ORIGINALARBEIT
Assessment of patient setup errors in IGRT in combination with a six degrees of freedom couch Daniel Schmidhalter ∗ , Marco Malthaner, Ernst J. Born, Alessia Pica, Michael Schmuecking, Daniel M. Aebersold, Michael K. Fix, Peter Manser Division of Medical Radiation Physics and Department of Radiation Oncology, Inselspital, Bern University Hospital, and University of Bern, Switzerland Received 12 March 2013; accepted 27 November 2013
Abstract Purpose: The range of patient setup errors in six dimensions detected in clinical routine for cranial as well as for extracranial treatments, were analyzed while performing linear accelerator based stereotactic treatments with frameless patient setup systems. Additionally, the need for re-verification of the patient setup for situations where couch rotations are involved was analyzed for patients treated in the cranial region. Methods and Materials: A total of 2185 initial (i.e. after pre-positioning the patient with the infrared system but before image guidance) patient setup errors (1705 in the cranial and 480 in the extracranial region) obtained by using ExacTrac (BrainLAB AG, Feldkirchen, Germany) were analyzed. Additionally, the patient setup errors as a function of the couch rotation angle were obtained by analyzing 242 setup errors in the cranial region. Before the couch was rotated, the patient setup error was corrected at couch rotation angle 0◦ with the aid of image guidance and the six degrees of freedom (6DoF) couch. For both situations attainment rates for two different tolerances (tolerance A: ±0.5 mm, ±0.5◦ ; tolerance B: ±1.0 mm, ±1.0◦ ) were calculated. Results: The mean (± one standard deviation) initial patient setup errors for the cranial cases were -0.24 ± 1.21◦ , -0.23 ± 0.91◦ and -0.03 ± 1.07◦ for the pitch, roll and couch rotation axes and 0.10 ± 1.17 mm, 0.10 ± 1.62 mm and 0.11 ± 1.29 mm for the lateral, longitudinal and vertical axes, respectively. Attainment
Erhebung von Patienten-Lagerungsunsicherheiten in IGRT in Kombination mit einem Tisch mit sechs Freiheitsgraden Zusammenfassung Zweck: Der klinisch relevante Bereich, in welchem Unsicherheiten der Patientenlagerung in sechs Dimensionen sowohl bei kraniellen wie auch bei extrakraniellen Bestrahlungen auftreten, wurde für Linearbeschleunigerbasierte stereotaktische Bestrahlungen mit rahmenlosen Patienten-Positionierungs-Systemen untersucht. Zusätzlich wurde die Lagerungsunsicherheit für Patienten, welche im kraniellen Bereich bestrahlt wurden, als Funktion des Tischrotationswinkels ausgewertet. Methoden und Material: Mit Hilfe von ExacTrac (BrainLAB AG, Feldkirchen, Deutschland) wurden insgesamt 2185 initiale (d.h. nach dem Vorpositionieren des Patienten mit dem Infrarotsystem aber vor der bildgestützten Lagerung) Lagerungsunsicherheiten gemessen (1705 im kraniellen und 480 im extrakraniellen Bereich). Die Unsicherheiten wurden zudem als Funktion des Tischrotationswinkels mit Hilfe von 242 Lagerungsunsicherheiten im kraniellen Bereich untersucht. Vor der Tischrotation wurde die Lagerungsunsicherheit der Patienten beim Tischwinkel 0◦ bildgestützt und mit Hilfe des Tisches mit sechs Freiheitsgraden (6D-Tisch) korrigiert. Für beide Situationen wurde die Rate, mit welcher zwei verschiedene
∗ Corresponding author: Daniel Schmidhalter, Division of Medical Radiation Physics and Department of Radiation Oncology, Inselspital, Bern University Hospital, and University of Bern, Switzerland. E-mail: [email protected] (D. Schmidhalter).
Z. Med. Phys. 24 (2014) 112–122 http://dx.doi.org/10.1016/j.zemedi.2013.11.002 http://journals.elsevier.de/zemedi
D. Schmidhalter et al. / Z. Med. Phys. 24 (2014) 112–122
113
rate (all six axes simultaneously) for tolerance A was 0.6% and 13.1% for tolerance B, respectively. For the extracranial cases the corresponding values were -0.21 ± 0.95◦ , -0.05 ± 1.08◦ and -0.14 ± 1.02◦ for the pitch, roll and couch rotation axes and 0.15 ± 1.77 mm, 0.62 ± 1.94 mm and -0.40 ± 2.15 mm for the lateral, longitudinal and vertical axes. Attainment rate (all six axes simultaneously) for tolerance A was 0.0% and 3.1% for tolerance B, respectively. After initial setup correction and rotation of the couch to treatment position a re-correction has to be performed in 77.4% of all cases to fulfill tolerance A and in 15.6% of all cases to fulfill tolerance B. Conclusion: The analysis of the data shows that all six axes of a 6DoF couch are used extensively for patient setup in clinical routine. In order to fulfill high patient setup accuracies (e.g. for stereotactic treatments), a 6DoF couch is recommended. Moreover, re-verification of the patient setup after rotating the couch is required in clinical routine.
Toleranzen (Toleranz A: ±0.5 mm, ±0.5◦ ; Toleranz B: ±1.0 mm, ±1.0◦ ) erfüllt waren, berechnet. Resultate: Die Mittelwerte (± eine Standardabweichung) der initialen Lagerungsunsicherheiten der kraniellen Fälle waren -0.24 ± 1.21◦ , -0.23 ± 0.91◦ und -0.03 ± 1.07◦ für die Achsen Pitch, Roll und Tischrotationswinkel und 0.10 ± 1.17 mm, 0.10 ± 1.62 mm und 0.11 ± 1.29 mm für die laterale, longitudinale und vertikale Achse. Die Lagerungsunsicherheit war in 0.6%, resp. 13.1% aller kraniellen Fälle für alle sechs Achsen gleichzeitig innerhalb der Toleranz A, resp. der Toleranz B. Die Mittelwerte (± eine Standardabweichung) der initialen Patienten-Lagerungsunsicherheiten der extrakraniellen Fälle waren -0.21 ± 0.95◦ , -0.05 ± 1.08◦ und -0.14 ± 1.02◦ für die Achsen Pitch, Roll und Tischrotationswinkel und 0.15 ± 1.77 mm, 0.62 ± 1.94 mm und -0.40 ± 2.15 mm für die laterale, longitudinale und vertikale Achse. Die Patienten-Lagerungsunsicherheit war in 0.0%, resp. 3.1% aller extrakraniellen Fälle für alle sechs Achsen gleichzeitig innerhalb der Toleranz A, resp. der Toleranz B. Nach initialer Lagerungskorrektur und anschliessendem Rotieren des Tisches in die Bestrahlungsposition waren in 77.4% resp. 15.6% aller Fälle erneute Lagerungskorrekturen nötig, um die Toleranz A resp. Toleranz B zu erfüllen. Schlussfolgerung: Unsere Daten zeigen, dass alle sechs Achsen eines 6D-Tisches extensiv für die Patientenlagerung in der klinischen Routine gebraucht werden. Um hohe Genauigkeiten bei der Patientenlagerung zu erreichen (z.B. für stereotaktische Bestrahlungen), ist ein 6D-Tisch empfohlen. Ausserdem zeigen die Resultate, dass nach einer Tischrotation eine erneute Verifikation der Patientenlagerung notwendig ist.
Keywords: Patient setup errors, 6DoF couch, IGRT, stereotactic radiotherapy
Schlüsselwörter: Patienten-Lagerungsunsicherheit, 6D-Tisch, IGRT, Stereotaktische Radiotherapie
1 Introduction Image guidance is a modern method to achieve high patient setup accuracy in radiotherapy [1]. Increasing the patient setup accuracy potentially allows margin reduction [2]. Margin reduction in turn has the potential of better normal tissue sparing and/or dose escalation in the tumor. Both may result in an improvement in patient care. A method where high patient setup accuracy is essential is stereotactic radiosurgery or stereotactic radiotherapy. Typically, high doses are delivered to small targets in one or a few sessions only. To achieve the required patient setup accuracy, invasive head frames are often used to immobilize the patient on the treatment couch. The invasive head frame is screwed to the patient’s skull as well as to the couch itself which leads to a high reproducibility in
patient setup between the acquisition of the reference planning CT and the treatment. Additionally, the inter- and intrafraction motion is minimized. Nevertheless, there are several drawbacks when using the invasive head frame. One drawback is that mounting the head frame is an invasive procedure such that local anesthesia is required. Once the invasive head frame is mounted, it has to stay in place until the whole treatment is finished. For that reason fractionation of the treatment is not possible. Additionally, the whole procedure is uncomfortable for the patient. Another drawback of the invasive head frame is that image guided radiation therapy (IGRT) is difficult because of the metallic part of the invasive head frame causing artifacts. A new approach to face these drawbacks is to use non-invasive frameless systems for patient setup [3,4]. With this non-invasive approach, e.g. by immobilizing
114
D. Schmidhalter et al. / Z. Med. Phys. 24 (2014) 112–122
the patient using a frameless mask system, it is possible to roughly position the patient on the treatment couch very easily within a certain range of accuracy (see later in this work). This approach is much more comfortable for the patient and allows a fractionated treatment scheme. In modern radiation therapy, stereotactic treatments are performed not only for cranial sites, but also for extracranial sites. Recently stereotactic body radiation therapy (SBRT) is of increasing interest. For both treatment sites (cranial and extracranial) IGRT is performed to achieve the required patient setup accuracy. Most IGRT systems employ a four degrees of freedom (4DoF) couch. However, modern IGRT systems are able to deliver patient setup errors in six dimensions. A 4DoF couch is only able to correct the patient setup regarding the three translational axes, i.e. longitudinal (long), lateral (lat) and vertical (vert), as well as the couch rotation angle (rot). A further improvement of the patient setup accuracy is achieved by combining IGRT with a 6DoF couch. The two additional rotational axes pitch and roll available by a 6DoF couch can also be used to correct patient setup errors [5]. Thus, with the available pitch and roll axes of the 6DoF couch, the patient benefits from the whole information delivered by the six degrees of freedom registration algorithm. Earlier studies showed the accuracy of the method with the aid of phantoms [6–8]. Only a few studies analyzed patient setup errors combining IGRT with a 6DoF couch. Two studies were published recently by Gevaert et al. [5] and Beltran et al. [9] where patient setup errors were investigated. However, the studies were restricted to the cranial region only. The aim of this work is to assess and analyze patient setup errors in six dimensions, i.e. to analyze in which range setup corrections are applied by means of a 6DoF couch in clinical routine for cranial as well as for extracranial cases in order to demonstrate the need of a 6DoF couch when performing linear accelerator based stereotactic treatments with frameless patient setup systems. Additionally, the patient setup error as a function of the couch rotation angle was analyzed for treatments in the cranial region, where noncoplanar fields are typically used. Attainment rates for two different patient setup accuracy tolerances are calculated.
2 Materials and Methods At our institution, treatments requiring high patient setup accuracy (e.g. stereotactic treatments) are performed using the Novalis TX system which is a combination of a modern linear accelerator from Varian (Varian Medical Systems, Inc., Palo Alto, USA) and the ExacTrac system from Brainlab (BrainLAB AG, Feldkirchen, Germany). ExacTrac is a patient positioning system consisting of an infrared system, an X-ray system and a robotic 6DoF couch [6–8,10–12]. The ExacTrac system was calibrated routinely following the Brainlab standard calibration procedures and using dedicated Brainlab calibration phantoms. The calibration procedure as well as the resulting accuracy of the ExacTrac system is reported in literature [6,8,12]. The verification of the calibration of the
infrared and the X-ray system was done on a weekly and daily frequency, respectively. To assess the range of setup corrections applied by means of a 6DoF couch in clinical routine, the ExacTrac was used as a measurement tool. To understand how and when (during the patient positioning process) the assessment was performed, an overview of the workflow for patient positioning using the ExacTrac in our clinic is described in the following. A flowchart of the workflow from patient setup to treatment when using the ExacTrac of the Novalis TX is given in Fig. 1. In a first step, the patient is positioned on the 6DoF couch. The immobilization of the patients treated in the cranial region was performed with the Brainlab frameless mask system [13,14]. A nose bridge and no mouthpiece were used. For patients treated in the extracranial region the immobilization was achieved with the aid of a vacuum bag cushion (VBC) system (IT-V, Innsbruck, Austria) on which the infrared markers are positioned. The couch is in the zero position during this step, which means that pitch, roll and couch rotation angle are all set to 0◦ . Afterwards, the patient is prepositioned with the aid of the infrared system such that the patient is shifted to treatment position by moving the couch. Note that only the three translational axes of the 6DoF couch are used during this step, i.e. the three rotational axes are inactive at that moment. This step is referred to as initial patient setup in the following. The initial patient setup is then verified by using the x-ray component of ExacTrac: two kV-images are acquired (one image per x-ray tube of the ExacTrac) and matched to the corresponding reference digitally reconstructed radiograph (DRR). This matching is performed automatically by the ExacTrac system using a corresponding matching algorithm [8]. The result of this procedure is the initial patient setup error in six dimensions. The verification of the initial patient setup is referred to as initial verification and the setup error resulting from this initial verification is referred to as initial setup error in this work. If the initial setup error is outside of a specified tolerance (see below) the determined correction is applied with the aid of the 6DoF couch including all six axes. In order to verify if the application of the correction has been performed accurately, two kV-images with ExacTrac have been acquired again. In the next step, all treatment fields for the actual couch rotation angle (which is 0◦ at this moment) are applied. Usually, the treatment plan contains non-coplanar fields when treating volumes in the cranial region. That means that the couch has to be rotated after the treatment of the fields at couch rotation angle 0◦ . Once the couch is rotated, the new patient position is verified with the aid of the x-ray system of ExacTrac. Again, if the resulting setup error is outside of the specified tolerance (see below), the corresponding correction is applied with the aid of the 6DoF couch, and again verified by kV- images using the ExacTrac until the setup error is within the specified tolerance. The whole procedure is repeated for every further couch rotation angle. To determine the clinical range in which a 6DoF couch is used to correct patient setup errors in six dimensions,
D. Schmidhalter et al. / Z. Med. Phys. 24 (2014) 112–122
initial patient setup infrared couch rotation angle 0º
115
couch rotation angle 0º
initial verification x-ray
setup error within tolerance?
no
verification x-ray
apply correction shift
setup error within tolerance?
no
yes
yes
apply treatment field(s)
additional treatment fields available? (rot ≠ 0º)
no
treatment finished
yes rotate couch to next couch rotation angle
couch rotation angle ≠ 0º
verification x-ray
setup error within tolerance?
no
verification x-ray
apply correction shift
no
yes
apply treatment field(s) for the corresponding couch rotation angle
setup error within tolerance?
yes
Figure 1. Flowchart of the workflow from patient setup to treatment when using the Novalis TX in combination with ExacTrac at our institution.
the initial verification step was analyzed in detail. 2185 initial verifications of 242 patients were analyzed in total. Patients undergoing radiosurgery (n = 108) as well as fractionated treatments (n = 134) were taken into account. 1705
initial verifications correspond to patients treated in the cranial region. 480 initial verifications correspond to treatments in the extracranial region. For all patients included in the extracranial subgroup, the target regions were located in that way, that
116
D. Schmidhalter et al. / Z. Med. Phys. 24 (2014) 112–122
Table 1 The table shows the attainment rates for the initial setup error according to tolerance A (±0.5 mm, ±0.5◦ ) and tolerance B (±1.0 mm, ±1.0◦ ). The investigated cases are arranged in three groups: All cases together, cranial cases only and extracranial cases only. Attainment rates are given for each group and for every single couch axis in the columns lat, long, vert, pitch, roll and rot. The attainment rates when all initial setup errors of the six axes have to be within tolerance simultaneously are given in the last column sim.
cranial & extracranial cranial extracranial
tolerance A tolerance B tolerance A tolerance B tolerance A tolerance B
lat [%]
long [%]
vert [%]
pitch [%]
roll [%]
rot [%]
sim [%]
38.3 63.7 43.0 70.8 21.7 38.3
26.4 48.1 28.9 51.9 17.5 34.4
30.8 51.8 36.0 60.1 12.1 22.1
37.1 63.6 34.3 61.4 47.3 71.5
45.0 72.1 45.7 73.4 42.7 67.5
36.5 66.9 36.0 65.8 38.3 70.6
0.5 10.9 0.6 13.1 0.0 3.1
matching on bony anatomy was feasible. No implanted markers were used. Typical examples for treated sites were spine or bone metastases in the pelvic region. Statistical analysis of the initial setup errors was performed by calculating corresponding histograms for each of the six couch axes separately. This analysis was performed for the cranial cases and extracranial cases separately as well as for all 2185 cases together. Two different tolerances were defined. The first allows the translational setup error to be within ±0.5 mm and the rotational setup error to be within ±0.5◦ (tolerance A), i.e. setup errors larger than ±0.5 mm and/or 0.5◦ will be corrected. This tolerance is a conservative lower limit of the accuracy of the ExacTrac system [6–8,10–12]. The second tolerance allows the translational setup error to be within ±1.0 mm and the rotational setup error to be within ±1.0◦ (tolerance B). When talking about radiosurgery and stereotactic radiotherapy, overall geometrical accuracy below 1 mm is often discussed [15–18]. Assuming that patient positioning with ExacTrac is the only error source contributing to this 1 mm error budget, this tolerance B was meant to be an upper limit for patient setup with ExacTrac in clinical routine. At this point, it is important to note that there exist other error sources which have an influence on the overall geometrical accuracy. According to these tolerances A and B, attainment rates for each of the six couch axes were calculated. In clinical routine a correction shift and rotation is applied if the initial setup error of only one of the six axes is outside the tolerance. Thus, the attainment rate was calculated for the situation where the initial setup errors of all the six axes were within tolerance simultaneously, i.e. where no correction has to be applied at all. As described above, the patient is just prepositioned with the aid of the infrared system. For that reason, the initial setup errors are expected to be the largest during the whole patient setup procedure and are suitable for determining the clinical range in which a 6DoF couch is used to correct patient setup errors in six dimensions. In the second part of this work the setup error after rotating the couch was analyzed. Note that at the moment of couch rotation, the patient setup has already been verified and corrected if necessary at couch rotation angle 0◦ . In a perfect world, no additional setup error would be expected when rotating the couch. Nevertheless, different error sources (instability of
the couch rotation center or patient movement) lead to the fact that such additional setup errors occur in clinical routine. For that reason, the setup errors found during patient setup verification after rotating the couch were analyzed as a function of the couch rotation angle. 242 verifications of 54 patients treated in the cranial region were analyzed. This analysis was performed for each of the six couch axes separately. In analogy as described in the previous paragraph, attainment rates for the two tolerances A and B were calculated.
3 Results The histograms showing the initial setup errors for all 2185 initial verifications (cranial and extracranial cases) are plotted in Fig. 2. The histograms for the three translational and the three rotational axes are shown separately. The mean initial setup error (± one standard deviation) for the lat, long and vert axes are 0.11 ± 1.32 mm, 0.21 ± 1.70 mm and 0.00 ± 1.54 mm, respectively. The mean initial setup error (± one standard deviation) for the pitch, roll and rot axes are -0.23 ± 1.16◦ , -0.19 ± 0.95◦ and -0.06 ± 1.06◦ , respectively. The areas in green and orange are visualizing the range where the initial setup error is within tolerance A (±0.5 mm, ±0.5◦ ) and/or tolerance B (±1.0 mm, ±1.0◦ ), respectively. The corresponding attainment rates are quantified in Table 1. The largest attainment rates were found for the roll axis. For this axis, 45.0% and 72.1% of all initial setup errors were within tolerance A and tolerance B, respectively. If the initial setup errors of all six axes have to be within tolerance simultaneously (which is the case in clinical routine), the attainment rate for tolerance A is 0.5%. For tolerance B this attainment rate was found to be 10.9%. Focusing on the two rotational axes pitch and roll only (which make the difference between a conventional 4DoF couch and a 6DoF couch), the data shows that during initial patient setup, in 50% of all cases the tolerance of 1◦ for the pitch and/or roll axis was exceeded. The data obtained during the initial verification was analyzed for cranial and extracranial cases separately. The histograms showing the initial setup errors for the 1705 initial verifications performed in the cranial region are plotted in Fig. 3. The mean initial setup error (± one standard deviation) for the lat, long and vert axes are 0.10±1.17 mm,
D. Schmidhalter et al. / Z. Med. Phys. 24 (2014) 112–122
9.09
117
(a) lateral longitudinal vertical
8.08 7.07
rel. frequency [%]
cranial and extracranial 6.06
tolerance A (± 0.5 mm)
2185 initial verifications 5.05
tolerance B (± 1.0 mm)
4.04 3.03 2.02
4.8
> 5.0
4.4
4.0
3.6
2.8
3.2
2.4
2.0
1.6
0.8
1.2
0.4
0.0
-0.4
-0.8
-1.2
-1.6
-2.0
-2.4
-2.8
-3.2
-3.6
-4.0
-4.4
-5.0-4.8-4.6-4.4-4.2-4.0-3.8-3.6-3.4-3.2-3.0-2.8-2.6-2.4-2.2-2.0-1.8-1.6-1.4-1.2-1.0-0.8-0.6-0.4-0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0 4.2 4.4 4.6 4.8 5.0
-4.8
0.00
> -5.0
1.01
initial setup error [mm]
12 12.0
(b) pitch roll rotation
rel. frequency [%]
10 10.0
cranial and extracranial 8.08
tolerance A (± 0.5º)
2185 initial verifications 6.06
tolerance B (± 1.0º) 4.04
> 4.0
3.6
3.2
2.8
2.4
2.0
1.2
1.6
0.8
0.4
0.0
-0.4
-0.8
-1.2
-1.6
-2.0
-2.4
-2.8
-3.2
< - -4.0 -3.8 -3.6 -3.4 -3.2 -3.0 -2.8 -2.6 -2.4 -2.2 -2.0 -1.8 -1.6 -1.4 -1.2 -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0 > 4.0 4.0
-3.6
0.00
< -4.0
2.02
initial setup error [º]
Figure 2. Histograms of the initial setup errors found during 2185 initial verifications of patients treated in the cranial (n = 1705) as well as in the extracranial (n = 480) region. The data was binned into 0.2 mm and 0.2◦ intervals, respectively. The areas in green and orange are visualizing the range where the initial setup error is within tolerance A (±0.5 mm, ±0.5◦ ) and tolerance B (±1.0 mm, ±1.0◦ ), respectively. (a) Histograms for the translational axes lat, long and vert. (b) Histograms for the rotational axes pitch, roll and rot.
0.10 ± 1.62 mm and 0.11 ± 1.29 mm, respectively. The mean initial setup error (± one standard deviation) for the pitch, roll and rot axes are -0.24 ± 1.21◦ , -0.23 ± 0.91◦ and 0.03 ± 1.07◦ , respectively. The attainment rates for tolerance A and B are quantified in Table 1. The largest attainment rates were found for the roll axis. For this axis, 45.7% and 73.4% of all initial setup errors were within tolerance A and tolerance B, respectively. If the initial setup errors of all six axes have to be within tolerance simultaneously, the attainment rate for tolerance A is 0.6%. For tolerance B an attainment rate of 13.1% was found. The histograms showing the initial setup errors for the 480 initial verifications performed in the extracranial region are plotted in Fig. 4. The mean initial setup errors (± one standard deviation) for the lat, long and vert axes are 0.15 ± 1.77 mm, 0.62 ± 1.94 mm and -0.40 ± 2.15 mm, respectively. The mean initial setup error (± one standard deviation) for the pitch, roll and rot axes are -0.21 ± 0.95◦ , -0.05 ± 1.08◦
and -0.14 ± 1.02◦ , respectively. The attainment rates for tolerance A and B are quantified in Table 1. The largest attainment rates were found for the pitch axis. For this axis 47.3% and 71.5% of all initial setup errors were within tolerance A and tolerance B, respectively. If the initial setup errors of all six axes have to be within tolerance simultaneously, the attainment rate for tolerance A is 0.0%. For tolerance B an attainment rate of 3.1% was found. All the mean initial setup errors are summarized in Table 2. The setup errors as a function of the couch rotation angle are plotted in Fig. 5 for the translational axes and in Fig. 6 for the rotational axes. The setup error per angle is visualized with the aid of a boxplot. For this analysis, the used couch rotation angle was binned into 10◦ -intervals. E.g. setup errors belonging to couch rotation angles in the interval]5◦ ,15◦ ] are visualized by the boxplot illustrated at 10◦ . The areas in green and orange are visualizing the range where the setup error is within tolerance A and/or tolerance B, respectively. The
118
D. Schmidhalter et al. / Z. Med. Phys. 24 (2014) 112–122
11 11.0
(a) 10 10.0
lateral longitudinal vertical
9.09
rel. frequency [%]
8.08
cranial
7.07
1705 initial verifications
6.06
tolerance A (± 0.5 mm)
5.05
tolerance B (± 1.0 mm)
4.04 3.03 2.02
> 5.0
4.8
4.4
4.0
3.2
3.6
2.8
2.4
2.0
1.2
1.6
0.8
0.4
0.0
-0.4
-0.8
-1.2
-1.6
-2.0
-2.4
-2.8
-3.2
-4.0
-3.6
-4.4
-5.0-4.8-4.6-4.4-4.2-4.0-3.8-3.6-3.4-3.2-3.0-2.8-2.6-2.4-2.2-2.0-1.8-1.6-1.4-1.2-1.0-0.8-0.6-0.4-0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0 4.2 4.4 4.6 4.8 5.0
-4.8
0.00
> -5.0
1.01
initial setup error [mm]
12 12.0
(b) pitch roll rotation
10 10.0
rel. frequency [%]
cranial 8.08
tolerance A (± 0.5º)
1705 initial verifications 6.06
tolerance B (± 1.0º) 4.04
3.6
3.2
2.8
2.4
2.0
> 4.0
initial setup error [º]
1.6
1.2
0.8
0.4
0.0
-0.4
-0.8
-1.2
-1.6
-2.0
-2.4
-2.8
-3.2
< - -4.0 -3.8 -3.6 -3.4 -3.2 -3.0 -2.8 -2.6 -2.4 -2.2 -2.0 -1.8 -1.6 -1.4 -1.2 -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0 > 4.0 4.0
-3.6
0.00
< -4.0
2.02
Figure 3. Histograms of the initial setup errors found during 1705 initial verifications of patients treated in the cranial region. The data was binned into 0.2 mm and 0.2◦ intervals, respectively. The areas in green and orange are visualizing the range where the initial setup error is within tolerance A (±0.5 mm, ±0.5◦ ) and tolerance B (±1.0 mm, ±1.0◦ ), respectively. (a) Histograms for the translational axes lat, long and vert. (b) Histograms for the rotational axes pitch, roll and rot. Table 2 The table summarizes the mean initial setup errors found for each of the six axes separately. The mean setup errors are given for the cranial and extracranial cases together as well as for the cranial and extracranial cases separately.
lat [mm] long [mm] vert [mm] pitch [◦ ] roll [◦ ] rot [◦ ]
cranial & extracranial
cranial
extracranial
0.11 0.21 0.00 -0.23 -0.19 -0.06
0.10 0.10 0.11 -0.24 -0.23 -0.03
0.15 0.62 -0.40 -0.21 -0.05 -0.14
corresponding attainment rates are quantified in Tab. 3. The largest attainment rates were found for the roll axis. For this axis, 94.6% and 100.0% of all setup errors after couch rotation were within tolerance A and tolerance B, respectively. If the setup errors of all six axes have to be within tolerance
simultaneously, the attainment rate for tolerance A is 32.6%. For tolerance B an attainment rate of 84.3% was found.
4 Discussion The clinical range in which a 6DoF couch is used to correct patient setup errors in six dimensions was investigated in this work. For this purpose, patient setup errors in six dimensions determined with the aid of ExacTrac in clinical routine were analyzed. The analysis of the patient setup errors was performed at two different points in the workflow: First during the initial patient setup at couch rotation angle 0◦ (where the biggest setup error is expected) and second after deflecting the couch. The histograms in Fig. 2 show that all six axes of the 6DoF couch are used extensively for patient setup in clinical routine. The variance of the distribution of the initial setup errors for the extracranial cases is higher than for the cranial cases,
D. Schmidhalter et al. / Z. Med. Phys. 24 (2014) 112–122
119
Table 3 The table shows the attainment rates for the setup error after couch rotation according to tolerance A (±0.5 mm, ±0.5◦ ) and tolerance B (±1.0 mm, ±1.0◦ ). Attainment rates are given for every single couch axis in the columns lat, long, vert, pitch, roll and rot. The attainment rates when all setup errors of the six axes have to be within tolerance simultaneously are given in the last column sim.
tolerance A tolerance B
lat [%]
long [%]
vert [%]
pitch [%]
roll [%]
rot [%]
sim [%]
64.0 90.5
65.3 95.5
85.5 97.9
91.5 98.8
94.6 100.0
89.3 99.2
32.6 84.3
which is confirmed and quantified by the attainment rates for the different tolerances in section III. One reason for this is the different immobilization strategy for the cranial and extracranial region. Another reason is the different deformability of the anatomy in the target region. E.g. the flex of the spine often differs slightly between planning CT and treatment. Changes in the flex of the spine were minimized with the aid of the VBC system, but a reduction to zero is most often not possible. Another example is the presence of the ribs, which are
positioned differently for each verification due to breathing. Even if most parts of the ribs are removed from the region of interest before performing the matching, the matching is affected by the ribs overlapping with the target region in the corresponding images. Compared to the deformable anatomy in the extracranial region the anatomy in the cranial region is much more rigid. The histograms in Figs. 3 and 4 shows that in clinical routine sometimes large initial setup errors which are even outside the travel range of the 6DoF couch
12 12.0 11 11.0
(a) lateral longitudinal vertical
10 10.0
rel. frequency [%]
9.09 8.08
extracranial
7.07
tolerance A (± 0.5 mm)
480 initial verifications
6.06
tolerance B (± 1.0 mm)
5.05 4.04 3.03 2.02
> 5.0
4.4
4.8
4.0
3.2
3.6
2.8
2.4
2.0
1.2
1.6
0.8
0.0
0.4
-0.4
-0.8
-1.2
-1.6
-2.0
-2.4
-2.8
-3.2
-4.0
-3.6
-4.4
-5.0-4.8-4.6-4.4-4.2-4.0-3.8-3.6-3.4-3.2-3.0-2.8-2.6-2.4-2.2-2.0-1.8-1.6-1.4-1.2-1.0-0.8-0.6-0.4-0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0 4.2 4.4 4.6 4.8 5.0
-4.8
0.00
> -5.0
1.01
initial setup error [mm]
16 16.0
(b) 14 14.0
pitch roll rotation
rel. frequency [%]
12 12.0
extracranial 10 10.0
tolerance A (± 0.5º)
480 initial verifications 8.08
tolerance B (± 1.0º) 6.06 4.04
3.6
3.2
2.8
2.4
2.0
1.6
1.2
0.8
0.4
0.0
-0.4
-0.8
-1.2
-1.6
-2.0
-2.4
-2.8
-3.2
-3.6
< - -4.0 -3.8 -3.6 -3.4 -3.2 -3.0 -2.8 -2.6 -2.4 -2.2 -2.0 -1.8 -1.6 -1.4 -1.2 -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0 > 4.0 4.0
< -4.0
0.00
> 4.0
2.02
initial setup error [º]
Figure 4. Histograms of the initial setup errors found during 480 initial verifications of patients treated in the extracranial region. The data was binned into 0.2 mm and 0.2◦ intervals, respectively. The areas in green and orange are visualizing the range where the initial setup error is within tolerance A (±0.5 mm, ±0.5◦ ) and tolerance B (±1.0 mm, ±1.0◦ ), respectively. (a) Histograms for the translational axes lat, long and vert. (b) Histograms for the rotational axes pitch, roll and rot.
120
D. Schmidhalter et al. / Z. Med. Phys. 24 (2014) 112–122
2.0
lateral 2.0
pitch
242 verifications
242 verifications 1.0 setup error [º]
setup error [mm]
1.0
0.0
0.0
-1.0 -1.0
tolerance A (± 0.5 mm) -2.0
tolerance A (± 0.5º)
tolerance B (± 1.0 mm)
tolerance B (± 1.0º) 40.0
50.0
60.0
70.0
80.0
90.0
40.0
50.0
60.0
70.0
80.0
90.0
40.0
50.0
60.0
70.0
80.0
90.0
20.0
30.0
10.0
-20.0
-10.0
couch rotation angle [º]
couch rotation angle [º]
2.0
longitudinal 2.0
-30.0
-50.0
-40.0
-60.0
-80.0
-70.0
-90.0
90.0
80.0
70.0
60.0
50.0
40.0
20.0
30.0
10.0
-20.0
-10.0
-30.0
-50.0
-40.0
-60.0
-80.0
-70.0
-90.0
-2.0
roll
242 verifications
242 verifications 1.0 setup error [º]
setup error [mm]
1.0
0.0
-1.0
0.0
-1.0
-2.0
tolerance A (± 0.5 mm)
tolerance A (± 0.5º)
tolerance B (± 1.0 mm)
tolerance B (± 1.0º) 30.0
20.0
10.0
-10.0
-20.0
couch rotation angle [º]
couch rotation angle [º]
2.0
vertical 2.0
-30.0
-40.0
-50.0
-60.0
-70.0
-80.0
-90.0
90.0
80.0
70.0
60.0
50.0
40.0
30.0
20.0
10.0
-10.0
-20.0
-30.0
-40.0
-50.0
-60.0
-70.0
-80.0
-90.0
-2.0
rotation 242 verifications
242 verifications 1.0 setup error [º]
setup error [mm]
1.0
0.0
0.0
-1.0 -1.0
tolerance A (± 0.5 mm) -2.0
tolerance A (± 0.5º)
tolerance B (± 1.0 mm)
tolerance B (± 1.0º)
couch rotation angle [º]
Figure 5. Setup errors as a function of the couch rotation angle for the three translational axes lat, long and vert. The setup error as a function of the couch rotation angle is visualized with the aid of a boxplot. Therefore, the couch rotation angle was binned into 10◦ intervals. E.g. setup errors belonging to couch rotation angles in the interval]5◦ ,15◦ ] are visualized by the boxplot illustrated at 10◦ . The areas in green and orange are visualizing the range where the setup error is within tolerance A (±0.5 mm, ±0.5◦ ) and/or tolerance B (±1.0 mm, ±1.0◦ ).
30.0
20.0
10.0
-10.0
-20.0
-30.0
-40.0
-50.0
-60.0
-70.0
-80.0
-90.0
90.0
80.0
70.0
60.0
50.0
40.0
30.0
20.0
10.0
-10.0
-30.0
-20.0
-40.0
-50.0
-60.0
-70.0
-80.0
-90.0
-2.0
couch rotation angle [º]
Figure 6. Setup errors as a function of the couch rotation angle for the three rotational axes pitch, roll and rot. The setup error as a function of the couch rotation angle is visualized with the aid of a boxplot. Therefore, the couch rotation angle was binned into 10◦ intervals. E.g. setup errors belonging to couch rotation angles in the interval]5◦ ,15◦ ] are visualized by the boxplot illustrated at 10◦ . The areas in green and orange are visualizing the range where the setup error is within tolerance A (±0.5 mm, ±0.5◦ ) and/or tolerance B (±1.0 mm, ±1.0◦ ).
D. Schmidhalter et al. / Z. Med. Phys. 24 (2014) 112–122
(e.g. for the pitch and/or roll axes) do occur in the extracranial as well as in the cranial region. Even if such large initial patient setup errors could be corrected with the aid of the 6DoF couch it is not a priori useful, especially not in the extracranial region. Large pitch and roll corrections could be noticed by the patient and the patient could try to compensate the correction rotation. Additionally, the anatomy of the patient may change in comparison with the reference CT if the pitch and/or roll angle differs too much from the situation during the acquisition of the reference CT. It is not possible to compensate such deformations in the patient anatomy with the aid of the 6DoF couch. In such cases, a possible approach is to restart the whole patient setup procedure from the beginning by taking the patient down from the couch. Large initial setup errors outside the travel range of the 6DoF couch in the cranial region do occur mainly during fractionated treatments where the patient is losing weight during the therapy which results in the fact that the mask is losing its tightness during treatment. In such a case, one possibility is to reduce the thickness of the spacers between the front and back part of the mask in order to tighten it. Nevertheless, in extreme cases a new mask has to be produced which means that a re-planning of the treatment has to be performed. When treating brain metastases, Gevaert et al. [5] reported mean translational setup errors (± one standard deviation) in the long, lat and vert axes of 0.41 ± 1.19 mm, -0.48 ± 1.58 mm and 0.06 ± 0.99 mm, respectively and mean rotational setup errors (± one standard deviation) in the pitch, roll and rot axes of -0.09 ± 0.72◦ , 0.23 ± 0.82◦ and -0.10 ± 1.03◦ , respectively. Beltran et al. [9] reported mean rotational errors (± on standard deviation) in the pitch roll and rot axes of 0.37 ± 1.04◦ , 0.29 ± 0.96◦ and 0.03 ± 0.99◦ , respectively for children with brain tumors treated in the supine position. These results are in concordance with the results found in our study. The patient setup error was analyzed as a function of the couch rotation angle. Before the couch is rotated, the patient setup was corrected at couch rotation angle 0◦ (initial patient setup correction). Our data shows that patient setup errors outside tolerance A (±0.5 mm, ±0.5◦ ) as well as outside tolerance B (±1.0 mm, ±1.0◦ ) occur after a couch rotation. There are mainly two error sources which lead to these results. The first error source is a discrepancy (within specified tolerances) between the treatment isocenter (which is represented by the ExacTrac isocenter) and the couch rotation center. The discrepancy between these two isocenters for our system was determined with a Winston-Lutz test and the results are in concordance with the data found in this work. The second reason is the patient itself who may move (intrafraction motion, e.g. due to the couch movement itself). E.g., Ramakrishna et al. [19] observed a mean (± standard deviation) three-dimensional intrafraction shift magnitude for the Brainlab frameless mask system of 0.7 ± 0.5 mm. Gevaert et al. [5] reported a mean (± standard deviation) three-dimensional intrafraction shift for the same frameless system of 0.58 ± 0.42 mm. Inaccuracies
121
in the system calibration would result in systematic setup errors as a function of the couch rotation error. The patient movement due to couch rotation would most probably result in a random error. Our data clearly shows these two types of errors. The inaccuracy in system calibration is confirmed by the plots in Fig. 5 and Fig. 6 where the setup error as a function of the couch rotation angle is shown. The axis of rotation when rotating the couch (change the couch rotation angle rot) is perpendicular to the lateral-longitudinal plane. For that reason, setup errors due to couch rotation are expected to be maximal for the lat and long axes and can be described by a periodic function. For geometrical reasons, the effect of inaccuracies in the system calibration due to couch rotation (systematic setup error) is negligible for the four other axes lat, long, vert, and rot. This is confirmed by Figs. 5 and 6. The median values of the setup errors per couch rotation angle are describing this systematic error. The boxplots are describing the random errors per couch rotation angle. Nevertheless, the systematic setup error resulting from miscalibration of the couch rotation center is detectable with the aid of ExacTrac, which is a room-based system. By following the workflow as described in the Materials and Methods section, that means by verifying the patient position always directly after the couch was shifted or rotated and before treatment, the setup error due to miscalibration of the couch will be detected and compensated with the aid of the 6DoF couch. The main limitation of a 4DoF couch in comparison to a 6DoF couch is that it is not possible to correct pitch and roll setup errors. To estimate the dosimetric consequences of pitch and/or roll rotational setup errors in general is difficult because the dosimetric consequences due to rotational setup are influenced by a lot of factors, e.g. the size and form of the PTV, the number of PTVs, the position of the organs at risks in relation to the PTV, the field arrangement, or the margin between the clinical target volume and the PTV. Several groups investigated these dosimetric consequences for specific cases. Beltran et. al [9] investigated the effect of these rotational errors for pediatric brain tumor patients and found undesirable changes in the gEUD for critical structures and planning target volume (PTV) for rotational setup errors larger than 2◦ . Gevaert et al. [5] investigated the dosimetric consequences for patients with brain metastases and found that the mean (± one standard deviation) conformity index decreased from 0.68 ± 0.08 (6DoF) to 0.59 ± 0.12 (4DoF). A loss of prescribed isodose coverage of 5% was found for 4DoF couch positioning. Half a degree for pitch and roll rotations was identified as a threshold for coverage loss. Guckenberger et al. [20] simulated the dosimetric improvements when using IGRT (including all six axes) for radiosurgery of brain metastases in comparison to non-IGRT radiosurgery.
5 Conclusion In this work we analyzed the range in which a 6DoF couch is used to correct patient setup errors in six dimensions in
122
D. Schmidhalter et al. / Z. Med. Phys. 24 (2014) 112–122
clinical routine. For that reason, initial patient setup errors were analyzed. Additionally, the patient setup error as a function of the couch rotation error was investigated. The analysis of the data shows that all six axes are extensively used for patient setup in clinical routine. In order to fulfill high patient setup accuracies (e.g. for stereotactic treatments), a 6DoF couch is recommended. We have shown that the high patient setup accuracy (realized at couch rotation angle 0◦ ) may be decreased when rotating the couch. That is why re-verification (including repositioning until the setup error is within tolerance) of the patient setup after rotating the couch is required in clinical routine.
References [1] Lamba M, Breneman JC, Warnick RE. Evaluation of image guided positioning for frameless intracranial radiosurgery. Int J Radiat Oncol Biol Phys 2009;74:913–9. [2] Verellen D, De Ridder M, Linthout N, Tournel K, Soete G, Storme G. Innovations in image-guided radiotherapy. Nat Rev Cancer 2007;7:949–60. [3] Kamath R, Ryken TC, Meeks SL, Pennington EC, Ritchie J, Buatti JM. Initial clinical experience with frameless radiosurgery for patients with intracranial metastases. Int J Radiat Oncol Biol Phys 2005;61:1467–72. [4] Breneman JC, Steinmetz R, Smith A, Lamba M, Warnick RE. Frameless image- guided intracranial stereotactic radiosurgery; clinical outcomes for brain metastases. Int J Radiat Oncol Biol Phys 2009;74:702–6. [5] Gevaert T, Verellen D, Engels B, Depuydt T, Heuninckx K, Duchateau M, et al. Clinical Evaluation of a Robotic 6-Degree of Freedom Treatment Couch for Frameless Radiosurgery. Int J Radiat Oncol Biol Phys 2012;83:467–74. [6] Jin JY, Yin FF, Tenn SE, Medin PM, Solberg TD. Use of the BrainLAB ExacTrac X- Ray 6D System in Image-Guided Radiotherapy. Med Dosim 2008;33:124–34. [7] Takakura T, Mizowaki T, Nakata M, Yano S, Fujimoto T, Miyabe Y, et al. The geometric accuracy of frameless stereotactic radiotherapy using a 6D robotic couch system. Phys Med Biol 2010;55:1–10. [8] Gevaert T, Verellen D, Tournel K, Linthout N, Bral S, Engels B, et al. Setup accuracy of the Novalis ExacTrac 6DOF system for frameless radiosurgery. Int J Radiat Oncol Biol Phys 2012;82:1627–35.
[9] Beltran C, Pegram A, Merchant TE. Dosimetric consequences of rotational errors in radiation therapy of pediatric brain tumors. Radiother Oncol 2012;102:206–9. [10] Schewe JE, Lam KL, Balter JM, Haken RKT. A room-based diagnostic imaging system for measurement of patient setup. Med Phys 1998;25:2385–8. [11] Kim J, Wen N, Jin JY, Walls N, Kim S, Li H, et al. Clinical commissioning and use of the Novalis TX linear accelerator for SRS and SBRT. J Appl Clin Med Phys 2012;13:124–51. [12] Verellen D, Soete G, Linthout N, Van Acker S, De Roover P, Vinh-Hung V, et al. Quality assurance of a system for improved target localization and patient set-up that combines real-time infrared tracking and stereoscopic X-ray imaging. Radiother Oncol 2003;67: 129–41. [13] Ali I, Tubbs J, Hibbitts K, Algan O, Thompson S, Herman T, et al. Evaluation of the setup accuracy of stereotactic head immobilization mask system using kV on- board imaging. J Appl Clin Med Phys 2010;10:3192. [14] Lightstone AW, Benedict AH, Bova FJ, Solberg TD, Stern RL. Intracranial stereotactic positioning systems: Report of the American Association of Physicists in Medicine Radiation Therapy Committee Task Group No. 68. Med Phys 2005;32:2380–98. [15] Bichay T, Dieterich S. Point/Counterpoint: Submillimeter accuracy in radiosurgery is not possible. Med Phys 2013;40:050601. [16] Antypas C, Pantelis E. Performance evaluation of a Cyberknife G4 image-guided robotic stereotactic radiosurgery system. Phys Med Biol 2008;53:4697–718. [17] Wang L, Kielar KN, Mok E, Hsu A, Dieterich S, Xing L. An end-toend examination of geometric accuracy of IGRT using a new digital accelerator equipped with onboard imaging system. Phys Med Biol 2012;57:757–69. [18] Ruschin M, Komljenovic PT, Ansell S, Ménard C, Bootsma G, Cho YB, et al. Cone beam computed tomography image guidance system for a dedicated intracranial radiosurgery treatment unit. Int J Radiat Oncol Biol Phys 2013;85:243–50. [19] Ramakrishna N, Rosca Florin, Friesen S, Tezcanli E, Zygmanszki P, Hacker F. A clinical comparison of patient setup and intra-fraction motion using frame-based radiosurgery versus a frameless imageguided radiosurgery system for intracranial lesions. Radiother Oncol 2010;95:109–15. [20] Guckenberger M, Roesch J, Baier K, Sweeney RA, Flentje M. Dosimetric consequences of translational and rotational errors in frame-less image-guided radiosurgery. Radiat Oncol 2012;7:63–78.
Available online at www.sciencedirect.com
ScienceDirect
Assessment of patient setup errors in IGRT in combination with a six degrees of freedom couch Daniel Schmidhalter ∗ , Marco Malthaner, Ernst J. Born, Alessia Pica, Michael Schmuecking, Daniel M. Aebersold, Michael K. Fix, Peter Manser Division of Medical Radiation Physics and Department of Radiation Oncology, Inselspital, Bern University Hospital, and University of Bern, Switzerland Received 12 March 2013; accepted 27 November 2013
Abstract Purpose: The range of patient setup errors in six dimensions detected in clinical routine for cranial as well as for extracranial treatments, were analyzed while performing linear accelerator based stereotactic treatments with frameless patient setup systems. Additionally, the need for re-verification of the patient setup for situations where couch rotations are involved was analyzed for patients treated in the cranial region. Methods and Materials: A total of 2185 initial (i.e. after pre-positioning the patient with the infrared system but before image guidance) patient setup errors (1705 in the cranial and 480 in the extracranial region) obtained by using ExacTrac (BrainLAB AG, Feldkirchen, Germany) were analyzed. Additionally, the patient setup errors as a function of the couch rotation angle were obtained by analyzing 242 setup errors in the cranial region. Before the couch was rotated, the patient setup error was corrected at couch rotation angle 0◦ with the aid of image guidance and the six degrees of freedom (6DoF) couch. For both situations attainment rates for two different tolerances (tolerance A: ±0.5 mm, ±0.5◦ ; tolerance B: ±1.0 mm, ±1.0◦ ) were calculated. Results: The mean (± one standard deviation) initial patient setup errors for the cranial cases were -0.24 ± 1.21◦ , -0.23 ± 0.91◦ and -0.03 ± 1.07◦ for the pitch, roll and couch rotation axes and 0.10 ± 1.17 mm, 0.10 ± 1.62 mm and 0.11 ± 1.29 mm for the lateral, longitudinal and vertical axes, respectively. Attainment
Erhebung von Patienten-Lagerungsunsicherheiten in IGRT in Kombination mit einem Tisch mit sechs Freiheitsgraden Zusammenfassung Zweck: Der klinisch relevante Bereich, in welchem Unsicherheiten der Patientenlagerung in sechs Dimensionen sowohl bei kraniellen wie auch bei extrakraniellen Bestrahlungen auftreten, wurde für Linearbeschleunigerbasierte stereotaktische Bestrahlungen mit rahmenlosen Patienten-Positionierungs-Systemen untersucht. Zusätzlich wurde die Lagerungsunsicherheit für Patienten, welche im kraniellen Bereich bestrahlt wurden, als Funktion des Tischrotationswinkels ausgewertet. Methoden und Material: Mit Hilfe von ExacTrac (BrainLAB AG, Feldkirchen, Deutschland) wurden insgesamt 2185 initiale (d.h. nach dem Vorpositionieren des Patienten mit dem Infrarotsystem aber vor der bildgestützten Lagerung) Lagerungsunsicherheiten gemessen (1705 im kraniellen und 480 im extrakraniellen Bereich). Die Unsicherheiten wurden zudem als Funktion des Tischrotationswinkels mit Hilfe von 242 Lagerungsunsicherheiten im kraniellen Bereich untersucht. Vor der Tischrotation wurde die Lagerungsunsicherheit der Patienten beim Tischwinkel 0◦ bildgestützt und mit Hilfe des Tisches mit sechs Freiheitsgraden (6D-Tisch) korrigiert. Für beide Situationen wurde die Rate, mit welcher zwei verschiedene
∗ Corresponding author: Daniel Schmidhalter, Division of Medical Radiation Physics and Department of Radiation Oncology, Inselspital, Bern University Hospital, and University of Bern, Switzerland. E-mail: [email protected] (D. Schmidhalter).
Z. Med. Phys. 24 (2014) 112–122 http://dx.doi.org/10.1016/j.zemedi.2013.11.002 http://journals.elsevier.de/zemedi
D. Schmidhalter et al. / Z. Med. Phys. 24 (2014) 112–122
113
rate (all six axes simultaneously) for tolerance A was 0.6% and 13.1% for tolerance B, respectively. For the extracranial cases the corresponding values were -0.21 ± 0.95◦ , -0.05 ± 1.08◦ and -0.14 ± 1.02◦ for the pitch, roll and couch rotation axes and 0.15 ± 1.77 mm, 0.62 ± 1.94 mm and -0.40 ± 2.15 mm for the lateral, longitudinal and vertical axes. Attainment rate (all six axes simultaneously) for tolerance A was 0.0% and 3.1% for tolerance B, respectively. After initial setup correction and rotation of the couch to treatment position a re-correction has to be performed in 77.4% of all cases to fulfill tolerance A and in 15.6% of all cases to fulfill tolerance B. Conclusion: The analysis of the data shows that all six axes of a 6DoF couch are used extensively for patient setup in clinical routine. In order to fulfill high patient setup accuracies (e.g. for stereotactic treatments), a 6DoF couch is recommended. Moreover, re-verification of the patient setup after rotating the couch is required in clinical routine.
Toleranzen (Toleranz A: ±0.5 mm, ±0.5◦ ; Toleranz B: ±1.0 mm, ±1.0◦ ) erfüllt waren, berechnet. Resultate: Die Mittelwerte (± eine Standardabweichung) der initialen Lagerungsunsicherheiten der kraniellen Fälle waren -0.24 ± 1.21◦ , -0.23 ± 0.91◦ und -0.03 ± 1.07◦ für die Achsen Pitch, Roll und Tischrotationswinkel und 0.10 ± 1.17 mm, 0.10 ± 1.62 mm und 0.11 ± 1.29 mm für die laterale, longitudinale und vertikale Achse. Die Lagerungsunsicherheit war in 0.6%, resp. 13.1% aller kraniellen Fälle für alle sechs Achsen gleichzeitig innerhalb der Toleranz A, resp. der Toleranz B. Die Mittelwerte (± eine Standardabweichung) der initialen Patienten-Lagerungsunsicherheiten der extrakraniellen Fälle waren -0.21 ± 0.95◦ , -0.05 ± 1.08◦ und -0.14 ± 1.02◦ für die Achsen Pitch, Roll und Tischrotationswinkel und 0.15 ± 1.77 mm, 0.62 ± 1.94 mm und -0.40 ± 2.15 mm für die laterale, longitudinale und vertikale Achse. Die Patienten-Lagerungsunsicherheit war in 0.0%, resp. 3.1% aller extrakraniellen Fälle für alle sechs Achsen gleichzeitig innerhalb der Toleranz A, resp. der Toleranz B. Nach initialer Lagerungskorrektur und anschliessendem Rotieren des Tisches in die Bestrahlungsposition waren in 77.4% resp. 15.6% aller Fälle erneute Lagerungskorrekturen nötig, um die Toleranz A resp. Toleranz B zu erfüllen. Schlussfolgerung: Unsere Daten zeigen, dass alle sechs Achsen eines 6D-Tisches extensiv für die Patientenlagerung in der klinischen Routine gebraucht werden. Um hohe Genauigkeiten bei der Patientenlagerung zu erreichen (z.B. für stereotaktische Bestrahlungen), ist ein 6D-Tisch empfohlen. Ausserdem zeigen die Resultate, dass nach einer Tischrotation eine erneute Verifikation der Patientenlagerung notwendig ist.
Keywords: Patient setup errors, 6DoF couch, IGRT, stereotactic radiotherapy
Schlüsselwörter: Patienten-Lagerungsunsicherheit, 6D-Tisch, IGRT, Stereotaktische Radiotherapie
1 Introduction Image guidance is a modern method to achieve high patient setup accuracy in radiotherapy [1]. Increasing the patient setup accuracy potentially allows margin reduction [2]. Margin reduction in turn has the potential of better normal tissue sparing and/or dose escalation in the tumor. Both may result in an improvement in patient care. A method where high patient setup accuracy is essential is stereotactic radiosurgery or stereotactic radiotherapy. Typically, high doses are delivered to small targets in one or a few sessions only. To achieve the required patient setup accuracy, invasive head frames are often used to immobilize the patient on the treatment couch. The invasive head frame is screwed to the patient’s skull as well as to the couch itself which leads to a high reproducibility in
patient setup between the acquisition of the reference planning CT and the treatment. Additionally, the inter- and intrafraction motion is minimized. Nevertheless, there are several drawbacks when using the invasive head frame. One drawback is that mounting the head frame is an invasive procedure such that local anesthesia is required. Once the invasive head frame is mounted, it has to stay in place until the whole treatment is finished. For that reason fractionation of the treatment is not possible. Additionally, the whole procedure is uncomfortable for the patient. Another drawback of the invasive head frame is that image guided radiation therapy (IGRT) is difficult because of the metallic part of the invasive head frame causing artifacts. A new approach to face these drawbacks is to use non-invasive frameless systems for patient setup [3,4]. With this non-invasive approach, e.g. by immobilizing
114
D. Schmidhalter et al. / Z. Med. Phys. 24 (2014) 112–122
the patient using a frameless mask system, it is possible to roughly position the patient on the treatment couch very easily within a certain range of accuracy (see later in this work). This approach is much more comfortable for the patient and allows a fractionated treatment scheme. In modern radiation therapy, stereotactic treatments are performed not only for cranial sites, but also for extracranial sites. Recently stereotactic body radiation therapy (SBRT) is of increasing interest. For both treatment sites (cranial and extracranial) IGRT is performed to achieve the required patient setup accuracy. Most IGRT systems employ a four degrees of freedom (4DoF) couch. However, modern IGRT systems are able to deliver patient setup errors in six dimensions. A 4DoF couch is only able to correct the patient setup regarding the three translational axes, i.e. longitudinal (long), lateral (lat) and vertical (vert), as well as the couch rotation angle (rot). A further improvement of the patient setup accuracy is achieved by combining IGRT with a 6DoF couch. The two additional rotational axes pitch and roll available by a 6DoF couch can also be used to correct patient setup errors [5]. Thus, with the available pitch and roll axes of the 6DoF couch, the patient benefits from the whole information delivered by the six degrees of freedom registration algorithm. Earlier studies showed the accuracy of the method with the aid of phantoms [6–8]. Only a few studies analyzed patient setup errors combining IGRT with a 6DoF couch. Two studies were published recently by Gevaert et al. [5] and Beltran et al. [9] where patient setup errors were investigated. However, the studies were restricted to the cranial region only. The aim of this work is to assess and analyze patient setup errors in six dimensions, i.e. to analyze in which range setup corrections are applied by means of a 6DoF couch in clinical routine for cranial as well as for extracranial cases in order to demonstrate the need of a 6DoF couch when performing linear accelerator based stereotactic treatments with frameless patient setup systems. Additionally, the patient setup error as a function of the couch rotation angle was analyzed for treatments in the cranial region, where noncoplanar fields are typically used. Attainment rates for two different patient setup accuracy tolerances are calculated.
2 Materials and Methods At our institution, treatments requiring high patient setup accuracy (e.g. stereotactic treatments) are performed using the Novalis TX system which is a combination of a modern linear accelerator from Varian (Varian Medical Systems, Inc., Palo Alto, USA) and the ExacTrac system from Brainlab (BrainLAB AG, Feldkirchen, Germany). ExacTrac is a patient positioning system consisting of an infrared system, an X-ray system and a robotic 6DoF couch [6–8,10–12]. The ExacTrac system was calibrated routinely following the Brainlab standard calibration procedures and using dedicated Brainlab calibration phantoms. The calibration procedure as well as the resulting accuracy of the ExacTrac system is reported in literature [6,8,12]. The verification of the calibration of the
infrared and the X-ray system was done on a weekly and daily frequency, respectively. To assess the range of setup corrections applied by means of a 6DoF couch in clinical routine, the ExacTrac was used as a measurement tool. To understand how and when (during the patient positioning process) the assessment was performed, an overview of the workflow for patient positioning using the ExacTrac in our clinic is described in the following. A flowchart of the workflow from patient setup to treatment when using the ExacTrac of the Novalis TX is given in Fig. 1. In a first step, the patient is positioned on the 6DoF couch. The immobilization of the patients treated in the cranial region was performed with the Brainlab frameless mask system [13,14]. A nose bridge and no mouthpiece were used. For patients treated in the extracranial region the immobilization was achieved with the aid of a vacuum bag cushion (VBC) system (IT-V, Innsbruck, Austria) on which the infrared markers are positioned. The couch is in the zero position during this step, which means that pitch, roll and couch rotation angle are all set to 0◦ . Afterwards, the patient is prepositioned with the aid of the infrared system such that the patient is shifted to treatment position by moving the couch. Note that only the three translational axes of the 6DoF couch are used during this step, i.e. the three rotational axes are inactive at that moment. This step is referred to as initial patient setup in the following. The initial patient setup is then verified by using the x-ray component of ExacTrac: two kV-images are acquired (one image per x-ray tube of the ExacTrac) and matched to the corresponding reference digitally reconstructed radiograph (DRR). This matching is performed automatically by the ExacTrac system using a corresponding matching algorithm [8]. The result of this procedure is the initial patient setup error in six dimensions. The verification of the initial patient setup is referred to as initial verification and the setup error resulting from this initial verification is referred to as initial setup error in this work. If the initial setup error is outside of a specified tolerance (see below) the determined correction is applied with the aid of the 6DoF couch including all six axes. In order to verify if the application of the correction has been performed accurately, two kV-images with ExacTrac have been acquired again. In the next step, all treatment fields for the actual couch rotation angle (which is 0◦ at this moment) are applied. Usually, the treatment plan contains non-coplanar fields when treating volumes in the cranial region. That means that the couch has to be rotated after the treatment of the fields at couch rotation angle 0◦ . Once the couch is rotated, the new patient position is verified with the aid of the x-ray system of ExacTrac. Again, if the resulting setup error is outside of the specified tolerance (see below), the corresponding correction is applied with the aid of the 6DoF couch, and again verified by kV- images using the ExacTrac until the setup error is within the specified tolerance. The whole procedure is repeated for every further couch rotation angle. To determine the clinical range in which a 6DoF couch is used to correct patient setup errors in six dimensions,
D. Schmidhalter et al. / Z. Med. Phys. 24 (2014) 112–122
initial patient setup infrared couch rotation angle 0º
115
couch rotation angle 0º
initial verification x-ray
setup error within tolerance?
no
verification x-ray
apply correction shift
setup error within tolerance?
no
yes
yes
apply treatment field(s)
additional treatment fields available? (rot ≠ 0º)
no
treatment finished
yes rotate couch to next couch rotation angle
couch rotation angle ≠ 0º
verification x-ray
setup error within tolerance?
no
verification x-ray
apply correction shift
no
yes
apply treatment field(s) for the corresponding couch rotation angle
setup error within tolerance?
yes
Figure 1. Flowchart of the workflow from patient setup to treatment when using the Novalis TX in combination with ExacTrac at our institution.
the initial verification step was analyzed in detail. 2185 initial verifications of 242 patients were analyzed in total. Patients undergoing radiosurgery (n = 108) as well as fractionated treatments (n = 134) were taken into account. 1705
initial verifications correspond to patients treated in the cranial region. 480 initial verifications correspond to treatments in the extracranial region. For all patients included in the extracranial subgroup, the target regions were located in that way, that
116
D. Schmidhalter et al. / Z. Med. Phys. 24 (2014) 112–122
Table 1 The table shows the attainment rates for the initial setup error according to tolerance A (±0.5 mm, ±0.5◦ ) and tolerance B (±1.0 mm, ±1.0◦ ). The investigated cases are arranged in three groups: All cases together, cranial cases only and extracranial cases only. Attainment rates are given for each group and for every single couch axis in the columns lat, long, vert, pitch, roll and rot. The attainment rates when all initial setup errors of the six axes have to be within tolerance simultaneously are given in the last column sim.
cranial & extracranial cranial extracranial
tolerance A tolerance B tolerance A tolerance B tolerance A tolerance B
lat [%]
long [%]
vert [%]
pitch [%]
roll [%]
rot [%]
sim [%]
38.3 63.7 43.0 70.8 21.7 38.3
26.4 48.1 28.9 51.9 17.5 34.4
30.8 51.8 36.0 60.1 12.1 22.1
37.1 63.6 34.3 61.4 47.3 71.5
45.0 72.1 45.7 73.4 42.7 67.5
36.5 66.9 36.0 65.8 38.3 70.6
0.5 10.9 0.6 13.1 0.0 3.1
matching on bony anatomy was feasible. No implanted markers were used. Typical examples for treated sites were spine or bone metastases in the pelvic region. Statistical analysis of the initial setup errors was performed by calculating corresponding histograms for each of the six couch axes separately. This analysis was performed for the cranial cases and extracranial cases separately as well as for all 2185 cases together. Two different tolerances were defined. The first allows the translational setup error to be within ±0.5 mm and the rotational setup error to be within ±0.5◦ (tolerance A), i.e. setup errors larger than ±0.5 mm and/or 0.5◦ will be corrected. This tolerance is a conservative lower limit of the accuracy of the ExacTrac system [6–8,10–12]. The second tolerance allows the translational setup error to be within ±1.0 mm and the rotational setup error to be within ±1.0◦ (tolerance B). When talking about radiosurgery and stereotactic radiotherapy, overall geometrical accuracy below 1 mm is often discussed [15–18]. Assuming that patient positioning with ExacTrac is the only error source contributing to this 1 mm error budget, this tolerance B was meant to be an upper limit for patient setup with ExacTrac in clinical routine. At this point, it is important to note that there exist other error sources which have an influence on the overall geometrical accuracy. According to these tolerances A and B, attainment rates for each of the six couch axes were calculated. In clinical routine a correction shift and rotation is applied if the initial setup error of only one of the six axes is outside the tolerance. Thus, the attainment rate was calculated for the situation where the initial setup errors of all the six axes were within tolerance simultaneously, i.e. where no correction has to be applied at all. As described above, the patient is just prepositioned with the aid of the infrared system. For that reason, the initial setup errors are expected to be the largest during the whole patient setup procedure and are suitable for determining the clinical range in which a 6DoF couch is used to correct patient setup errors in six dimensions. In the second part of this work the setup error after rotating the couch was analyzed. Note that at the moment of couch rotation, the patient setup has already been verified and corrected if necessary at couch rotation angle 0◦ . In a perfect world, no additional setup error would be expected when rotating the couch. Nevertheless, different error sources (instability of
the couch rotation center or patient movement) lead to the fact that such additional setup errors occur in clinical routine. For that reason, the setup errors found during patient setup verification after rotating the couch were analyzed as a function of the couch rotation angle. 242 verifications of 54 patients treated in the cranial region were analyzed. This analysis was performed for each of the six couch axes separately. In analogy as described in the previous paragraph, attainment rates for the two tolerances A and B were calculated.
3 Results The histograms showing the initial setup errors for all 2185 initial verifications (cranial and extracranial cases) are plotted in Fig. 2. The histograms for the three translational and the three rotational axes are shown separately. The mean initial setup error (± one standard deviation) for the lat, long and vert axes are 0.11 ± 1.32 mm, 0.21 ± 1.70 mm and 0.00 ± 1.54 mm, respectively. The mean initial setup error (± one standard deviation) for the pitch, roll and rot axes are -0.23 ± 1.16◦ , -0.19 ± 0.95◦ and -0.06 ± 1.06◦ , respectively. The areas in green and orange are visualizing the range where the initial setup error is within tolerance A (±0.5 mm, ±0.5◦ ) and/or tolerance B (±1.0 mm, ±1.0◦ ), respectively. The corresponding attainment rates are quantified in Table 1. The largest attainment rates were found for the roll axis. For this axis, 45.0% and 72.1% of all initial setup errors were within tolerance A and tolerance B, respectively. If the initial setup errors of all six axes have to be within tolerance simultaneously (which is the case in clinical routine), the attainment rate for tolerance A is 0.5%. For tolerance B this attainment rate was found to be 10.9%. Focusing on the two rotational axes pitch and roll only (which make the difference between a conventional 4DoF couch and a 6DoF couch), the data shows that during initial patient setup, in 50% of all cases the tolerance of 1◦ for the pitch and/or roll axis was exceeded. The data obtained during the initial verification was analyzed for cranial and extracranial cases separately. The histograms showing the initial setup errors for the 1705 initial verifications performed in the cranial region are plotted in Fig. 3. The mean initial setup error (± one standard deviation) for the lat, long and vert axes are 0.10±1.17 mm,
D. Schmidhalter et al. / Z. Med. Phys. 24 (2014) 112–122
9.09
117
(a) lateral longitudinal vertical
8.08 7.07
rel. frequency [%]
cranial and extracranial 6.06
tolerance A (± 0.5 mm)
2185 initial verifications 5.05
tolerance B (± 1.0 mm)
4.04 3.03 2.02
4.8
> 5.0
4.4
4.0
3.6
2.8
3.2
2.4
2.0
1.6
0.8
1.2
0.4
0.0
-0.4
-0.8
-1.2
-1.6
-2.0
-2.4
-2.8
-3.2
-3.6
-4.0
-4.4
-5.0-4.8-4.6-4.4-4.2-4.0-3.8-3.6-3.4-3.2-3.0-2.8-2.6-2.4-2.2-2.0-1.8-1.6-1.4-1.2-1.0-0.8-0.6-0.4-0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0 4.2 4.4 4.6 4.8 5.0
-4.8
0.00
> -5.0
1.01
initial setup error [mm]
12 12.0
(b) pitch roll rotation
rel. frequency [%]
10 10.0
cranial and extracranial 8.08
tolerance A (± 0.5º)
2185 initial verifications 6.06
tolerance B (± 1.0º) 4.04
> 4.0
3.6
3.2
2.8
2.4
2.0
1.2
1.6
0.8
0.4
0.0
-0.4
-0.8
-1.2
-1.6
-2.0
-2.4
-2.8
-3.2
< - -4.0 -3.8 -3.6 -3.4 -3.2 -3.0 -2.8 -2.6 -2.4 -2.2 -2.0 -1.8 -1.6 -1.4 -1.2 -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0 > 4.0 4.0
-3.6
0.00
< -4.0
2.02
initial setup error [º]
Figure 2. Histograms of the initial setup errors found during 2185 initial verifications of patients treated in the cranial (n = 1705) as well as in the extracranial (n = 480) region. The data was binned into 0.2 mm and 0.2◦ intervals, respectively. The areas in green and orange are visualizing the range where the initial setup error is within tolerance A (±0.5 mm, ±0.5◦ ) and tolerance B (±1.0 mm, ±1.0◦ ), respectively. (a) Histograms for the translational axes lat, long and vert. (b) Histograms for the rotational axes pitch, roll and rot.
0.10 ± 1.62 mm and 0.11 ± 1.29 mm, respectively. The mean initial setup error (± one standard deviation) for the pitch, roll and rot axes are -0.24 ± 1.21◦ , -0.23 ± 0.91◦ and 0.03 ± 1.07◦ , respectively. The attainment rates for tolerance A and B are quantified in Table 1. The largest attainment rates were found for the roll axis. For this axis, 45.7% and 73.4% of all initial setup errors were within tolerance A and tolerance B, respectively. If the initial setup errors of all six axes have to be within tolerance simultaneously, the attainment rate for tolerance A is 0.6%. For tolerance B an attainment rate of 13.1% was found. The histograms showing the initial setup errors for the 480 initial verifications performed in the extracranial region are plotted in Fig. 4. The mean initial setup errors (± one standard deviation) for the lat, long and vert axes are 0.15 ± 1.77 mm, 0.62 ± 1.94 mm and -0.40 ± 2.15 mm, respectively. The mean initial setup error (± one standard deviation) for the pitch, roll and rot axes are -0.21 ± 0.95◦ , -0.05 ± 1.08◦
and -0.14 ± 1.02◦ , respectively. The attainment rates for tolerance A and B are quantified in Table 1. The largest attainment rates were found for the pitch axis. For this axis 47.3% and 71.5% of all initial setup errors were within tolerance A and tolerance B, respectively. If the initial setup errors of all six axes have to be within tolerance simultaneously, the attainment rate for tolerance A is 0.0%. For tolerance B an attainment rate of 3.1% was found. All the mean initial setup errors are summarized in Table 2. The setup errors as a function of the couch rotation angle are plotted in Fig. 5 for the translational axes and in Fig. 6 for the rotational axes. The setup error per angle is visualized with the aid of a boxplot. For this analysis, the used couch rotation angle was binned into 10◦ -intervals. E.g. setup errors belonging to couch rotation angles in the interval]5◦ ,15◦ ] are visualized by the boxplot illustrated at 10◦ . The areas in green and orange are visualizing the range where the setup error is within tolerance A and/or tolerance B, respectively. The
118
D. Schmidhalter et al. / Z. Med. Phys. 24 (2014) 112–122
11 11.0
(a) 10 10.0
lateral longitudinal vertical
9.09
rel. frequency [%]
8.08
cranial
7.07
1705 initial verifications
6.06
tolerance A (± 0.5 mm)
5.05
tolerance B (± 1.0 mm)
4.04 3.03 2.02
> 5.0
4.8
4.4
4.0
3.2
3.6
2.8
2.4
2.0
1.2
1.6
0.8
0.4
0.0
-0.4
-0.8
-1.2
-1.6
-2.0
-2.4
-2.8
-3.2
-4.0
-3.6
-4.4
-5.0-4.8-4.6-4.4-4.2-4.0-3.8-3.6-3.4-3.2-3.0-2.8-2.6-2.4-2.2-2.0-1.8-1.6-1.4-1.2-1.0-0.8-0.6-0.4-0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0 4.2 4.4 4.6 4.8 5.0
-4.8
0.00
> -5.0
1.01
initial setup error [mm]
12 12.0
(b) pitch roll rotation
10 10.0
rel. frequency [%]
cranial 8.08
tolerance A (± 0.5º)
1705 initial verifications 6.06
tolerance B (± 1.0º) 4.04
3.6
3.2
2.8
2.4
2.0
> 4.0
initial setup error [º]
1.6
1.2
0.8
0.4
0.0
-0.4
-0.8
-1.2
-1.6
-2.0
-2.4
-2.8
-3.2
< - -4.0 -3.8 -3.6 -3.4 -3.2 -3.0 -2.8 -2.6 -2.4 -2.2 -2.0 -1.8 -1.6 -1.4 -1.2 -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0 > 4.0 4.0
-3.6
0.00
< -4.0
2.02
Figure 3. Histograms of the initial setup errors found during 1705 initial verifications of patients treated in the cranial region. The data was binned into 0.2 mm and 0.2◦ intervals, respectively. The areas in green and orange are visualizing the range where the initial setup error is within tolerance A (±0.5 mm, ±0.5◦ ) and tolerance B (±1.0 mm, ±1.0◦ ), respectively. (a) Histograms for the translational axes lat, long and vert. (b) Histograms for the rotational axes pitch, roll and rot. Table 2 The table summarizes the mean initial setup errors found for each of the six axes separately. The mean setup errors are given for the cranial and extracranial cases together as well as for the cranial and extracranial cases separately.
lat [mm] long [mm] vert [mm] pitch [◦ ] roll [◦ ] rot [◦ ]
cranial & extracranial
cranial
extracranial
0.11 0.21 0.00 -0.23 -0.19 -0.06
0.10 0.10 0.11 -0.24 -0.23 -0.03
0.15 0.62 -0.40 -0.21 -0.05 -0.14
corresponding attainment rates are quantified in Tab. 3. The largest attainment rates were found for the roll axis. For this axis, 94.6% and 100.0% of all setup errors after couch rotation were within tolerance A and tolerance B, respectively. If the setup errors of all six axes have to be within tolerance
simultaneously, the attainment rate for tolerance A is 32.6%. For tolerance B an attainment rate of 84.3% was found.
4 Discussion The clinical range in which a 6DoF couch is used to correct patient setup errors in six dimensions was investigated in this work. For this purpose, patient setup errors in six dimensions determined with the aid of ExacTrac in clinical routine were analyzed. The analysis of the patient setup errors was performed at two different points in the workflow: First during the initial patient setup at couch rotation angle 0◦ (where the biggest setup error is expected) and second after deflecting the couch. The histograms in Fig. 2 show that all six axes of the 6DoF couch are used extensively for patient setup in clinical routine. The variance of the distribution of the initial setup errors for the extracranial cases is higher than for the cranial cases,
D. Schmidhalter et al. / Z. Med. Phys. 24 (2014) 112–122
119
Table 3 The table shows the attainment rates for the setup error after couch rotation according to tolerance A (±0.5 mm, ±0.5◦ ) and tolerance B (±1.0 mm, ±1.0◦ ). Attainment rates are given for every single couch axis in the columns lat, long, vert, pitch, roll and rot. The attainment rates when all setup errors of the six axes have to be within tolerance simultaneously are given in the last column sim.
tolerance A tolerance B
lat [%]
long [%]
vert [%]
pitch [%]
roll [%]
rot [%]
sim [%]
64.0 90.5
65.3 95.5
85.5 97.9
91.5 98.8
94.6 100.0
89.3 99.2
32.6 84.3
which is confirmed and quantified by the attainment rates for the different tolerances in section III. One reason for this is the different immobilization strategy for the cranial and extracranial region. Another reason is the different deformability of the anatomy in the target region. E.g. the flex of the spine often differs slightly between planning CT and treatment. Changes in the flex of the spine were minimized with the aid of the VBC system, but a reduction to zero is most often not possible. Another example is the presence of the ribs, which are
positioned differently for each verification due to breathing. Even if most parts of the ribs are removed from the region of interest before performing the matching, the matching is affected by the ribs overlapping with the target region in the corresponding images. Compared to the deformable anatomy in the extracranial region the anatomy in the cranial region is much more rigid. The histograms in Figs. 3 and 4 shows that in clinical routine sometimes large initial setup errors which are even outside the travel range of the 6DoF couch
12 12.0 11 11.0
(a) lateral longitudinal vertical
10 10.0
rel. frequency [%]
9.09 8.08
extracranial
7.07
tolerance A (± 0.5 mm)
480 initial verifications
6.06
tolerance B (± 1.0 mm)
5.05 4.04 3.03 2.02
> 5.0
4.4
4.8
4.0
3.2
3.6
2.8
2.4
2.0
1.2
1.6
0.8
0.0
0.4
-0.4
-0.8
-1.2
-1.6
-2.0
-2.4
-2.8
-3.2
-4.0
-3.6
-4.4
-5.0-4.8-4.6-4.4-4.2-4.0-3.8-3.6-3.4-3.2-3.0-2.8-2.6-2.4-2.2-2.0-1.8-1.6-1.4-1.2-1.0-0.8-0.6-0.4-0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0 4.2 4.4 4.6 4.8 5.0
-4.8
0.00
> -5.0
1.01
initial setup error [mm]
16 16.0
(b) 14 14.0
pitch roll rotation
rel. frequency [%]
12 12.0
extracranial 10 10.0
tolerance A (± 0.5º)
480 initial verifications 8.08
tolerance B (± 1.0º) 6.06 4.04
3.6
3.2
2.8
2.4
2.0
1.6
1.2
0.8
0.4
0.0
-0.4
-0.8
-1.2
-1.6
-2.0
-2.4
-2.8
-3.2
-3.6
< - -4.0 -3.8 -3.6 -3.4 -3.2 -3.0 -2.8 -2.6 -2.4 -2.2 -2.0 -1.8 -1.6 -1.4 -1.2 -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0 > 4.0 4.0
< -4.0
0.00
> 4.0
2.02
initial setup error [º]
Figure 4. Histograms of the initial setup errors found during 480 initial verifications of patients treated in the extracranial region. The data was binned into 0.2 mm and 0.2◦ intervals, respectively. The areas in green and orange are visualizing the range where the initial setup error is within tolerance A (±0.5 mm, ±0.5◦ ) and tolerance B (±1.0 mm, ±1.0◦ ), respectively. (a) Histograms for the translational axes lat, long and vert. (b) Histograms for the rotational axes pitch, roll and rot.
120
D. Schmidhalter et al. / Z. Med. Phys. 24 (2014) 112–122
2.0
lateral 2.0
pitch
242 verifications
242 verifications 1.0 setup error [º]
setup error [mm]
1.0
0.0
0.0
-1.0 -1.0
tolerance A (± 0.5 mm) -2.0
tolerance A (± 0.5º)
tolerance B (± 1.0 mm)
tolerance B (± 1.0º) 40.0
50.0
60.0
70.0
80.0
90.0
40.0
50.0
60.0
70.0
80.0
90.0
40.0
50.0
60.0
70.0
80.0
90.0
20.0
30.0
10.0
-20.0
-10.0
couch rotation angle [º]
couch rotation angle [º]
2.0
longitudinal 2.0
-30.0
-50.0
-40.0
-60.0
-80.0
-70.0
-90.0
90.0
80.0
70.0
60.0
50.0
40.0
20.0
30.0
10.0
-20.0
-10.0
-30.0
-50.0
-40.0
-60.0
-80.0
-70.0
-90.0
-2.0
roll
242 verifications
242 verifications 1.0 setup error [º]
setup error [mm]
1.0
0.0
-1.0
0.0
-1.0
-2.0
tolerance A (± 0.5 mm)
tolerance A (± 0.5º)
tolerance B (± 1.0 mm)
tolerance B (± 1.0º) 30.0
20.0
10.0
-10.0
-20.0
couch rotation angle [º]
couch rotation angle [º]
2.0
vertical 2.0
-30.0
-40.0
-50.0
-60.0
-70.0
-80.0
-90.0
90.0
80.0
70.0
60.0
50.0
40.0
30.0
20.0
10.0
-10.0
-20.0
-30.0
-40.0
-50.0
-60.0
-70.0
-80.0
-90.0
-2.0
rotation 242 verifications
242 verifications 1.0 setup error [º]
setup error [mm]
1.0
0.0
0.0
-1.0 -1.0
tolerance A (± 0.5 mm) -2.0
tolerance A (± 0.5º)
tolerance B (± 1.0 mm)
tolerance B (± 1.0º)
couch rotation angle [º]
Figure 5. Setup errors as a function of the couch rotation angle for the three translational axes lat, long and vert. The setup error as a function of the couch rotation angle is visualized with the aid of a boxplot. Therefore, the couch rotation angle was binned into 10◦ intervals. E.g. setup errors belonging to couch rotation angles in the interval]5◦ ,15◦ ] are visualized by the boxplot illustrated at 10◦ . The areas in green and orange are visualizing the range where the setup error is within tolerance A (±0.5 mm, ±0.5◦ ) and/or tolerance B (±1.0 mm, ±1.0◦ ).
30.0
20.0
10.0
-10.0
-20.0
-30.0
-40.0
-50.0
-60.0
-70.0
-80.0
-90.0
90.0
80.0
70.0
60.0
50.0
40.0
30.0
20.0
10.0
-10.0
-30.0
-20.0
-40.0
-50.0
-60.0
-70.0
-80.0
-90.0
-2.0
couch rotation angle [º]
Figure 6. Setup errors as a function of the couch rotation angle for the three rotational axes pitch, roll and rot. The setup error as a function of the couch rotation angle is visualized with the aid of a boxplot. Therefore, the couch rotation angle was binned into 10◦ intervals. E.g. setup errors belonging to couch rotation angles in the interval]5◦ ,15◦ ] are visualized by the boxplot illustrated at 10◦ . The areas in green and orange are visualizing the range where the setup error is within tolerance A (±0.5 mm, ±0.5◦ ) and/or tolerance B (±1.0 mm, ±1.0◦ ).
D. Schmidhalter et al. / Z. Med. Phys. 24 (2014) 112–122
(e.g. for the pitch and/or roll axes) do occur in the extracranial as well as in the cranial region. Even if such large initial patient setup errors could be corrected with the aid of the 6DoF couch it is not a priori useful, especially not in the extracranial region. Large pitch and roll corrections could be noticed by the patient and the patient could try to compensate the correction rotation. Additionally, the anatomy of the patient may change in comparison with the reference CT if the pitch and/or roll angle differs too much from the situation during the acquisition of the reference CT. It is not possible to compensate such deformations in the patient anatomy with the aid of the 6DoF couch. In such cases, a possible approach is to restart the whole patient setup procedure from the beginning by taking the patient down from the couch. Large initial setup errors outside the travel range of the 6DoF couch in the cranial region do occur mainly during fractionated treatments where the patient is losing weight during the therapy which results in the fact that the mask is losing its tightness during treatment. In such a case, one possibility is to reduce the thickness of the spacers between the front and back part of the mask in order to tighten it. Nevertheless, in extreme cases a new mask has to be produced which means that a re-planning of the treatment has to be performed. When treating brain metastases, Gevaert et al. [5] reported mean translational setup errors (± one standard deviation) in the long, lat and vert axes of 0.41 ± 1.19 mm, -0.48 ± 1.58 mm and 0.06 ± 0.99 mm, respectively and mean rotational setup errors (± one standard deviation) in the pitch, roll and rot axes of -0.09 ± 0.72◦ , 0.23 ± 0.82◦ and -0.10 ± 1.03◦ , respectively. Beltran et al. [9] reported mean rotational errors (± on standard deviation) in the pitch roll and rot axes of 0.37 ± 1.04◦ , 0.29 ± 0.96◦ and 0.03 ± 0.99◦ , respectively for children with brain tumors treated in the supine position. These results are in concordance with the results found in our study. The patient setup error was analyzed as a function of the couch rotation angle. Before the couch is rotated, the patient setup was corrected at couch rotation angle 0◦ (initial patient setup correction). Our data shows that patient setup errors outside tolerance A (±0.5 mm, ±0.5◦ ) as well as outside tolerance B (±1.0 mm, ±1.0◦ ) occur after a couch rotation. There are mainly two error sources which lead to these results. The first error source is a discrepancy (within specified tolerances) between the treatment isocenter (which is represented by the ExacTrac isocenter) and the couch rotation center. The discrepancy between these two isocenters for our system was determined with a Winston-Lutz test and the results are in concordance with the data found in this work. The second reason is the patient itself who may move (intrafraction motion, e.g. due to the couch movement itself). E.g., Ramakrishna et al. [19] observed a mean (± standard deviation) three-dimensional intrafraction shift magnitude for the Brainlab frameless mask system of 0.7 ± 0.5 mm. Gevaert et al. [5] reported a mean (± standard deviation) three-dimensional intrafraction shift for the same frameless system of 0.58 ± 0.42 mm. Inaccuracies
121
in the system calibration would result in systematic setup errors as a function of the couch rotation error. The patient movement due to couch rotation would most probably result in a random error. Our data clearly shows these two types of errors. The inaccuracy in system calibration is confirmed by the plots in Fig. 5 and Fig. 6 where the setup error as a function of the couch rotation angle is shown. The axis of rotation when rotating the couch (change the couch rotation angle rot) is perpendicular to the lateral-longitudinal plane. For that reason, setup errors due to couch rotation are expected to be maximal for the lat and long axes and can be described by a periodic function. For geometrical reasons, the effect of inaccuracies in the system calibration due to couch rotation (systematic setup error) is negligible for the four other axes lat, long, vert, and rot. This is confirmed by Figs. 5 and 6. The median values of the setup errors per couch rotation angle are describing this systematic error. The boxplots are describing the random errors per couch rotation angle. Nevertheless, the systematic setup error resulting from miscalibration of the couch rotation center is detectable with the aid of ExacTrac, which is a room-based system. By following the workflow as described in the Materials and Methods section, that means by verifying the patient position always directly after the couch was shifted or rotated and before treatment, the setup error due to miscalibration of the couch will be detected and compensated with the aid of the 6DoF couch. The main limitation of a 4DoF couch in comparison to a 6DoF couch is that it is not possible to correct pitch and roll setup errors. To estimate the dosimetric consequences of pitch and/or roll rotational setup errors in general is difficult because the dosimetric consequences due to rotational setup are influenced by a lot of factors, e.g. the size and form of the PTV, the number of PTVs, the position of the organs at risks in relation to the PTV, the field arrangement, or the margin between the clinical target volume and the PTV. Several groups investigated these dosimetric consequences for specific cases. Beltran et. al [9] investigated the effect of these rotational errors for pediatric brain tumor patients and found undesirable changes in the gEUD for critical structures and planning target volume (PTV) for rotational setup errors larger than 2◦ . Gevaert et al. [5] investigated the dosimetric consequences for patients with brain metastases and found that the mean (± one standard deviation) conformity index decreased from 0.68 ± 0.08 (6DoF) to 0.59 ± 0.12 (4DoF). A loss of prescribed isodose coverage of 5% was found for 4DoF couch positioning. Half a degree for pitch and roll rotations was identified as a threshold for coverage loss. Guckenberger et al. [20] simulated the dosimetric improvements when using IGRT (including all six axes) for radiosurgery of brain metastases in comparison to non-IGRT radiosurgery.
5 Conclusion In this work we analyzed the range in which a 6DoF couch is used to correct patient setup errors in six dimensions in
122
D. Schmidhalter et al. / Z. Med. Phys. 24 (2014) 112–122
clinical routine. For that reason, initial patient setup errors were analyzed. Additionally, the patient setup error as a function of the couch rotation error was investigated. The analysis of the data shows that all six axes are extensively used for patient setup in clinical routine. In order to fulfill high patient setup accuracies (e.g. for stereotactic treatments), a 6DoF couch is recommended. We have shown that the high patient setup accuracy (realized at couch rotation angle 0◦ ) may be decreased when rotating the couch. That is why re-verification (including repositioning until the setup error is within tolerance) of the patient setup after rotating the couch is required in clinical routine.
References [1] Lamba M, Breneman JC, Warnick RE. Evaluation of image guided positioning for frameless intracranial radiosurgery. Int J Radiat Oncol Biol Phys 2009;74:913–9. [2] Verellen D, De Ridder M, Linthout N, Tournel K, Soete G, Storme G. Innovations in image-guided radiotherapy. Nat Rev Cancer 2007;7:949–60. [3] Kamath R, Ryken TC, Meeks SL, Pennington EC, Ritchie J, Buatti JM. Initial clinical experience with frameless radiosurgery for patients with intracranial metastases. Int J Radiat Oncol Biol Phys 2005;61:1467–72. [4] Breneman JC, Steinmetz R, Smith A, Lamba M, Warnick RE. Frameless image- guided intracranial stereotactic radiosurgery; clinical outcomes for brain metastases. Int J Radiat Oncol Biol Phys 2009;74:702–6. [5] Gevaert T, Verellen D, Engels B, Depuydt T, Heuninckx K, Duchateau M, et al. Clinical Evaluation of a Robotic 6-Degree of Freedom Treatment Couch for Frameless Radiosurgery. Int J Radiat Oncol Biol Phys 2012;83:467–74. [6] Jin JY, Yin FF, Tenn SE, Medin PM, Solberg TD. Use of the BrainLAB ExacTrac X- Ray 6D System in Image-Guided Radiotherapy. Med Dosim 2008;33:124–34. [7] Takakura T, Mizowaki T, Nakata M, Yano S, Fujimoto T, Miyabe Y, et al. The geometric accuracy of frameless stereotactic radiotherapy using a 6D robotic couch system. Phys Med Biol 2010;55:1–10. [8] Gevaert T, Verellen D, Tournel K, Linthout N, Bral S, Engels B, et al. Setup accuracy of the Novalis ExacTrac 6DOF system for frameless radiosurgery. Int J Radiat Oncol Biol Phys 2012;82:1627–35.
[9] Beltran C, Pegram A, Merchant TE. Dosimetric consequences of rotational errors in radiation therapy of pediatric brain tumors. Radiother Oncol 2012;102:206–9. [10] Schewe JE, Lam KL, Balter JM, Haken RKT. A room-based diagnostic imaging system for measurement of patient setup. Med Phys 1998;25:2385–8. [11] Kim J, Wen N, Jin JY, Walls N, Kim S, Li H, et al. Clinical commissioning and use of the Novalis TX linear accelerator for SRS and SBRT. J Appl Clin Med Phys 2012;13:124–51. [12] Verellen D, Soete G, Linthout N, Van Acker S, De Roover P, Vinh-Hung V, et al. Quality assurance of a system for improved target localization and patient set-up that combines real-time infrared tracking and stereoscopic X-ray imaging. Radiother Oncol 2003;67: 129–41. [13] Ali I, Tubbs J, Hibbitts K, Algan O, Thompson S, Herman T, et al. Evaluation of the setup accuracy of stereotactic head immobilization mask system using kV on- board imaging. J Appl Clin Med Phys 2010;10:3192. [14] Lightstone AW, Benedict AH, Bova FJ, Solberg TD, Stern RL. Intracranial stereotactic positioning systems: Report of the American Association of Physicists in Medicine Radiation Therapy Committee Task Group No. 68. Med Phys 2005;32:2380–98. [15] Bichay T, Dieterich S. Point/Counterpoint: Submillimeter accuracy in radiosurgery is not possible. Med Phys 2013;40:050601. [16] Antypas C, Pantelis E. Performance evaluation of a Cyberknife G4 image-guided robotic stereotactic radiosurgery system. Phys Med Biol 2008;53:4697–718. [17] Wang L, Kielar KN, Mok E, Hsu A, Dieterich S, Xing L. An end-toend examination of geometric accuracy of IGRT using a new digital accelerator equipped with onboard imaging system. Phys Med Biol 2012;57:757–69. [18] Ruschin M, Komljenovic PT, Ansell S, Ménard C, Bootsma G, Cho YB, et al. Cone beam computed tomography image guidance system for a dedicated intracranial radiosurgery treatment unit. Int J Radiat Oncol Biol Phys 2013;85:243–50. [19] Ramakrishna N, Rosca Florin, Friesen S, Tezcanli E, Zygmanszki P, Hacker F. A clinical comparison of patient setup and intra-fraction motion using frame-based radiosurgery versus a frameless imageguided radiosurgery system for intracranial lesions. Radiother Oncol 2010;95:109–15. [20] Guckenberger M, Roesch J, Baier K, Sweeney RA, Flentje M. Dosimetric consequences of translational and rotational errors in frame-less image-guided radiosurgery. Radiat Oncol 2012;7:63–78.
Available online at www.sciencedirect.com
ScienceDirect
