Original article Strahlenther Onkol 2013 DOI 10.1007/s00066-013-0441-z Received: 11 July 2013 Accepted: 31 July 2013 © Springer-Verlag Berlin Heidelberg 2013
F. Stieler · F. Wenz · M. Shi · F. Lohr Department of Radiation Oncology, University Medical Center Mannheim, University of Heidelberg, Mannheim
A novel surface imaging system for patient positioning and surveillance during radiotherapy A phantom study and clinical evaluation
Image-guided radiotherapy (IGRT) reduces setup errors and thus minimizes the margin between clinical target volume (CTV) and planning target volume (PTV). Two-dimensional megavoltage imaging with the therapy beam enables matching/positioning relative to bony structures only [1]. Cone-beam computed tomography (CBCT) has been widely adopted and provides the most accurate patient positioning with a relatively low extra imaging dose to the patient [1, 2]. A remaining positioning issue is target motion during dose delivery in the treatment of lung and liver metastases. Multiple strategies have been developed to compensate for this intrafractional tumor motion [3]. An alternative positioning strategy is based on surface tracking. The current surface scan is compared to the reference surface (based on planning CT) and a shift vector is calculated [4, 5, 6, 7]. These systems may reduce the number of CBCT scans and thus limit the imaging dose to patients. The system described in this study uses a new scanning method with a near-visible light projector and a charge-coupled device (CCD) camera. It projects the calculated regional patient shift directly onto the patient’s surface in order to simplify the patient positioning process. It also provides a surveillance function to detect patient movement or breathing during treatment (intrafractional movement); a functional modality that can also be used to drive the gating interface of a linear accelerator.
The surveillance function, the new scanning approach and gating may further improve the accuracy of liver and lung treatments [3], provided that the inherent accuracy of the system is sufficient. As a first step, we investigated the basic performance and accuracy of the new scanning method of the Catalyst (C-RAD, Uppsala, Sweden) system in a non-gated environment. These issues were addressed in both phantom experiments mimicking different clinical situations and in a prospective clinical study covering three anatomical regions.
Materials and methods Phantom and clinical studies were performed on an Elekta Synergy (Elekta AB, Stockholm, Sweden) accelerator with CBCT. The Catalyst optical system is mounted to the ceiling above the foot end of the treatment table (. Fig. 1). Instead of using laser light to scan the surfaces,
Catalyst employs three high-power LEDs to project light with wavelengths of 405 (blue), 528 (green) and 624 nm (red) onto the object. The blue component is the measuring light for object scanning and is detected by a monochrome CCD camera with an acquisition speed of 202 frames per second. The green and red lights project surface mismatches (actual vs. reference scan) onto the area where the mismatch is detected to aid patient positioning. Two custom settings, gain and integration time (IT), inside the Catalyst software can influence scan quality. The gain is the quantity of captured electrons required on a pixel of the CCD camera to convert light into electronic charge and a digital readout. IT defines the time of light absorption. The maximal scan volume is 80 cm wide, 130 cm long and 70 cm high. An individual region of interest related to the paradigm can be defined. The phantom part of the study analyzed scanning quality, reproducibility
Fig. 1 8 Optical system Catalyst (left) and its setup in linear accelerator room (right) Strahlentherapie und Onkologie 2013
| 1
Original article
Fig. 2 8 a Color test setup, b shape test setup
Pink–white
Pink–1
Pink–2
Pink–3
Gain (%)
600 500 400 300 200 100
1000 2000 3000 4000 5000 6000 7000 1000 2000 3000 4000 5000 6000 7000 1000 2000 3000 4000 5000 6000 7000 1000 2000 3000 4000 5000 6000 7000
Integration time (μs)
Non-area detected Half-area detected Full-area detected
Gray–white
Gray–1
Gray–2
Gray–3
Gain (%)
600 500 400 300 200 100
1000 2000 3000 4000 5000 6000 7000 1000 2000 3000 4000 5000 6000 7000 1000 2000 3000 4000 5000 6000 7000 1000 2000 3000 4000 5000 6000 7000
Integration time (μs) Fig. 3 8 Plots of the color tests depending on gain and integration time. Scans with settings that result in values in the blue area are not visible; scans with results in the red area are fully visible; green indicates half-visible scans
and therefore the inherent accuracy of the system in a static, anthropomorphic situation. The influence of body shape and the color of the scanned surface on these parameters was measured and revealed that scan quality depends on the surface color and the surface shape. We tested surface detection with series of pink and gray surfaces (. Fig. 2a) and by adjusting the gain (range 100–600%) and IT (range 1000– 7000 μs) parameters. The shape test comprised four different concave and convex objects (cylinder, cube, wave and trape zoid cylinder) representing potentially problematic parts of the patient geometry
2 |
Strahlentherapie und Onkologie 2013
(chin/neck transition, different thorax geometries etc.). These objects were scanned with the system operating with different scanning parameters (. Fig. 2b). For the reproducibility test, w e scanned an object every 15 s, 11 times and registered the differences. For the accuracy test, an isocentrically positioned phantom was shifted 95 times in well-defined directions. These manually applied shifts were compared to the measured shifts of the optical system. After obtaining institutional review board (IRB) approval and informed patient consent, a total of 224 radiothera-
py fractions in 3 patients being treated for head and neck cancer, 5 patients treated for thoracic tumors and 5 patients with pelvic tumors were selected for the clinical evaluation. All patients were placed in the supine position and had no fiducial markers. The workflow was as follows: the planning CT dataset was sent to the treatment planning system (Monaco 3.2; Elekta AB); CTVs, PTVs and organs at risk were contoured and a treatment plan was created. Afterwards, the CT images, structure set and treatment plan were sent to a recordand-verify system (Mosaiq 2.41; Elekta AB) and the Catalyst optical scanner. To
Abstract · Zusammenfassung Strahlenther Onkol 2013 · [jvn]:[afp]–[alp] DOI 10.1007/s00066-013-0441-z © Springer-Verlag Berlin Heidelberg 2013 F. Stieler · F. Wenz · M. Shi · F. Lohr
A novel surface imaging system for patient positioning and surveillance during radiotherapy. A phantom study and clinical evaluation Abstract Background. The use of optical surface positioning to support or replace X-ray-based image-guided radiotherapy (IGRT) may reduce patient exposure to extra dose. In specifically designed phantom tests, we analyzed the potential of a new scanning device preclinically. The system’s clinical performance was evaluated in comparison to cone-beam computed tomography (CBCT) in a prospective study. Materials and methods. We first evaluated the scanning performance in terms of accuracy and reproducibility using phantom tests. An institutional review board (IRB)-approved clinical evaluation encompassing 224 fractions in 13 patients treated in three different regions (head and neck, thorax, pelvis)
was then performed. Patients were first positioned using CBCT and then scanned with the CatalystTM (C-RAD, Uppsala, Sweden) optical system to define the resulting difference vector. Results. Individual system settings are necessary for different scanning conditions. Reproducibility tests with phantoms showed a mean difference of 0.25±0.21 cm. Accuracy tests showed a mean difference of less than 0.52±0.41 cm. Considering all patients, clinical data showed residual target position differences between CatalystTM (surface-driven) and CBCT (target-driven) systems within 0.07±0.28 cm/−0.13±0.40 cm/0.15±0.36 cm/
0.11±1.57°/−0.43±1.68/−0.10±1.67° (lateral/ longitudinal/vertical/rotation/roll/pitch). Conclusion. Scanning quality depends on the color and shape of the scanned surface. Upon prospective clinical evaluation, excellent agreement between target- and contour driven positioning was observed. CatalystTM may reduce CBCT scan frequency in patients where tumor location is fixed relative to the surface. Keywords Cone-bean computed tomography · Imageguided radiotherapy · Head and neck cancer · Linear accelerator · Target volume
Neues oberflächenbasiertes Bildsystem zur Positionierung und Kontrolle der Patienten während der Strahlentherapie. Phantomstudie und klinische Bewertung Zusammenfassung Hintergrund. Die optische Oberflächenpositionierung zur Unterstützung oder zum Ersatz von röntgenstrahlenbasierender IGRT kann die Strahlenbelastung des Patienten reduzieren. In speziellen Phantomtest wurde das Potential eines neuen Systems untersucht und das klinische Potential im Vergleich zur ConeBeam-Computertomographie (CBCT) in einer prospektiven Studie evaluiert. Materialien und Methoden. Wir evaluierten das Potential des Systems bezüglich Genauigkeit und Reproduzierbarkeit in Phantomtests und analysierten eine durch die IRB geneh migte Studie, welche 224 Fraktionen aus 13 Patienten in 3 unterschiedlichen Regionen umfasste (Kopf-Hals, Thorax und Abdomen). Die Patienten wurden zuerst mit CBCT posi-
compare patient positioning based on surface matching to patient positioning based on CBCT matching, the following procedure was adhered to: patients were positioned initially with the room lasers aided by the mismatch projection function of the Catalyst system. Patients were then scanned with the CBCT system and shifted to the planned position based on the location of the target volume in the planning CT. Afterwards, a Catalyst scan was performed and the resulting difference vector automatically derived by the sys-
tioniert und anschließend mit dem optischen System CatalystTM (C-RAD, Uppsala, Schweden) gescannt, um den Unterschied zu ermitteln. Ergebnisse. Individuelle Systemeinstellungen sind für unterschiedliche Abtastbedingungen notwendig. Die Reprodu zierbarkeitstests anhand Phantomen zeigten eine mittlere Abweichung von 0,25±0,21 cm. Genauigkeitsanalysen ergaben eine mittlere Abweichung von weniger als 0,52±0,41 cm. Die klinischen Ergebnisse wiesen eine Abweichung von CatalystTM (oberflächenbasiert) zu CBCT (zielvolumenbasiert) über alle Patienten von 0,07±0,28 cm/−0,13±0,40 cm/0,15±0,36 cm /0,11±1,57°/−0,43±1,68/−0,10±1,67° (late
tem in three translational directions (lateral, longitudinal and vertical) and three rotational axes (rotation = yaw, roll and pitch) was recorded. An ideal match between surface scan and soft tissue matching results in zero shift vectors. We recorded the shifts to calculate mean values and standard deviations.
Results The phantom tests quantified positioning accuracy and reproducibility depend-
ral/longitudinal/vertikal/Rotation/Rollen/Kippen) auf. Schlussfolgerung. Die Abtastqualität hängt von Farbe und Kontur der Oberfläche ab. Die Auswertung der klinischen Studie zeigte hervorragende Übereinstimmung zwischen zielvolumen- und oberflächenbasierter Positio nierung. CatalystTM ermöglicht eine Reduzie rung der CBCT-Anzahl bei Patienten mit fester Tumoroberflächenrelation. Schlüsselwörter Cone-Beam-Computertomographie · Bildgesteuerte Strahlentherapie · Kopf-HalsKarzinom · Linearbeschleuniger · Zielvolumen
ing on object shape and color, respectively. The detection quality for different colors depending on the choice of the customizable settings is displayed in . Fig. 3. Classification into the non-, half- and fully-detected surface was performed by a human observer. For the pink-white and pink-1 colors, the surface is not fully detectable when IT and gain are beyond certain thresholds (as indicated by the green halfvisible area). Overexposure of the surface is then the result. For pink-2, underexposure starts at very low IT and gain values, Strahlentherapie und Onkologie 2013
| 3
Original article
Horizontal plane
Vertical plane
Face to catalyst Good resolution
Test shapes
Resolution 1
Resolution 2
Resolution 3
a Vertical plane
Gain (%)
600
Horizontal plane
600
Face to catalyst
600
500
500
500
400
400
400
300
300
300
200
200
200
100 100 100 1000 2000 3000 4000 5000 6000 7000 1000 2000 3000 4000 5000 6000 7000 1000 2000 3000 4000 5000 6000 7000 Integration time (μs) non-area detected half-area detected full-area detected
b
Lat-Long (mm)
Long-Vert (mm)
10
10 Vert
10
Vert
20
Long
20
0
–10
–10 –20 –10
0 Lat
10
20
–20 –20
Manually applied shift
but overexposure is also possible for high IT and gain settings. For pink-3 the surface is only fully detectable at high IT and gain; if IT and gain are too low, underexStrahlentherapie und Onkologie 2013
0
–10
–20 –20
4 |
Lat-Vert (mm)
20
0
–10
0 Long
Measured shift
10
Fig. 4 9 a Setup and different resolutions depending on gain and integration time (good resolution: gain 200%, IT 2000 μs; resolution 1: gain 200%, IT 4000 μs; resolution 2: gain 200%, IT 6000 μs; resolution 3: gain 200%, IT 8000 μs). b Plots of the surface shape test depending on integration time and gain. Red fully visible, blue not visible, green half visible
20
–20
Manually applied shift
–10
0 Lat
10
Measured shift
posure occurs and the surface is not visible. The gray colors behave similarly. The darker the surface is, the higher the IT and gain settings have to be.
20
Fig. 5 9 Defined shifts in three dimensions were manually applied to the respective phantom and then measured with Catalyst. The figure graphically depicts the comparison of the manually applied vs. measured shifts. Vert vertical, Lat lateral, Long longitudinal
Based on these semiquantitative diagrams it becomes evident that although scan quality depends on color, it can be improved by manual adjustment of the
Fig. 6 9 Shift vector in cm or degree, as appropriate, for three translation directions (lateral, longitudinal and vertical) and three rotational directions (rotation, roll and pitch) for the 224 fractions. Left section: head and neck targets, middle section: pelvic targets and right section: thoracic targets. Sections are divided by dashed lines
optical settings (gain and IT). For extremely dark surfaces (e.g. gray-3) there is almost no detection because not enough light is reflected by the objects. A mean IT of 4000 μs and 300% gain seams to provide good results over a wide spectrum of scanning situations. Similar results were obtained for the surface shape tests (. Fig. 4a, b). Scan quality depends on surface orientation and the optical settings. A vertical surface and high IT and gain settings result in overexposure; a horizontal surface allows longer IT and higher gain with acceptable detection. System settings therefore have to be customized for the different standard target regions (head and neck/thorax/pelvis etc.). The reproducibility test showed good stability with the following mean variations over all test shifts in lateral, longitudinal and vertical directions: 0.90±0.32, −0.11±0.23 and −0.26±0.09 mm, respectively. The rotational deviations (rotation, roll and pitch) were: 0.12±0.08°, −0.19±0.10° and −0.01±0.05°, respectively. The accuracy test showed the following mean deviations between manually applied shifts and the measured shifts: in the lateral direction 0.52±0.41 mm, in the longitudinal direction 0.44±0.36 mm and in
the vertical direction 0.26±0.19 mm. The distributions between applied and evaluated shifts are shown in . Fig. 5. The clinical study showed good agreement between Catalyst and CBCT. Mean values ± standard deviations of the resulting shifts (translations and rotations) associated with the 224 recorded fractions are shown in . Fig. 6. No recorded Catalyst shifts exceeded a 1 cm deviation from CBCT measurements. As shown in . Tab. 1, longitudinal and vertical shifts were larger than lateral shifts due to surface changes caused by patient breathing. The largest longitudinal/vertical shifts were observed for patients with pelvic targets due to abdominal respiration in supine position. Rotational shifts had a mean value around 0 and stayed within 4°. The only rotational shift suggested by Catalyst larger than 6° was found in a head and neck cancer patient with misaligned shoulders. Analyzing the CBCT with respect to target and overall surface, there was no clinical need to reposition the patient. The absence of fractions with shifts outside the tolerances and mean shift values that are always close to zero show that a systematic error does not exist.
Mean shifts including standard deviations, separated into scanned regions and as recorded across all patients are shown in . Tab. 1. The overall mean deviation between CBCT and Catalyst measurements stayed within 1.5 mm and 0.4°. The standard deviations were within 4 mm and 1.7°.
Discussion Optical scanning systems are based on surface scanning with laser light or speckle projectors. Palotta et al. [4] compared the Sentinel (C-Rad, Uppsala, Sweden) system to CBCT and portal imaging based on measurements with a rigid phantom. A second study from this group was based on real clinical patients and analyzed thoracic and pelvic target regions separately with a total of 192 fractions [8]. These authors found mean deviations between patient positions measured by CBCT and Sentinel of less than 0.37 cm and 2.1°. This system was also evaluated (under its rebranded name Galaxy; LAP Laser, Lüneburg, Germany) in comparison to megavoltage CT (MVCT) in a TomoTherapy (Tomotherapy, Madison, WI, USA) unit in a study based on 200 treatment fractions [7]. This group found a systematStrahlentherapie und Onkologie 2013
| 5
Original article Tab. 1 Mean shift vector of Catalyst after cone-beam computed tomography and associated
standard deviation (in cm or degree, as appropriate) Direction/axis Latitudinal (cm) Longitudinal (cm) Vertical (cm) Rotation (degree) Roll (degree) Pitch (degree)
Mean shift vector Head & neck 0.03±0.24 −0.37±0.34 −0.02±0.34 −0.5±1.8 −0.9±1.5 −0.5±1.7
Pelvis 0.12±0.25 0.02±0.37 0.18±0.37 0.0±0.7 0.1±1.3 0.5±1.2
ic shift of 3–9 mm, which they suggested to be due to the single camera setup and patient breathing. We had previously also evaluated the same system [5] with similar results (mean deviation less than of 0.51 cm and 2.1°). The system analyzed in the present study shows slightly better results; not for the mean values but for the standard deviations, which means that the statistical error of this system/case study is smaller than the previous laser-based version/study. This most likely results from the improved scanning quality using LED light, the higher resolution of the tracking camera and an improved surface matching algorithm. Gopan et al. [6] performed a clinical head and neck study with the AlignRT (Vision RT Ltd, London, UK) system. They found a translational error of 2.4–4.5 mm and a rotational error of 0.8–2.2°, thus indicating somewhat worse agreement than our study. Kauweloa et al. used the GateCT tracking system (VisionRT, London, UK) for respiratory signal reconstruction in four-dimensional CT imaging. This group could show that the system allowed consistent temporal/ phase tracking [9]. The time aspect is also an important issue in radiotherapy treatment. The precision of accurate target positioning may be lost due to patient/target movement if treatment duration is too long. Modern irradiation techniques such as volumetric modulated arc therapy (VMAT) [10, 11], faster multileaf collimators (MLC) or flattening filter free (FFF) linear accelerators were introduced into the clinical environment to accelerate the patient treatment. A standard CBCT procedure takes 3–5 min including scanning and matching. The optical system introduced in this study needs seconds to scan the patient and calculate the corresponding shift vec-
6 |
Strahlentherapie und Onkologie 2013
Thorax 0.07±0.33 −0.07±0.38 0.24±0.32 0.7±1.5 −0.6±1.9 −0.3±1.8
Overall 0.07±0.28 −0.13±0.40 0.15±0.36 0.1±1.5 −0.4±1.7 −0.1±1.7
tor, thereby potentially reducing the time required for image guidance. Several clinical aspects may influence scanning results. The exclusive use of scanning systems must therefore be controlled carefully. Patients may lose weight during radiation therapy [12, 13]. This dramatically changes the patient surface, thus the changed target-to-surface distances may lead to inaccurate patient positioning. During our clinical evaluation, 1 patient unexpectedly lost a large amount of weight and the matching deviation increased from 4.8 to 13.4 mm in the vertical direction, prompting replanning. Therefore, combining surface scanning and CBCT may provide accurate patient positioning at a low imaging dose. Portal imaging instead of CBCT is not adequate, due to limited soft tissue contrast unless fiducial markers are implanted. A second clinical aspect is patient breathing during positioning and treatment, particularly for lung/liver lesions. For these patients special treatment techniques (breath holding and gating) were introduced [3, 14]. In order to use optical scanners to survey breathing motion, image quality and scan accuracy must be optimal. To date, the matching algorithm of the described system uses rigid matching of the surfaces. An elastic matching algorithm under development will likely further improve the already good accuracy of the matching procedure. For thoracic targets (. Tab. 1), the mean deviation in the vertical direction is already good but may be further improved through elastic matching. This, in turn, might enable reduction of the margin associated with random and systematic geometrical deviations in target doses [15]. The supplementary use of surface scanners to reduce
CBCT scan frequency for patients where imaging dose should be minimized, such as children, improves positioning compared to no IGRT and patient surveillance may reduce intrafractional positioning uncertainties.
Conclusion This study analyzed the basic scanning parameters of the Catalyst surface imaging system, including scan quality, accuracy and reproducibility. Furthermore, it assesses the system’s clinical applicability and accuracy. These analyses showed good agreements for both basic tests and the clinical evaluation; the possible reasons for deviations are discussed. In a controlled environment, the introduction of the optical scanner allows a reduction in periodical CBCT scan frequency. The surveillance function minimizes intrafractional positioning errors and allows monitoring of moving tumors for gating.
Corresponding address F. Stieler, Ph.D. Department of Radiation Oncology, University Medical Center Mannheim, University of Heidelberg Theodor-Kutzer-Ufer 1–3, 68167 Mannheim Germany
[email protected]
Financial support. This work was supported within the framework of a Research Cooperation Agreement between the Department of Radiation Oncology, Mannheim University Medical Center and C-Rad (Sweden).
Compliance with ethical guidelines Conflict of interest. F. Stieler, F. Wenz, M. Shi and F. Lohr state that there are no conflicts of interest. All studies on humans described in the present manuscript were carried out with the approval of the responsible ethics committee and in accordance with national law and the Helsinki Declaration of 1975 (in its current, revised form). Informed consent was obtained from all patients included in studies.
References 1. Walter C et al (2007) Phantom and in-vivo measurements of dose exposure by image-guided radiotherapy (IGRT): MV portal images vs. kV portal images vs. cone-beam CT. Radiother Oncol 85(3):418–423 2. Amer A et al (2007) Imaging doses from the Elekta Synergy X-ray cone beam CT system. Br J Radiol 80(954):476–482 3. Guckenberger M et al (2012) Motion compensation in radiotherapy. Crit Rev Biomed Eng 40(3):187–197 4. Pallotta S et al (2012) A phantom evaluation of Sentinel(™), a commercial laser/camera surface imaging system for patient setup verification in radiotherapy. Med Phys 39(2):706–712 5. Stieler F et al (2012) Clinical evaluation of a commercial surface-imaging system for patient positioning in radiotherapy. Strahlenther Onkol 188(12):1080–1084 6. Gopan O, Wu Q (2012) Evaluation of the accuracy of a 3D surface imaging system for patient setup in head and neck cancer radiotherapy. Int J Radiat Oncol Biol Phys 84(2):547–552 7. Moser T et al (2012) Clinical evaluation of a laser surface scanning system in 120 patients for improving daily setup accuracy in fractionated radiation therapy. Int J Radiat Oncol Biol Phys 85(3):846–853 8. Pallotta S et al (2013) Accuracy of a 3D lasercamera surface imaging system for setup verification of the pelvic and thoracic regions in radiotherapy treatments. Med Phys 40(1):011710 9. Kauweloa KI et al (2012) GateCT surface tracking system for respiratory signal reconstruction in 4DCT imaging. Med Phys 39(1):492–502 10. Wolff D et al (2009) Volumetric modulated arc therapy (VMAT) vs. serial tomotherapy, step-andshoot IMRT and 3D-conformal RT for treatment of prostate cancer. Radiother Oncol 93(2):226–233 11. Stieler F et al (2011) A comparison of several modulated radiotherapy techniques for head and neck cancer and dosimetric validation of VMAT. Radiother Oncol 101(3):388–393 12. Hansen EK et al (2006) Repeat CT imaging and replanning during the course of IMRT for head-andneck cancer. Int J Radiat Oncol Biol Phys 64(2):355– 362 13. Ho KF et al (2012) Monitoring dosimetric impact of weight loss with kilovoltage (kV) cone beam CT (CBCT) during parotid-sparing IMRT and concurrent chemotherapy. Int J Radiat Oncol Biol Phys 82(3):e375–e382 14. Giraud P et al (2003) Respiration-gated radiotherapy: current techniques and potential benefits. Cancer Radiother 7(Suppl 1):15s–25s 15. Herk M van et al (2000) The probability of correct target dosage: dose-population histograms for deriving treatment margins in radiotherapy. Int J Radiat Oncol Biol Phys 47(4):1121–1135
Strahlentherapie und Onkologie 2013
| 7
