 1). This is reassuring because it suggests that the dental impression material does not become loose or damaged, and that other factors also remain unchanged.
A separate question is whether the CTV to PTV margin could be tailored to suit the patient, relying on measurements from the first few fractions. In a few cases displacements increased for a few fractions in the later part of the treatment, so that early measurements cannot be used to individualize the CTV to PTV margin.
Relocation of the Mouth Bite
The accuracy of the GTC frame is based on accurate relocation of the mouth bite. There is information specifically regarding the accuracy of mouth bite relocation, which supports use of this technique, from development work for a repositioning system for thermoplastic shells. Bova et al. [13] reported that a plate could be repositioned with a mean error of 0.48 mm, and a standard deviation of 0.39 mm. These measurements are consistent with our data.
37
siderably greater accuracy than standard head shells, particularly those of the thermoplastic type. Additional Advantages of the Depth Helmet
Readings from the depth helmet have allowed us to refit loosened mouth bite and occipital plates, and to repair a broken dental impression. These are hazards when the same frame is used for several patients, which is realistic in busy departments given the cost of commercial frame systems. Use of the depth helmet, with the vector displacement spreadsheet, makes it clear which patients set up accurately and for which patients the technique confers no advantage. Use of the depth helmet with the spreadsheet is a simple way to measure set-up accuracy to define the CTV to PTV margin, which is needed for implementation of stereotactic radiotherapy. It could therefore permit technology transfer of stereotactic radiotherapy into centres starting this techniques.
Comparison with Head Shell Systems
Time Taken for Depth Helmet Measurements
The Heidelberg group have developed a relocatable stereotactic radiotherapy system using a mask substantially more rigid than conventional thermoplastic, together with a stereotactic frame and localizer [10]. Using this immobilization device, with a photogrammetric apparatus to detect displacements, they report standard deviations of reproducibility in all 3 dimensions of AP 0.9 mm, LR 0.6 mm and SI 1.3 mm. Hamilton et al. [11] assessed the reproducibility of a head shell and frame system, akin to the Heidelberg type, using radio-opaque spheres embedded in a custom mouthpiece which was independent of the mask and frame. The reproducibility was analysed using portal films on which images of the spheres could be seen. In 104 set-ups on 12 patients they found a mean deviation in the position of the isocentre of 1.8 mm, with a standard deviation of 1.4 mm. The maximum displacement seen was 6 mm. Thermoplastic shells immobilize much less accurately. In the hands of the Heidelberg group the largest deviations occurred laterally, with a standard deviation of 4.6 mm, and a maximum of 8 mm [10]. This level of accuracy is consistent with other reports of cranial set-up accuracy [11,14–16]. Byhart et al. [14] found errors greater than 5 mm in 7% of treatments. Verhey et al. [15] reported reproducibility with errors of up to roughly 5 mm using a mask system. Hanna et al. [12] recorded set-up inaccuracy of the same order, with deviations of greater than 5 mm occurring in 1% of measurements in the AP direction but 13% in the SI direction. Gildersleve et al. [16] reported errors greater than 5 mm in 8% of set-ups in the AP and 14% in the SI directions. These published results clearly demonstrate that relocatable stereotactic radiotherapy frames, in a number of manifestations, can reposition patients with con-
A major concern with the system is the additional time required for the depth helmet measurements. The mean total treatment time, including depth helmet measurements, for 39 stereotactically-guided fixed field treatment episodes in 3 patients was 15.7 min (median 15.2, standard deviation 3.2, range 11.1–25 min). This is consistent with timings reported in the literature [6,13]. The equivalent standard treatment using a clear thermoplastic shell, that is a three-field plan with blocks on each field, takes an average of 12.5 min in our department [17]. Since use of the depth helmet includes assessment of accuracy, a more direct comparison should include portal films of each field. This increases the average treatment time to 18.8 min [17]. Thus, use of the stereotactic frame with the depth helmet takes on average 3 min longer than using a shell, but is quicker than taking portal films. Since the depth measurements provide daily verification of patient position for high precision treatments, this represents an excellent balance between speed and high precision. Use of Portal Imaging for Measurement of Positioning Accuracy
It is our standard practice to carry out simulator verification of fixed field radiotherapy treatment plans, prior to starting treatment. In this way, most systematic errors can be accounted for. In addition, portal images are routinely taken on each field, at the start of treatment. However, we have found that, with small field treatments, it is often impossible to find sufficient bony anatomy to compare the treatment set and simulator films. Where blocks are being used, double exposure films can be taken, but the field has to be opened very wide, causing unnecessary irradiation to normal tissue.
38
Beam films taken with the same frequency so depth helmet measurements could add sufficient dose to the lens of the eye, for example, to approach the threshold for cataract formation. After the first treatment, we prefer to rely on the depth helmet measurements, rather than irradiate more normal tissue. For arc treatments, the isocentre position is verified with CT. It is impractical with our equipment to take portal films of arc treatments. Whilst low-weighted orthogonal fixed fields could be added to arc treatment plans, and imaged on set, we prefer to use the depth helmet and spreadsheet to verify patient positioning. Defining the CTV to PTV Margins for Treatment
The discrepancies measured by the depth helmet relate to patient position within the frame. Set-up inaccuracies due to the treatment machine also occur, though they are not measured by this method. For fractionated treatment, we estimate the maximum discrepancy due to machine inaccuracies to be of the order of 1 mm. These should be added in quadrature to the patient set up inaccuracies measured from the head frame. For single fraction stereotactic radiosurgery, which we carry out using multiple arcs, we estimate possible machine inaccuracies to be up to 1.6 mm, and this should be added arithmetically to the measured head frame deviations. In both situations, using the mean measured displacement vector length, the total errors are extremely small. For fixed fields, these would amount to 1.6 mm, and for single fraction multiple arcs to 2.8 mm. Use of the mean displacement, however, would allow some fractions in some patients to under treat the PTV, albeit by a small amount. Ninety-seven percent of vector displacements are encompassed by a 2.5 mm margin. We have elected to routinely use a 2.5 mm margin for vector discrepancies. For fractionated treatment the total PTV margin becomes 2.5 mm plus 1 mm for machine inaccuracy; added in quadrature this gives a PTV margin of 2.7 mm. As a result of these calculations, for fixed field stereotactically-guided conformal radiotherapy our routine practice is to use a 3 mm margin. Use of the Relocatable Frame for Single Fraction Stereotactic Radiosurgery
A margin of 4.1 mm (2.5+1.6 mm added arithmetically), needed to guarantee coverage of the PTV for stereotactic radiosurgery, is larger than clinically acceptable. There is delicate balance between adequate coverage of the PTV and risk of normal tissue toxicity. For this reason the Boston group [4] recommend against the use of the relocatable frame for radiosurgery. Therefore, in patients receiving radiosurgery, we now routinely process the depth helmet measurements to verify that positioning is accurate to better than 1.4 mm vector displacement prior to treatment. This acceptance
criterion allows a maximum error of 3 mm, when added arithmetically to 1.6 mm for machine inaccuracy. Ideally, for radical radiosurgery where coverage of the PTV is critical, use of a skull-fixed frame to reduce the CTV–PTV margin to a minimum is preferable. CONCLUSIONS
Use of the depth helmet with the vector displacement spreadsheet has proved extremely effective as an adjunct to setting up patients in the relocatable stereotactic head frame. The technique is relatively quick and produces excellent day-to-day accuracy. It is valuable in selecting patients and developing stereotactic readiotherapy techniques. It may help centres wishing to start stereotactic radiotherapy. It has allowed us to choose a value for the CTV and PTV margin for stereotactic radiotherapy and radiosurgery in our department. For fractionated treatments we suggest a CTV to PTV margin of 3 mm. For single fraction stereotactic radiosurgery we assess the accuracy of relocation before the start of treatment, enabling us to use a minimum CTV–PTV margin. Acknowledgements. We wish to thank Mrs Carol Moxey for her help in preparing the manuscript, Mrs Kath Walker for her support, the radiographers in the department who carried out the depth helmet measurements, and Miss Eleanor Pinto for statistical advice. REFERENCES 1 Burnet NG, Wurm R, Nyman J, Peacock JH. Normal tissue radiosensitivity – how important is it? Clin Oncol 1996;8:25–34. 2 Thomson ES, Gill SS, Doughty D. Stereotactic multiple arc radiotherapy. Br J Radiol 1990;63(754):745–751. 3 Gill SS, Thomas DG, Warrington AP, Brada M. Relocatable frame for stereotactic external beam radiotherapy. Int J Radiat Oncol Biol Phys 1991;20(3):599–603. 4 Kooy HM, Dunbar SF, Tarbell NJ, et al. Adaptation and verification of the relocatable Gill–Thomas–Cosman frame in stereotactic radiotherapy. Int J Radiat Oncol Biol Phys 1994; 30(3):685–691. 5 Graham JD, Warrington AP, Gill SS, Brada M. A non-invasive, relocatable stereotactic frame for fractionated radiotherapy and multiple imaging. Radiother Oncol 1991;21(1):60–62. 6 Laing RW, Thompson V, Warrington AP, Brada M. Feasibility of patient immobilization for conventional cranial irradiation with a relocatable stereotactic frame. Br J Radiol 1993;66(791):1020–1024. 7 Warrington AP, Laing RW, Brada M. Quality assurance in fractionated stereotactic radiotherapy. Radiother Oncol 1994; 30(3):239–246. 8 Scott TW, Beach JL, Mendiondo OA. A precision repeat localization head frame for fractionated stereotactic radiotherapy. Med Dosim 1997;22(1):5–8. 9 Theodorou K, Kappas C, Tsokas C. A new non-invasive and relocatable immobilization frame for fractionated stereotactic radiotherapy. Radiother Oncol 1998 Jun;47(3):313–317. 10 Schlegel W, Pastyr O, Bortfeld T, Gademann G, Menke M, Maier-Borst W. Stereotactically guided fractionated radiotherapy: technical aspects. Radiother Oncol 1993;29(2):197–204. 11 Hamilton RJ, Kuchnir FT, Pelizzari CA, Sweeney PJ, Rubin SJ. Repositioning accuracy of a noninvasive head fixation system for stereotactic radiotherapy. Med Phys 1996;23(11):1909–1917. 12 Hanna CL, Slade S, Mason MD, Burnet NG. Translating radiotherapy Clinical Target Volumes into Planning Target Volumes for bladder and brain tumour patients. Clin Oncol 1999;11:93–98.
13 Bova FJ, Buatti JM, Friedman WA, Mendenhall WM, Yang CC, Liu C. The University of Florida frameless high-precision stereotactic radiotherapy system. Int J Radiat Oncol Biol Phys 1997; 38(4):875–882. 14 Byhart RW, Cox JD, Hornburg AG, et al. Weekly localisation films and detection of field placement deviations. Int J Radiat Oncol Biol Phys 1978;4:881–887. 15 Verhey LJ, Goitein M, McNulty P, Munzenrider JE, Suit HD. Precise positioning of patients for radiation therapy. Int J Radiat Oncol Biol Phys 1982;8(2):289–294. 16 Gildersleve J, Dearneley DP, Evans PM, et al. Reproducibility of patient positioning during routine radiotherapy, as assessed by an integrated megavoltage imaging system. Radiother Oncol 1995; 35:151–160. 17 Burnet NG, Routsis DS, Murrell P, et al. A tool to measure radiotherapy complexity and workload: derivation from the Basic Treatment Equivalent (BTE) concept. Clin Oncol 2001;13:14–23.
APPENDIX 1
Figure 4 shows horizontal and vertical sections through the helmet, to illustrate the method used to combine the depth helmet readings. The angle applies to all vertical planes spaced around the helmet. For each probe position a unit displacement ri in the direction shown can be divided into its x, y and z components: xi =ri cosi cosi yi =ri sini cosi zi =ri sini For each depth probe position, the vectors can be added, and normalized to give the overall displacement. Horizontal
θ
Vertical
φ
Fig. 4 – Horizontal and vertical sections through the helmet, to illustrate the method used to combine the depth helmet readings
39
Thus:
The effect of the algorithm can be demonstrated by considering first the 2D situation of the horizontal positions around the base of the helmet, for a patient displacement of (2.0 mm, 1.0 mm). For a set of depth differences of (1, 0.71, 2, 2.12, 1, 0.71, 2, 2.12), the algorithm would give: x =(0+0.5+2+1.5+0+0.5+2+1.5)/ (0+0.5+1+0.5+0+0.5+1+0.5) =8/4 =2.0 mm y =(1+0.5+01.51+0.5+01.5)/ (1+0.5+0+0.5+1+0.5+0+0.5) = 4/4 = 1.0 mm. If another operator pressed less hard, and all depths were decremented by 0.5 mm, the algorithm still returns the values x=2.0, y= 1.0. In the 3D case, for all 24 points, if a set of depths that gave (2, 1.0) were all decremented by 0.5 mm, the result would be (2.0, 1.0, 0.66). This is because the z dimension (superior–inferior) is measured in one direction only, whereas the x and y directions have opposing points. If all 24 points are present, an error of 0.5 mm in a single point can change the calculated overall displacement by at most 0.06 mm in any direction. In practice there are always a few points missing, typically 4, and care should be taken that the number of points is roughly balanced between anterior and posterior, and between left and right.