Home
Search
Collections
Journals
About
Contact us
My IOPscience
Standardized accuracy assessment of the calypso wireless transponder tracking system
This content has been downloaded from IOPscience. Please scroll down to see the full text. 2014 Phys. Med. Biol. 59 6797 (http://iopscience.iop.org/0031-9155/59/22/6797) View the table of contents for this issue, or go to the journal homepage for more Download details: IP Address: 128.119.168.112 This content was downloaded on 29/05/2017 at 12:07 Please note that terms and conditions apply.
You may also be interested in: Assessment of a new electromagnetic tracking system J Hummel, M Figl, W Birkfellner et al. Simultaneous, synchronized tumour tracking A W Rau, S Nill, R S Eidens et al. Volumetric characterization of the Aurora magnetic tracker system N C Atuegwu and R L Galloway Electromagnetic tracking system accuracy protocols D D Frantz, A D Wiles, S E Leis et al. An automated method for adaptive radiation therapy for prostate cancer patients using continuous fiducial-based tracking C E Noel, L Santanam, J R Olsen et al. 3D--2D registration for surgical guidance: effect of projection view angles on registration accuracy A Uneri, Y Otake, A S Wang et al. A Kinect™ camera based navigation system for percutaneous abdominal puncture Deqiang Xiao, Huoling Luo, Fucang Jia et al. A new hand-tracking method for primates S Schaffelhofer and H Scherberger Evaluation of motion management strategies based on required margins D Sawkey, M Svatos and C Zankowski
Institute of Physics and Engineering in Medicine Phys. Med. Biol. 59 (2014) 6797–6810
Physics in Medicine & Biology doi:10.1088/0031-9155/59/22/6797
Standardized accuracy assessment of the calypso wireless transponder tracking system A M Franz1, D Schmitt2, A Seitel3, M Chatrasingh4, G Echner2, U Oelfke2,5, S Nill5, W Birkfellner4,6 and L Maier-Hein1 1
Junior Group Computer-assisted Interventions, DKFZ, 69121 Heidelberg, Germany Division of Medical Physics in Radiation Oncology, DKFZ, 69121 Heidelberg, Germany 3 Robotics and Control Laboratory, University of British Columbia, 2332 Main Hall, Vancouver, BC, V6T 1ZA Canada 4 Center for Medical Physics and Biomedical Engineering, Medical University Vienna, 1090 Wien, Austria 5 Joint Department of Physics at The Institute of Cancer Research and The Royal Marsden NHS Foundation Trust, London, UK 6 Christian Doppler Laboratory for Medical Radiation Research for Radiation Oncology, Medical University Vienna, 1090 Wien, Austria 2
E-mail: [email protected] and [email protected] Received 7 March 2014, revised 21 July 2014 Accepted for publication 2 September 2014 Published 21 October 2014 Abstract
Electromagnetic (EM) tracking allows localization of small EM sensors in a magnetic field of known geometry without line-of-sight. However, this technique requires a cable connection to the tracked object. A wireless alternative based on magnetic fields, referred to as transponder tracking, has been proposed by several authors. Although most of the transponder tracking systems are still in an early stage of development and not ready for clinical use yet, Varian Medical Systems Inc. (Palo Alto, California, USA) presented the Calypso system for tumor tracking in radiation therapy which includes transponder technology. But it has not been used for computer-assisted interventions (CAI) in general or been assessed for accuracy in a standardized manner, so far. In this study, we apply a standardized assessment protocol presented by Hummel et al (2005 Med. Phys. 32 2371–9) to the Calypso system for the first time. The results show that transponder tracking with the Calypso system provides a precision and accuracy below 1 mm in ideal clinical environments, which is comparable with other EM tracking systems. Similar to other systems the tracking accuracy was affected by metallic distortion, which led to errors of up to 3.2 mm. The potential of the wireless transponder 0031-9155/14/226797+14$33.00 © 2014 Institute of Physics and Engineering in Medicine Printed in the UK & the USA
6797
A M Franz et al
Phys. Med. Biol. 59 (2014) 6797
tracking technology for use in many future CAI applications can be regarded as extremely high. Keywords: transponder tracking, electromagnetic tracking, calypso, computer-assisted interventions, standardized assessment, wireless tracking (Some figures may appear in colour only in the online journal) 1. Introduction Computer-Assisted Interventions (CAI) are increasingly gaining importance for today’s patient care (Peters and Cleary 2008). CAI systems usually combine medical imaging data of the patient with positional data of instruments to provide guidance information during an intervention. The accurate localization of objects of interest is accomplished by using tracking technology such as optical or electromagnetic (EM) tracking which thus is a key enabling technology for CAI. Optical trackers that use cameras for localization of visual markers are widespread in current CAI systems. However, these systems always require a line-of-sight to the tracked object and are unsuitable for many applications, such as interventions with flexible endoscopes, catheters or bendable needles, where line-of-sight to the tip of the instrument is blocked. EM tracking allows for localization of small EM sensors in a magnetic field of known geometry without line-of-sight (Franz et al 2014). But this technique relies on accurate measurements through the EM sensors and requires a cable connection to the tracked object. This cable often hampers the clinical workflow as it can block a working channel of an endoscope or require a thick catheter. An alternative technology, which is also based on EM fields, but additionally uses transponder localization signals was proposed by several authors (Mate et al 2004, Peters and Cleary 2008, Sanpechuda and Kovavisaruch 2008). It is referred to as transponder tracking and does not require a cable connection to the tracked object. Although most of the presented techniques are still in an early stage of development and not ready for clinical use yet, Varian Medical Systems Inc. (Palo Alto, California, USA) presented a system for tumor tracking in radiation therapy which includes transponder technology (Mate et al 2004). Various studies tested the Calypso system with custom assessment protocols. Balter et al (2005) and Santanam et al (2009) focused on tracking accuracy and precision, while Murphy et al (2008) assessed the effects of transponder motion. Wang et al (2012) investigated the effects of metallic nails as well as inflatable penile prostheses near the transponder and Bittner et al tested possible distortions of metal hip prostheses (Bittner et al 2014). The Calypso system was also tested for endoscopic surgery in feasibility studies (Sonnenday and Kaufman 2003, Nakamoto et al 2008) and the effects of a CT room have been assessed (Wen 2010). While these studies showed good tracking results with sub-millimeter accuracy and relatively small distortion effects, they all used different experimental setups and did not evaluate their data in the same way, which makes the results hardly comparable especially in relation to studies with other tracking systems. For EM tracking systems in general, different standardized assessment protocols have been presented and were applied many times (Franz et al 2014). The most established protocols were proposed by Wilson et al (2007) and Hummel et al (2005). The Wilson protocol provides a cube-shaped phantom (18 cm × 18 cm × 18 cm) with 225 holes of different depths. A tracked sensor is inserted into these holes and the tracking data of the known positions is acquired and evaluated. The Hummel protocol provides a Perspex board with drilled holes, where the tracked sensor can be fixed by means of a special sensor mount. Measurements on this board can be taken and evaluated on different height levels (Maier-Hein et al 2012, Sirokai et al 2012). Both protocols are relatively easy to apply and have been used to assess most of the 6798
A M Franz et al
Phys. Med. Biol. 59 (2014) 6797
Figure 1. Tracking principle of the Calypso system as described in Mate et al (2004).
available EM tracking systems (Hummel et al 2005, Wilson et al 2007, Yaniv et al 2009, Bø et al 2012, Maier-Hein et al 2012, Sirokai et al 2012). However, a comparative evaluation of transponder tracking to other EM tracking techniques has not yet been performed, which makes it difficult to judge the current potential of transponder systems in the context of CAI. Thus, the goal of this study was to adapt the Hummel protocol to the Calypso system and to perform a standardized comparable assessment of a transponder tracking system for the first time. 2. Materials and methods This section describes the Calypso GPS for the Body™ tracking system we evaluated (section 2.1), reviews the applied standardized assessment phantom and protocol developed by Hummel et al (2005) together with the adaptions to the Calypso system (section 2.2) and presents our experimental design in detail (sections 2.3–2.5). 2.1. Calypso GPS for the body system
The Calypso GPS for the Body is a system designed for radiation therapy, e.g. of the prostate (Balter et al 2005). It enables localization of a target structure, such as a tumor, by means of three implanted transponders. These transponders, referred to as beacons™, are excited by an external magnetic field provided by EM source coils included in the tracking system (Mate et al 2004) as shown in figure 1. Once excited, the beacons emit a location signal, which is measured by a sensor array. Based on the sensor data a tracking algorithm estimates the location of the beacon while no line-of-sight and no cable connection is needed. The system supports three beacons with resonance frequencies of 300, 400 and 500 kHz (Mate et al 2004), which are excited and localized sequentially. The update rate of the Calypso system used in this study was 10 Hz, leading to approximately 3.3 Hz per beacon; in a newer version of the system the rate was improved to 25 Hz. The system internally computes the center position of all localized beacons using a special algorithm depending on the selected tracking mode. For this study the isocenter tracking 6799
A M Franz et al
Phys. Med. Biol. 59 (2014) 6797
Figure 2. Pictures of the beacon mount (upper left), one single Calypso beacon (lower
left) and the modified Hummel Board (right). Due to the limited tracking volume of the Calypso system, only a central area on the board (dashed framed in grey) was used. Inside this area additional holes spaced at 2.5 cm were added to the board, leading to 5 × 5 possible positions of the beacon mount inside the grid.
mode (Mate et al 2004) was used. The center position is returned as tracking data and used to estimate the position of the target structure during medical application of the Calypso system in radiation therapy. In addition to the center position we also had access to the single beacon positions as a special feature for research purposes. The tracking volume of the Calypso system covers an area of 14 cm × 14 cm × 19 cm (Balter et al 2005, Mate et al 2004)6. It is worth mentioning that the Calypso system also includes an optical tracker which tracks the sensor array to calibrate it to the Linear Accelerator (LINAC) respectively to the radiotherapy room. However, the basic tracking component is the transponder tracking system, which is the only clinically approved tracking technology that is able to track miniaturized transponders without a cable connection. This study therefore focuses on the transponder tracking subsystem. 2.2. Assessment phantom
The phantom presented by Hummel et al (2005) consists of a measurement board (Hummel Board) and a mount which allows for positioning tracked objects on this board. Due to special properties of the Calypso system, some modifications of the original Hummel phantom were needed (see figure 2): Hummel Board: The tracking volume of the Calypso system is small compared to the tracking systems the Hummel Board was originally designed for. Under consideration of this restriction we only used an area of 10 cm × 10 cm in the central region of the Hummel Board, as shown in figure 2 on the right. Because this would lead to only 3 × 3 measurements on the board, we added further holes spaced at 2.5 cm—instead of the original 5 cm—in the central area. This leads to 5 × 5 = 25 measurements on the board. 6 The height of the tracking volume was wrong in some previous publications. Due to the specifications given by Varian the volume for beacon localization ranges from 8 cm below the array to 27 cm below the array resulting in a height of 19 cm.
6800
A M Franz et al
Phys. Med. Biol. 59 (2014) 6797
Figure 3. Drawing of the beacon mount, showing the positions of the three beacons
(B1, B2, B3) relative to the beacon center (C). All values are given in mm.
Beacon Mount: The Calypso system typically tracks three beacons in a known configuration to estimate the target position. This approach is caused by the fact that the system is a medical product for prostate radiotherapy, in which three beacons are usually implanted to gain 6 degree-of-freedom (DoF) tracking data. For this study we designed a special beacon mount for the Hummel Board, as shown in figures 2 and 3. It includes three beacons in a configuration that is similar to that used in a clinical application of the Calypso system. To obtain the beacon configuration represented by the beacon mount (see figure 3), the beacon configurations of 17 patients that were treated in the German Cancer Research Center (DKFZ) were averaged. The Hummel Board used for this study as well as the beacon mount were built in the workshop of the German Cancer Research Center using a CNC milling machine (type: Mikron WF 31C) providing a manufacturing tolerance of ± 0.01 mm. 2.3. Experimental setup
The Calypso assessment was performed in a radiotherapy room of the DKFZ, the usual operational environment of the Calypso system. The setup is shown in figure 4(a). As proposed by Maier-Hein et al (2012), the Hummel Board was placed on three different levels inside the tracking volume by using wooden height adapters. The middle level was defined such that the beacon center on the central position of the Hummel Board was in the center of the tracking volume. This led to a board pose 16 cm below the Calypso array, while the upper level was 5 cm above the middle and the lower level was 5 cm beneath. The other equipment in the room, including the LINAC, was running, but the radiation beam was not activated. Initial alignment of the phantom was performed using the laser alignment system of the radiotherapy room. A passive optical tracking system (MicronTracker 2, 6801
A M Franz et al
Phys. Med. Biol. 59 (2014) 6797
Figure 4. (a) Photograph of the experimental setup inside the radiation therapy room of the DKFZ. The passive optical tracker (MicronTracker 2, Claron Technology) was used to assure that neither phantom nor Calypso array was moved during the measurements. (b) For the metallic distortions experiment, four sample cylinders of steel (SST 303 and SST 416), aluminum, and bronze were placed in five different positions (H1–H5) between the beacon mount and the Calypso array.
Claron Technology Inc., Carlton, Ontario, Canada), that does not disturb the optical component of the Calypso system was used to assure that neither assessment phantom nor Calypso array was moved during the measurements. In addition to the experiment described above, the Hummel protocol (Hummel et al 2005) proposes to assess metallic distortion in a separate experimental setup, as shown in figure 4(b). For this purpose different sample cylinders of 300- and 400-series steel (SST 303 and SST 416), bronze, and aluminum—each with a length of 5 cm and a diameter of 1 cm—were placed between the beacon mount and the Calypso array. The beacon mount was placed on the central position of the Hummel Board on the middle level. A plastic holder was used to place the sample cylinders in five different positions. 2.4. Data acquisition
The acquisition protocol by Hummel et al (2005) was applied with the beacon mount on three different levels as described above. On each level, measurements were recorded on 5 × 5 = 25 positions. The Calypso system provides the center of the three beacons as well as single positions of each beacon. The center, hereinafter referred to as beacon center (C), is calculated by the Calypso system similar to a running mean. That means the center is always updated after new tracking data from one beacon arrives, using this data together with the latest tracking data from the other beacons. The beacons are updated sequentially with an update rate of 10 Hz. According to the protocol, 150 values per position should be recorded (Maier-Hein et al 2012). To gain 150 values per single beacon, we took 450 beacon center values per position in this study. To evaluate the beacon center, we then took every third tracked position, so that it was completely updated for each datum. 2.5. Data evaluation
For each of the three levels inside the tracking volume (see figure 4(a)), data acquisition yielded a set of 150 (j = 0..150) measurements pBx, i, j per beacon Bx = (B1, B2, B3) for each 6802
A M Franz et al
Phys. Med. Biol. 59 (2014) 6797
pose i(i = 1..25) of the beacon mount. In addition to every beacons' position the Calypso system also provides the position of the beacon center (C) pC, i, j. For each pose i the mean of C, B1, B2, and B3 was calculated. This led to 4 × 25 mean values p[C ∣ Bx ], i i = 1..25; p[C ∣ Bx ], i ∈ R3 for each level.
(
)
Precision: The jitter error for each position i was defined as the root mean square (RMS) error of the N = 150 values: ϵ jitter, ( [C∣Bx ] , i ) =
1 N
N
∑
j=1
p[C ∣ Bx ], i, j − p[C ∣ Bx ], i, j
2 2
(1)
where ∣ … ∣2 denotes the Euclidean norm. Distance accuracy: The distance accuracy was determined by comparing the Euclidean distances between mean tracked locations to the known physical distances on the measurement plate. For the tracking volumes of previously tested EM tracking devices, the Hummel protocol proposes to evaluate 5 cm, 15 cm and 30 cm distances. However, due to the smaller tracking volume we added further holes to the measurement plate and evaluated 2.5 cm and 5 cm distances. For both, all possible distances were calculated. In case of the 2.5 cm distances this led to 5 × 4 distances in the rows of the grid (see figure 2) and 4 × 5 distances in the columns, and, hence, 40 distances in total. Similarly, 3 × 2 + 2 × 3 = 12 distances were obtained for 5 cm. For a more detailed description of how distances are defined by the Hummel protocol the reader is referred to Maier-Hein et al (2012). Grid accuracy: As another measure of accuracy, the Hummel protocol recommends determining so-called accumulated distances and plotting the result in a 3D graph (Hummel et al 2005). Due to the tracking volume of the Calypso system, we omitted this measure and performed a grid matching instead, as proposed by Maier-Hein et al for smaller tracking volumes (Maier-Hein et al 2012). The grid accuracy was determined by matching the set of 25 measured mean grid positions p[C ∣ Bx ], i (i = 1..25) of one level to the set of reference positions with the optimal transformation in a least square sense (Maier-Hein et al 2012). The fiducial registration error (FRE) of this transformation is used as measure of grid accuracy. The grid accuracy was determined for the beacon center (C) and the three single beacon positions (Bx) on each of the three levels upper, middle, lower. Comparison: For a meaningful comparison of the Calypso system to another EM tracking system, we used data from a previous study with the NDI (Northern Digital Inc., Waterloo, ON, Canada) Aurora system (Maier-Hein et al 2012) and evaluated it on the same 3 × 3 = 9 positions (5.0 cm distances) on three levels as we did for the Calypso system in this study. For good comparability we took data from a field generator (FG) with a similar tracking volume, the NDI Compact FG (Maier-Hein et al 2012). Based on this data, the precision, distance accuracy (5.0 cm), and grid accuracy (of 9 positions) of the NDI Aurora system were determined as described above. Distortion: To evaluate the metallic distortions, positional data was aquired while the beacon mount was fixed on the central position on the middle level as described in section 2.4. The mean position of the beacon center (C) and the three single beacons (Bx) was computed for each possible position Hx = (H1, H2, H3, H4, H5) (see figure 4(b)) of the four metal cylinders m = (SST303, SST416, aluminum, bronze). The mean positions were compared to a reference measurement pref which was taken without any metal probe. ϵdistortion([C∣Bx], m, Hx) was defined as Euclidian distance between the mean position p[C ∣ Bx ], m, Hx and pref . 6803
A M Franz et al
Phys. Med. Biol. 59 (2014) 6797
Table 1. Results of the jitter evaluation over 25 RMS error values per level in mm (μ ± σ). Every RMS error value was computed from 150 separate measurements.
[mm]
Center
Beacon 1
Beacon 2
Beacon 3
Upper Middle Lower Overall
0.3 ± 0.3 0.1 ± 0.0 0.8 ± 0.1 0.4 ± 0.4
0.2 ± 0.0 0.1 ± 0.0 0.6 ± 0.1*2 0.3 ± 0.2
1.1 ± 2.1*29 0.1 ± 0.0 0.3 ± 0.0*1 0.5 ± 1.3
0.2 ± 0.0 0.1 ± 0.0 0.4 ± 0.0*1 0.2 ± 0.1
Note: The number of invalid samples is given superscripted (*) if ≠ 0. Table 2. Results
of the direct comparison between the Calypso and the NDI Aurora system. To enable this comparison, evaluation was limited to the positions with 5 cm distances of the original Hummel protocol (Hummel et al 2005). This led to 3 × 3 = 9 positions per level inside the tracking volume of the Calypso system, on which precision (n = 9 points), 5 cm distance accuracy (n = 12 distances) and grid accuracy (n = 9 points) were computed based on the tracking data from one beacon (beacon 1 from this study) and one sensor (a 5 DoF NDI Aurora sensor from a previous study (Maier-Hein et al 2012)). For the NDI Aurora system a Compact FG (Maier-Hein et al 2012) with a tracking volume similar to those of the Calypso system was used. Precision (μ ± σ)
Distance acc. (μ ± σ)
Grid acc. (RMS)
[mm]
Calypso
Aurora
Calypso
Aurora
Calypso Aurora
Upper Middle Lower Overall
0.2 ± 0.0 0.1 ± 0.0 0.6 ± 0.1 0.3 ± 0.2
0.0 ± 0.0 0.1 ± 0.0 0.1 ± 0.0 0.1 ± 0.0
0.1 ± 0.1 0.1 ± 0.0 0.1 ± 0.1 0.1 ± 0.1
0.6 ± 0.6 0.2 ± 0.1 0.5 ± 0.5 0.4 ± 0.5
0.1 0.1 0.2 0.1
0.9 0.2 0.8 0.6
3. Results Concerning precision, the jitter of the beacon center averaged over all 3 × 25 = 75 grid positions was 0.4 ± 0.4 mm (μ ± σ). The results separated by pose of the Hummel Board and by the single beacons are shown in table 1. When limited to the original grid positions of the Hummel Board with 5 cm distances (3 × 9 = 27 positions), the results look similar with an overall mean value of 0.4 ± 0.4 mm. For beacon 1 these results are listed in table 2 for direct comparison with the NDI Aurora system. The measurements included some invalid samples on the lower level and on the upper level (
Search
Collections
Journals
About
Contact us
My IOPscience
Standardized accuracy assessment of the calypso wireless transponder tracking system
This content has been downloaded from IOPscience. Please scroll down to see the full text. 2014 Phys. Med. Biol. 59 6797 (http://iopscience.iop.org/0031-9155/59/22/6797) View the table of contents for this issue, or go to the journal homepage for more Download details: IP Address: 128.119.168.112 This content was downloaded on 29/05/2017 at 12:07 Please note that terms and conditions apply.
You may also be interested in: Assessment of a new electromagnetic tracking system J Hummel, M Figl, W Birkfellner et al. Simultaneous, synchronized tumour tracking A W Rau, S Nill, R S Eidens et al. Volumetric characterization of the Aurora magnetic tracker system N C Atuegwu and R L Galloway Electromagnetic tracking system accuracy protocols D D Frantz, A D Wiles, S E Leis et al. An automated method for adaptive radiation therapy for prostate cancer patients using continuous fiducial-based tracking C E Noel, L Santanam, J R Olsen et al. 3D--2D registration for surgical guidance: effect of projection view angles on registration accuracy A Uneri, Y Otake, A S Wang et al. A Kinect™ camera based navigation system for percutaneous abdominal puncture Deqiang Xiao, Huoling Luo, Fucang Jia et al. A new hand-tracking method for primates S Schaffelhofer and H Scherberger Evaluation of motion management strategies based on required margins D Sawkey, M Svatos and C Zankowski
Institute of Physics and Engineering in Medicine Phys. Med. Biol. 59 (2014) 6797–6810
Physics in Medicine & Biology doi:10.1088/0031-9155/59/22/6797
Standardized accuracy assessment of the calypso wireless transponder tracking system A M Franz1, D Schmitt2, A Seitel3, M Chatrasingh4, G Echner2, U Oelfke2,5, S Nill5, W Birkfellner4,6 and L Maier-Hein1 1
Junior Group Computer-assisted Interventions, DKFZ, 69121 Heidelberg, Germany Division of Medical Physics in Radiation Oncology, DKFZ, 69121 Heidelberg, Germany 3 Robotics and Control Laboratory, University of British Columbia, 2332 Main Hall, Vancouver, BC, V6T 1ZA Canada 4 Center for Medical Physics and Biomedical Engineering, Medical University Vienna, 1090 Wien, Austria 5 Joint Department of Physics at The Institute of Cancer Research and The Royal Marsden NHS Foundation Trust, London, UK 6 Christian Doppler Laboratory for Medical Radiation Research for Radiation Oncology, Medical University Vienna, 1090 Wien, Austria 2
E-mail: [email protected] and [email protected] Received 7 March 2014, revised 21 July 2014 Accepted for publication 2 September 2014 Published 21 October 2014 Abstract
Electromagnetic (EM) tracking allows localization of small EM sensors in a magnetic field of known geometry without line-of-sight. However, this technique requires a cable connection to the tracked object. A wireless alternative based on magnetic fields, referred to as transponder tracking, has been proposed by several authors. Although most of the transponder tracking systems are still in an early stage of development and not ready for clinical use yet, Varian Medical Systems Inc. (Palo Alto, California, USA) presented the Calypso system for tumor tracking in radiation therapy which includes transponder technology. But it has not been used for computer-assisted interventions (CAI) in general or been assessed for accuracy in a standardized manner, so far. In this study, we apply a standardized assessment protocol presented by Hummel et al (2005 Med. Phys. 32 2371–9) to the Calypso system for the first time. The results show that transponder tracking with the Calypso system provides a precision and accuracy below 1 mm in ideal clinical environments, which is comparable with other EM tracking systems. Similar to other systems the tracking accuracy was affected by metallic distortion, which led to errors of up to 3.2 mm. The potential of the wireless transponder 0031-9155/14/226797+14$33.00 © 2014 Institute of Physics and Engineering in Medicine Printed in the UK & the USA
6797
A M Franz et al
Phys. Med. Biol. 59 (2014) 6797
tracking technology for use in many future CAI applications can be regarded as extremely high. Keywords: transponder tracking, electromagnetic tracking, calypso, computer-assisted interventions, standardized assessment, wireless tracking (Some figures may appear in colour only in the online journal) 1. Introduction Computer-Assisted Interventions (CAI) are increasingly gaining importance for today’s patient care (Peters and Cleary 2008). CAI systems usually combine medical imaging data of the patient with positional data of instruments to provide guidance information during an intervention. The accurate localization of objects of interest is accomplished by using tracking technology such as optical or electromagnetic (EM) tracking which thus is a key enabling technology for CAI. Optical trackers that use cameras for localization of visual markers are widespread in current CAI systems. However, these systems always require a line-of-sight to the tracked object and are unsuitable for many applications, such as interventions with flexible endoscopes, catheters or bendable needles, where line-of-sight to the tip of the instrument is blocked. EM tracking allows for localization of small EM sensors in a magnetic field of known geometry without line-of-sight (Franz et al 2014). But this technique relies on accurate measurements through the EM sensors and requires a cable connection to the tracked object. This cable often hampers the clinical workflow as it can block a working channel of an endoscope or require a thick catheter. An alternative technology, which is also based on EM fields, but additionally uses transponder localization signals was proposed by several authors (Mate et al 2004, Peters and Cleary 2008, Sanpechuda and Kovavisaruch 2008). It is referred to as transponder tracking and does not require a cable connection to the tracked object. Although most of the presented techniques are still in an early stage of development and not ready for clinical use yet, Varian Medical Systems Inc. (Palo Alto, California, USA) presented a system for tumor tracking in radiation therapy which includes transponder technology (Mate et al 2004). Various studies tested the Calypso system with custom assessment protocols. Balter et al (2005) and Santanam et al (2009) focused on tracking accuracy and precision, while Murphy et al (2008) assessed the effects of transponder motion. Wang et al (2012) investigated the effects of metallic nails as well as inflatable penile prostheses near the transponder and Bittner et al tested possible distortions of metal hip prostheses (Bittner et al 2014). The Calypso system was also tested for endoscopic surgery in feasibility studies (Sonnenday and Kaufman 2003, Nakamoto et al 2008) and the effects of a CT room have been assessed (Wen 2010). While these studies showed good tracking results with sub-millimeter accuracy and relatively small distortion effects, they all used different experimental setups and did not evaluate their data in the same way, which makes the results hardly comparable especially in relation to studies with other tracking systems. For EM tracking systems in general, different standardized assessment protocols have been presented and were applied many times (Franz et al 2014). The most established protocols were proposed by Wilson et al (2007) and Hummel et al (2005). The Wilson protocol provides a cube-shaped phantom (18 cm × 18 cm × 18 cm) with 225 holes of different depths. A tracked sensor is inserted into these holes and the tracking data of the known positions is acquired and evaluated. The Hummel protocol provides a Perspex board with drilled holes, where the tracked sensor can be fixed by means of a special sensor mount. Measurements on this board can be taken and evaluated on different height levels (Maier-Hein et al 2012, Sirokai et al 2012). Both protocols are relatively easy to apply and have been used to assess most of the 6798
A M Franz et al
Phys. Med. Biol. 59 (2014) 6797
Figure 1. Tracking principle of the Calypso system as described in Mate et al (2004).
available EM tracking systems (Hummel et al 2005, Wilson et al 2007, Yaniv et al 2009, Bø et al 2012, Maier-Hein et al 2012, Sirokai et al 2012). However, a comparative evaluation of transponder tracking to other EM tracking techniques has not yet been performed, which makes it difficult to judge the current potential of transponder systems in the context of CAI. Thus, the goal of this study was to adapt the Hummel protocol to the Calypso system and to perform a standardized comparable assessment of a transponder tracking system for the first time. 2. Materials and methods This section describes the Calypso GPS for the Body™ tracking system we evaluated (section 2.1), reviews the applied standardized assessment phantom and protocol developed by Hummel et al (2005) together with the adaptions to the Calypso system (section 2.2) and presents our experimental design in detail (sections 2.3–2.5). 2.1. Calypso GPS for the body system
The Calypso GPS for the Body is a system designed for radiation therapy, e.g. of the prostate (Balter et al 2005). It enables localization of a target structure, such as a tumor, by means of three implanted transponders. These transponders, referred to as beacons™, are excited by an external magnetic field provided by EM source coils included in the tracking system (Mate et al 2004) as shown in figure 1. Once excited, the beacons emit a location signal, which is measured by a sensor array. Based on the sensor data a tracking algorithm estimates the location of the beacon while no line-of-sight and no cable connection is needed. The system supports three beacons with resonance frequencies of 300, 400 and 500 kHz (Mate et al 2004), which are excited and localized sequentially. The update rate of the Calypso system used in this study was 10 Hz, leading to approximately 3.3 Hz per beacon; in a newer version of the system the rate was improved to 25 Hz. The system internally computes the center position of all localized beacons using a special algorithm depending on the selected tracking mode. For this study the isocenter tracking 6799
A M Franz et al
Phys. Med. Biol. 59 (2014) 6797
Figure 2. Pictures of the beacon mount (upper left), one single Calypso beacon (lower
left) and the modified Hummel Board (right). Due to the limited tracking volume of the Calypso system, only a central area on the board (dashed framed in grey) was used. Inside this area additional holes spaced at 2.5 cm were added to the board, leading to 5 × 5 possible positions of the beacon mount inside the grid.
mode (Mate et al 2004) was used. The center position is returned as tracking data and used to estimate the position of the target structure during medical application of the Calypso system in radiation therapy. In addition to the center position we also had access to the single beacon positions as a special feature for research purposes. The tracking volume of the Calypso system covers an area of 14 cm × 14 cm × 19 cm (Balter et al 2005, Mate et al 2004)6. It is worth mentioning that the Calypso system also includes an optical tracker which tracks the sensor array to calibrate it to the Linear Accelerator (LINAC) respectively to the radiotherapy room. However, the basic tracking component is the transponder tracking system, which is the only clinically approved tracking technology that is able to track miniaturized transponders without a cable connection. This study therefore focuses on the transponder tracking subsystem. 2.2. Assessment phantom
The phantom presented by Hummel et al (2005) consists of a measurement board (Hummel Board) and a mount which allows for positioning tracked objects on this board. Due to special properties of the Calypso system, some modifications of the original Hummel phantom were needed (see figure 2): Hummel Board: The tracking volume of the Calypso system is small compared to the tracking systems the Hummel Board was originally designed for. Under consideration of this restriction we only used an area of 10 cm × 10 cm in the central region of the Hummel Board, as shown in figure 2 on the right. Because this would lead to only 3 × 3 measurements on the board, we added further holes spaced at 2.5 cm—instead of the original 5 cm—in the central area. This leads to 5 × 5 = 25 measurements on the board. 6 The height of the tracking volume was wrong in some previous publications. Due to the specifications given by Varian the volume for beacon localization ranges from 8 cm below the array to 27 cm below the array resulting in a height of 19 cm.
6800
A M Franz et al
Phys. Med. Biol. 59 (2014) 6797
Figure 3. Drawing of the beacon mount, showing the positions of the three beacons
(B1, B2, B3) relative to the beacon center (C). All values are given in mm.
Beacon Mount: The Calypso system typically tracks three beacons in a known configuration to estimate the target position. This approach is caused by the fact that the system is a medical product for prostate radiotherapy, in which three beacons are usually implanted to gain 6 degree-of-freedom (DoF) tracking data. For this study we designed a special beacon mount for the Hummel Board, as shown in figures 2 and 3. It includes three beacons in a configuration that is similar to that used in a clinical application of the Calypso system. To obtain the beacon configuration represented by the beacon mount (see figure 3), the beacon configurations of 17 patients that were treated in the German Cancer Research Center (DKFZ) were averaged. The Hummel Board used for this study as well as the beacon mount were built in the workshop of the German Cancer Research Center using a CNC milling machine (type: Mikron WF 31C) providing a manufacturing tolerance of ± 0.01 mm. 2.3. Experimental setup
The Calypso assessment was performed in a radiotherapy room of the DKFZ, the usual operational environment of the Calypso system. The setup is shown in figure 4(a). As proposed by Maier-Hein et al (2012), the Hummel Board was placed on three different levels inside the tracking volume by using wooden height adapters. The middle level was defined such that the beacon center on the central position of the Hummel Board was in the center of the tracking volume. This led to a board pose 16 cm below the Calypso array, while the upper level was 5 cm above the middle and the lower level was 5 cm beneath. The other equipment in the room, including the LINAC, was running, but the radiation beam was not activated. Initial alignment of the phantom was performed using the laser alignment system of the radiotherapy room. A passive optical tracking system (MicronTracker 2, 6801
A M Franz et al
Phys. Med. Biol. 59 (2014) 6797
Figure 4. (a) Photograph of the experimental setup inside the radiation therapy room of the DKFZ. The passive optical tracker (MicronTracker 2, Claron Technology) was used to assure that neither phantom nor Calypso array was moved during the measurements. (b) For the metallic distortions experiment, four sample cylinders of steel (SST 303 and SST 416), aluminum, and bronze were placed in five different positions (H1–H5) between the beacon mount and the Calypso array.
Claron Technology Inc., Carlton, Ontario, Canada), that does not disturb the optical component of the Calypso system was used to assure that neither assessment phantom nor Calypso array was moved during the measurements. In addition to the experiment described above, the Hummel protocol (Hummel et al 2005) proposes to assess metallic distortion in a separate experimental setup, as shown in figure 4(b). For this purpose different sample cylinders of 300- and 400-series steel (SST 303 and SST 416), bronze, and aluminum—each with a length of 5 cm and a diameter of 1 cm—were placed between the beacon mount and the Calypso array. The beacon mount was placed on the central position of the Hummel Board on the middle level. A plastic holder was used to place the sample cylinders in five different positions. 2.4. Data acquisition
The acquisition protocol by Hummel et al (2005) was applied with the beacon mount on three different levels as described above. On each level, measurements were recorded on 5 × 5 = 25 positions. The Calypso system provides the center of the three beacons as well as single positions of each beacon. The center, hereinafter referred to as beacon center (C), is calculated by the Calypso system similar to a running mean. That means the center is always updated after new tracking data from one beacon arrives, using this data together with the latest tracking data from the other beacons. The beacons are updated sequentially with an update rate of 10 Hz. According to the protocol, 150 values per position should be recorded (Maier-Hein et al 2012). To gain 150 values per single beacon, we took 450 beacon center values per position in this study. To evaluate the beacon center, we then took every third tracked position, so that it was completely updated for each datum. 2.5. Data evaluation
For each of the three levels inside the tracking volume (see figure 4(a)), data acquisition yielded a set of 150 (j = 0..150) measurements pBx, i, j per beacon Bx = (B1, B2, B3) for each 6802
A M Franz et al
Phys. Med. Biol. 59 (2014) 6797
pose i(i = 1..25) of the beacon mount. In addition to every beacons' position the Calypso system also provides the position of the beacon center (C) pC, i, j. For each pose i the mean of C, B1, B2, and B3 was calculated. This led to 4 × 25 mean values p[C ∣ Bx ], i i = 1..25; p[C ∣ Bx ], i ∈ R3 for each level.
(
)
Precision: The jitter error for each position i was defined as the root mean square (RMS) error of the N = 150 values: ϵ jitter, ( [C∣Bx ] , i ) =
1 N
N
∑
j=1
p[C ∣ Bx ], i, j − p[C ∣ Bx ], i, j
2 2
(1)
where ∣ … ∣2 denotes the Euclidean norm. Distance accuracy: The distance accuracy was determined by comparing the Euclidean distances between mean tracked locations to the known physical distances on the measurement plate. For the tracking volumes of previously tested EM tracking devices, the Hummel protocol proposes to evaluate 5 cm, 15 cm and 30 cm distances. However, due to the smaller tracking volume we added further holes to the measurement plate and evaluated 2.5 cm and 5 cm distances. For both, all possible distances were calculated. In case of the 2.5 cm distances this led to 5 × 4 distances in the rows of the grid (see figure 2) and 4 × 5 distances in the columns, and, hence, 40 distances in total. Similarly, 3 × 2 + 2 × 3 = 12 distances were obtained for 5 cm. For a more detailed description of how distances are defined by the Hummel protocol the reader is referred to Maier-Hein et al (2012). Grid accuracy: As another measure of accuracy, the Hummel protocol recommends determining so-called accumulated distances and plotting the result in a 3D graph (Hummel et al 2005). Due to the tracking volume of the Calypso system, we omitted this measure and performed a grid matching instead, as proposed by Maier-Hein et al for smaller tracking volumes (Maier-Hein et al 2012). The grid accuracy was determined by matching the set of 25 measured mean grid positions p[C ∣ Bx ], i (i = 1..25) of one level to the set of reference positions with the optimal transformation in a least square sense (Maier-Hein et al 2012). The fiducial registration error (FRE) of this transformation is used as measure of grid accuracy. The grid accuracy was determined for the beacon center (C) and the three single beacon positions (Bx) on each of the three levels upper, middle, lower. Comparison: For a meaningful comparison of the Calypso system to another EM tracking system, we used data from a previous study with the NDI (Northern Digital Inc., Waterloo, ON, Canada) Aurora system (Maier-Hein et al 2012) and evaluated it on the same 3 × 3 = 9 positions (5.0 cm distances) on three levels as we did for the Calypso system in this study. For good comparability we took data from a field generator (FG) with a similar tracking volume, the NDI Compact FG (Maier-Hein et al 2012). Based on this data, the precision, distance accuracy (5.0 cm), and grid accuracy (of 9 positions) of the NDI Aurora system were determined as described above. Distortion: To evaluate the metallic distortions, positional data was aquired while the beacon mount was fixed on the central position on the middle level as described in section 2.4. The mean position of the beacon center (C) and the three single beacons (Bx) was computed for each possible position Hx = (H1, H2, H3, H4, H5) (see figure 4(b)) of the four metal cylinders m = (SST303, SST416, aluminum, bronze). The mean positions were compared to a reference measurement pref which was taken without any metal probe. ϵdistortion([C∣Bx], m, Hx) was defined as Euclidian distance between the mean position p[C ∣ Bx ], m, Hx and pref . 6803
A M Franz et al
Phys. Med. Biol. 59 (2014) 6797
Table 1. Results of the jitter evaluation over 25 RMS error values per level in mm (μ ± σ). Every RMS error value was computed from 150 separate measurements.
[mm]
Center
Beacon 1
Beacon 2
Beacon 3
Upper Middle Lower Overall
0.3 ± 0.3 0.1 ± 0.0 0.8 ± 0.1 0.4 ± 0.4
0.2 ± 0.0 0.1 ± 0.0 0.6 ± 0.1*2 0.3 ± 0.2
1.1 ± 2.1*29 0.1 ± 0.0 0.3 ± 0.0*1 0.5 ± 1.3
0.2 ± 0.0 0.1 ± 0.0 0.4 ± 0.0*1 0.2 ± 0.1
Note: The number of invalid samples is given superscripted (*) if ≠ 0. Table 2. Results
of the direct comparison between the Calypso and the NDI Aurora system. To enable this comparison, evaluation was limited to the positions with 5 cm distances of the original Hummel protocol (Hummel et al 2005). This led to 3 × 3 = 9 positions per level inside the tracking volume of the Calypso system, on which precision (n = 9 points), 5 cm distance accuracy (n = 12 distances) and grid accuracy (n = 9 points) were computed based on the tracking data from one beacon (beacon 1 from this study) and one sensor (a 5 DoF NDI Aurora sensor from a previous study (Maier-Hein et al 2012)). For the NDI Aurora system a Compact FG (Maier-Hein et al 2012) with a tracking volume similar to those of the Calypso system was used. Precision (μ ± σ)
Distance acc. (μ ± σ)
Grid acc. (RMS)
[mm]
Calypso
Aurora
Calypso
Aurora
Calypso Aurora
Upper Middle Lower Overall
0.2 ± 0.0 0.1 ± 0.0 0.6 ± 0.1 0.3 ± 0.2
0.0 ± 0.0 0.1 ± 0.0 0.1 ± 0.0 0.1 ± 0.0
0.1 ± 0.1 0.1 ± 0.0 0.1 ± 0.1 0.1 ± 0.1
0.6 ± 0.6 0.2 ± 0.1 0.5 ± 0.5 0.4 ± 0.5
0.1 0.1 0.2 0.1
0.9 0.2 0.8 0.6
3. Results Concerning precision, the jitter of the beacon center averaged over all 3 × 25 = 75 grid positions was 0.4 ± 0.4 mm (μ ± σ). The results separated by pose of the Hummel Board and by the single beacons are shown in table 1. When limited to the original grid positions of the Hummel Board with 5 cm distances (3 × 9 = 27 positions), the results look similar with an overall mean value of 0.4 ± 0.4 mm. For beacon 1 these results are listed in table 2 for direct comparison with the NDI Aurora system. The measurements included some invalid samples on the lower level and on the upper level (
