Brachytherapy 13 (2014) 640e650

Electromagnetic tracking for catheter reconstruction in ultrasound-guided high-dose-rate brachytherapy of the prostate Shyam Bharat1,*, Cynthia Kung1, Ehsan Dehghan1, Ananth Ravi2, Niranjan Venugopal2, Antonio Bonillas1, Doug Stanton1, Jochen Kruecker1 1

Department of Ultrasound Imaging and Interventions, Philips Research North America, Briarcliff Manor, NY 2 Department of Radiation Oncology, Sunnybrook Health Sciences Center, Toronto, ON, Canada

ABSTRACT

PURPOSE: The accurate delivery of high-dose-rate brachytherapy is dependent on the correct identification of the position and shape of the treatment catheters. In many brachytherapy clinics, transrectal ultrasound (TRUS) imaging is used to identify the catheters. However, manual catheter identification on TRUS images can be time consuming, subjective, and operator dependent because of calcifications and distal shadowing artifacts. We report the use of electromagnetic (EM) tracking technology to map the position and shape of catheters inserted in a tissuemimicking phantom. METHODS AND MATERIALS: The accuracy of the EM system was comprehensively quantified using a three-axis robotic system. In addition, EM tracks acquired from catheters in a phantom were compared with catheter positions determined from TRUS and CT images to compare EM system performance to standard clinical imaging modalities. The tracking experiments were performed in a controlled laboratory environment and also in a typical brachytherapy operating room to test for potential EM distortions. RESULTS: The robotic validation of the EM system yielded a mean accuracy of !0.5 mm for a clinically acceptable field of view in a nondistorting environment. The EM-tracked catheter representations were found to have an accuracy of !1 mm when compared with TRUS- and CTidentified positions, both in the laboratory environment and in the brachytherapy operating room. The achievable accuracy depends to a large extent on the calibration of the TRUS probe, geometry of the tracked devices relative to the EM field generator, and locations of surrounding clinical equipment. To address the issue of variable accuracy, a robust calibration algorithm has been developed and integrated into the workflow. The proposed mapping technique was also found to improve the workflow efficiency of catheter identification. CONCLUSIONS: The high baseline accuracy of the EM system, the consistent agreement between EM-tracked, TRUS- and CT-identified catheters, and the improved workflow efficiency illustrate the potential value of using EM tracking for catheter mapping in high-dose-rate brachytherapy. Ó 2014 American Brachytherapy Society. Published by Elsevier Inc. All rights reserved.

Keywords:

High-dose-rate brachytherapy; Electromagnetic tracking; Treatment planning; Catheter mapping

Introduction

Received 7 January 2014; received in revised form 8 April 2014; accepted 6 May 2014. Financial disclosure/conflict of interest: AR has a funded research agreement with Philips Healthcare. SB, CK, ED, AB, DS, and JK are salaried employees of Philips. * Corresponding author. Philips Research North America, 345 Scarborough Road, Briarcliff Manor, NY 10510. Tel.: þ914-945-6374; fax: þ914-945-6330. E-mail address: [email protected] (S. Bharat).

Because of the ablative doses encountered in high-doserate (HDR) brachytherapy and to minimize the dose to nearby organs at risk, it is necessary to be able to generate and deliver dosimetric distributions with sharp gradients (1e4). To this end, accurate placement and localization of the catheters is critical. In prostate HDR brachytherapy, transrectal ultrasound (TRUS) image guidance is used for transperineal insertion of 10e20 catheters, often in a generic template-based pattern. The ability to visualize the catheters simultaneously

1538-4721/$ - see front matter Ó 2014 American Brachytherapy Society. Published by Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.brachy.2014.05.012

S. Bharat et al. / Brachytherapy 13 (2014) 640e650

with the prostate, rectal wall, and urethra in real time makes TRUS the preferred modality for image guidance during catheter insertion (5). After insertion, CT, MRI, or TRUS imaging may be used to localize the catheters and contour the prostate, urethra, bladder, and rectum (6e9). Next, an anatomy-based inverse planning algorithm uses the catheter positions and the organ contours to optimize the dwell positions and dwell times of radioactive sources in the catheters (10). Therefore, it is vital to be able to accurately implant the catheters relative to the patient’s anatomy and then digitize this geometry in the treatment planning system. CT provides good quality visualization of the catheters with high geometric fidelity (5). However, the use of CT for catheter localization and organ contouring has the disadvantages of having to transport the patient to the CT scanner (5, 7), the associated possibility of catheter displacement (11), and the subjectivity of prostate visualization on CT (12). The use of MRI for prostate HDR treatment planning has the benefit of exquisite soft tissue definition and the ability to visualize voids for the implanted catheters. However, patient transfer issues similar to that of CT treatment planning and limited access to this high value resource make MR approaches difficult to implement. Manual catheter identification on TRUS images can be subjective, challenging, and operator dependent because of problems of calcifications and distal shadowing artifacts. Figure 1 illustrates a typical ultrasound (US) B-mode axial image of the prostate, with multiple catheters implanted. Having many catheters in a small volume increases the level of difficulty in accurately identifying the catheters in axial and sagittal TRUS images.

641

Typically in clinical practice, the operator scrolls through several US slices in the three-dimensional (3D) data set and sometimes adjusts the TRUS probe in the live twodimensional (2D) mode to get a more complete picture of the catheter locations, which can be a time-consuming process. To improve efficiency and reduce potential interoperator variability in catheter identification, we propose an electromagnetic (EM) tracking solution that has the ability to provide enhanced intraoperative image guidance and accurately map the 3D locations of the catheters while also improving workflow efficiency in HDR prostate brachytherapy. EM tracking technology works on the principle of an electric field being induced in response to a time-varying magnetic field (Faraday’s law). A field generator (FG) produces a weak magnetic field that is calibrated so that the position and orientation of an EM sensor/tool in its field can be determined by analyzing the amplitude of the weak electric current that is generated in the tool. EM technology has previously been used in procedures such as interventional radiology, biopsy guidance, radiofrequency ablation, and external-beam radiation therapy (13e17). The performance accuracy of EM tracking in an HDR operating room (OR) was reported by Zhou et al. (18) in a recent publication. In this article, we describe an EM system customized for use in HDR brachytherapy procedures, including a novel framework for registration with TRUS imaging. The accuracy of the EM system was tested using a threeaxis robotic system. We also report feasibility results on the use of EM tracking for mapping the 3D positions of catheters in a prostate phantom, concurrent with TRUS imaging. A framework for registering EM catheter positions to the TRUS image and generating a calibration that is robust against environmental distortions has also been developed. The EM-reported catheter positions were validated against manually identified catheter positions on TRUS and CT images. The experiments were performed in an ideal laboratory environment and in a brachytherapy OR, the latter to test the potential impact of the presence of clinical equipment around the treatment table on the EM measurements.

Methods and materials The EM system

Fig. 1. Axial transrectal ultrasound image of the prostate, with catheters in place (illustrated by the dots). The bright spots represent the proximal edge of the catheters. Note the difficulty in identifying some of the catheters.

The Aurora EM system (Northern Digital, Inc., Waterloo, ON, Canada) was used in these experiments. The system consists of an FG, which generates a 50  50  50 cm3 tracking field, within which up to four separate EM sensors can be tracked simultaneously. In our experiments, we used three EM sensors: a flexible tracked guidewire to map the catheters, a customized EM sensor pair to track the TRUS probe, and a reference sensor to aid in combined EMeTRUS data acquisition. The details of the use of these sensors are described later.

642

S. Bharat et al. / Brachytherapy 13 (2014) 640e650

Robotic validation A three-axis robotic system was used to comprehensively test the accuracy of the EM system. The setup and data acquisition protocol are shown in Figs. 2a and 2b, respectively. Two EM sensors (the flexible guidewire and the reference sensor) were attached to the end of a robotic arm. The EM FG was positioned on the table, and the robot was programed to perform a 3D sweep of the field of view (FOV) of the FG. A 20  20  20 cm3 volume, centered in front of the FG with a minimum distance of about 8 cm, was tested. The volume comprises 729 equally spaced measurement points. At each spatial location, 100 EM samples were obtained. The performance of the EM system was quantified in the following ways. Absolute accuracy At each EM measurement position, the mean of 100 measurement points over 2.5 s (at 40 Hz) was acquired. The mean EM positions were then rigidly registered to the robotic positions. The absolute accuracy was defined as the root mean square (RMS) error of this registration. This parameter is representative of the average accuracy of an individual sensor. Relative accuracy The standard deviation of the mean position and orientation of the EM guidewire relative to that of the reference sensor was computed as a measure of relative accuracy. This is a relevant parameter to quantify because the EM

catheter tracks in our system are acquired with respect to the reference sensor (described in the subsection ‘‘Imagebased validation’’). Noise The standard deviation of the position and orientation measurement of the EM sensor at each position was averaged over all positions to quantify the noise of the system. Out-of-plane accuracy Catheter tracks were simulated using the robot, with multiple measurement points along the length of each track. The cumulative EM track length from a simulated grid plane was compared with robotic measurements, after registering the two modalities as described earlier. This represents a quantification of the ability of the EM system to estimate the tip of a catheter in a brachytherapy procedure. The robotic setup was also used to test the performance of the EM system in the presence of distorting equipment. A typical brachytherapy OR consists of different kinds of equipment in close proximity to the treatment table. A 20-inch liquid crystal display (LCD) monitor was used as an example of a distorter (Fig. 2c). The monitor was first placed at a distance of 30 cm from the center of the FG. Data were then obtained from the entire measurement volume, as described earlier. This procedure was repeated with the monitor at distances of 40, 50, 60, and 70 cm from the center of the FG. The trends in system accuracy as a function of location of the monitor provide useful preliminary

Fig. 2. (a) Experimental setup for testing the EM system using a three-stage robot. (b) The robotic data acquisition protocol. (c) Testing the impact of different positions of a 2000 liquid crystal display monitor on EM tracking accuracy. EM 5 electromagnetic; FG 5 field generator.

S. Bharat et al. / Brachytherapy 13 (2014) 640e650

643

Fig. 3. (a) The experimental phantom setup in a controlled ideal environment. (b) The flexible EM-tracked guidewire alongside the plastic treatment catheter. (c) Catheters inserted into the prostate model through the grid. (d) The experimental setup positioned on the treatment table in the OR (i.e., ‘‘low distortion’’ OR environment). (e) Mimicking a typical brachytherapy setup (i.e., distorting OR environment). EM 5 electromagnetic; TRUS 5 transrectal ultrasound; OR 5 operating room.

indicators of safe operating distances for equipment during clinical use of the EM system. Image-based validation In addition to the robotic validation of the EM system, catheter positions mapped using the EM system were compared with manually identified catheters from TRUS and CT images to benchmark the performance of the

EM system compared with standard clinical imaging modalities. Experimental setup The mapping experiments were performed on CIRS Model 053 ultrasound prostate training phantoms (CIRS, Inc., Norfolk, VA). As shown in Fig. 3a, the phantom was positioned in front of the brachytherapy grid/template

644

S. Bharat et al. / Brachytherapy 13 (2014) 640e650

(North American Scientific, Inc., Chatsworth, CA) with the rectal opening facing the grid to permit insertion of the TRUS probe. The TRUS probe (BK Medical, Peabody, MA) was mounted on a custom-built and EM-compatible stepper that allowed the manual translation of the probe in a direction perpendicular to the grid. The position of the TRUS probe was tracked using a pair of EM sensor coils attached externally to the probe. 3D TRUS image sets of the prostate model were reconstructed using tracked 2D images obtained by translating the TRUS probe through the phantom. Each point in the 3D TRUS image was obtained from the nearest points on the 2D TRUS images using distance-weighted interpolation. Catheter mapping was performed by retracting a 1.2 mm diameter flexible EMtracked guidewire (Northern Digital, Inc., Waterloo, ON, Canada), shown in Fig. 3b, through each implanted catheter (Fig. 3c shows the implanted catheters). A reference EM tracker (Northern Digital, Inc., Waterloo, ON, Canada) was attached to the side of the acrylic phantom holder. The EM FG was positioned alongside the rest of the apparatus so that all the EM-tracked devices (TRUS probe, guidewire [when located in an implanted catheter], and reference tracker) were in the FOV. The experiments were conducted both in a controlled laboratory environment (Fig. 3a) and in a brachytherapy OR. The OR experiments were carried out to determine the potential impact of the presence of clinical equipment around the tracked region. The OR experiments were performed on the treatment table. The FG was held above the tracked region of interest (i.e., the phantom) by a commercial FG holder (Northern Digital, Inc., Waterloo, ON, Canada). Data were acquired under two operating conditions: ‘‘low distortion’’ and distorting. For the tests corresponding to the ‘‘low distortion’’ environment (Fig. 3d), the phantom and stepper with the TRUS probe were positioned directly on the treatment table. No other equipment that could potentially distort the EM measurements was positioned in or near the EM FOV. Therefore, the tracking accuracy achieved in this configuration would represent the upper limit of accuracy in the OR. To simulate a potentially distorting environment, a typical brachytherapy procedural

setup was recreated (Fig. 3e). Leg stirrups and a stepper holder (in which the stepper was mounted) were attached to the treatment table. Other equipment such as the treatment planning computer, treatment afterloader, and instrument table were placed in their typical positions. Workflow The use of EM tracking technology for catheter position and shape determination was evaluated by comparing the EM catheter tracks with manually delineated catheter positions on TRUS and CT images. Five to six catheters were inserted through the grid to different depths in the prostate phantom. A TRUS cross section of the prostate phantom with five implanted catheters and the EM guidewire tip superimposed in real time on one of the catheters is shown in Fig. 4a. Figure 4b illustrates the superposition of EM tracks on a transverse TRUS image. EM catheter tracks were acquired by sequentially inserting the EM-tracked guidewire into each catheter. After insertion, the guidewire was retracted while simultaneously recording EM data from the sensor located at the tip of the guidewire. For manual catheter identification on TRUS images, the hyperechoic flash of each catheter was digitized on the axial images every few millimeters; reconstructed sagittal views were simultaneously available to inform the delineation of the catheters on the axial views. Both EM and TRUS data were obtained directly in the grid coordinate system CGrid, which is defined in the next subsection. The prostate phantom with the implanted catheters was then imaged in the CT scanner. The CT images were rigidly registered to the 3D TRUS image by registering the centroids of the segmented prostates in CT and 3D TRUS. The CTeTRUS registration enabled the combined interpretation of CT, US, and EM data in a common coordinate system (CGrid). Further manual refinements were carried out in some of the experiments to optimize the in-plane overlap of the catheter visualizations on the 3D TRUS and CT images. In one of the experiments, the CTeTRUS registration was repeated nine times, and the final registrations were compared to assess the uncertainty in the registration. The fact that TRUS visualizes the proximal catheter edge

Fig. 4. (a) Axial TRUS image of catheters with the real-time overlay of the EM guidewire tip on one of the catheters. (b) Recorded EM tracks superimposed (blue circles) on axial TRUS image. TRUS 5 transrectal ultrasound; EM 5 electromagnetic; US 5 ultrasound. (For interpretation of references to color in this figure legend, the reader is referred to the web version of this article.)

S. Bharat et al. / Brachytherapy 13 (2014) 640e650

645

Fig. 5. Schematic of EMeTRUS registration. EM 5 electromagnetic; TRUS 5 transrectal ultrasound.

was accounted for during the manual refinement process. Finally, each catheter in the CT image was manually delineated, and its coordinates transformed to CGrid. The results reported here represent data from 10 independent experiments, performed on multiple phantoms.

thereby replacing the positional encoders used in conventional steppers. The various transformations for all data recorded are defined below: pEMguidewire; Grid 5 T1 GridCalib  pEMguidewire;Ref

ð1Þ

EMeTRUS registration The system consists of imaging and tracking components that are interlinked to record all data in a common frame of reference. Figure 5 illustrates the various EM and TRUS system components, their coordinate frames of reference, and the transformations between these frames of reference. Each sensor represents its own coordinate system (Cxx), with a specific position and orientation relative to the EM FG or reference sensor. All EM-tracking data are transformed to the coordinate frame of a reference sensor (CRef) before conversion to the grid coordinate system (CGrid). A reference sensor is used to remove dependence of the coordinate system on the FG location, which is liable to be changed during the procedure. In clinical procedures, the reference sensor may be attached to the stepper because the stepper may move during the procedure. The transformation between the grid and the coordinate frame of the EM reference sensor (TGridCalib) is defined by manually pointing an EM-tracked tool at three or more points in the grid. TUScalib provides the transformation between the TRUS imaging plane (C2DTRUS) and the probe tracker (CProbe). This TRUS probe calibration is a one-time procedure, which involves manually picking the tip of an EMtracked needle on US images multiple times (performed in a water tank or in a phantom). Knowing the corresponding EM-reported positions of the needle tip, a rigid transformation between the two sets of data (US and EM) is derived (TUScalib) (19). Note that the TRUS probe is tracked using a pair of EM sensors attached externally to the probe,

1 p2DTRUS; Grid 5 T1 GridCalib  TProbe/Ref  TUSCalib  p2DTRUS

ð2Þ

pCT; Grid 5 TCT/3DTRUS  pCT :

ð3Þ

where pEMguidewire;Ref is the position of the EM guidewire tip in CRef, p2DTRUS is any point in the live 2D TRUS image coordinate system (C2DTRUS), and pCT is any point in the original CT coordinate space. TProbe/Ref is the EM tracking transformation of the probe sensor relative to the reference sensor, and TCT/3DTRUS is the registration transformation from the CT image to the 3D TRUS image. Also, p3DTRUS is any point in the 3D TRUS coordinate system, which is defined to be identical to CGrid. Calibration update The TRUS probe calibration (TUScalib) provides the primary conversion from the TRUS coordinate system to an EM frame of reference. If the position of the FG (with respect to the individual EM sensors) during TRUS probe calibration differs substantially from its position during the brachytherapy procedure, inhomogeneities in the tracking performance of the EM system may cause the calibration, and thus the EM mapping of the catheters to be inaccurate. This is especially the case if the calibration is performed in a water tank where the TRUS probe has to be vertically positioned, as opposed to its roughly horizontal configuration in brachytherapy procedures. Therefore,

646

S. Bharat et al. / Brachytherapy 13 (2014) 640e650

we define a framework to update the TRUS probe calibration intraprocedurally, based on the current position of the FG. A few catheters that are clearly visible on TRUS images are manually identified, and their corresponding EM tracks obtained. In our experiments, the use of at least four catheters yielded accurate calibrations. The calibration correction (TCorr) is obtained through a point-based registration of these two point sets in CGrid. Note that the correction is computed in the in-plane direction only (where in plane is defined to be parallel to the grid). Therefore, the updated TRUS probe calibration is calculated as TUSCalib;updated 5 TUSCalib;original  T1 Probe/Ref  TGridCalib 

T1 Corr

 T1 GridCalib

ð4Þ

 TProbe/Ref Accuracy analysis For the in-plane accuracy evaluation, the catheter positions on all modalities (EM, TRUS, and CT) were interpolated along the craniocaudal (Z) direction in increments of 0.5 mm. Approximately 100 comparison points were used per catheter, thereby resulting in approximately 5000 comparison points for all the experiments combined. The in-plane error was calculated as the Euclidean distance between the EM-tracked points and their corresponding CT or TRUS points on the same slice. The out-of-plane (craniocaudal direction or perpendicular to the grid) error was computed by comparing the catheter tips in the different modalities.

Results Tables 1e3 list the absolute accuracy, relative accuracy, and noise, respectively, of the EM system, based on the robotic tests. For the reference sensor, the mean absolute accuracy was 0.23 mm, over the entire spatial range tested. The mean RMS error was 0.25 mm. The performance of the flexible guidewire was comparable, with mean absolute and RMS accuracies of 0.26 and 0.31 mm, respectively (Table 1). The accuracy of the guidewire relative to the reference sensor was 0.19 mm (position) and 0.1 (orientation) (Table 2). The measured noise of both the guidewire Table 1 Average accuracy values of individual EM sensors obtained using the robotic system Absolute accuracy Absolute distance (mm) Configuration

Mean

Maximum

Standard deviation

RMS distance (mm)

Reference vs. robot Guidewire vs. robot

0.23 0.26

0.82 0.98

0.11 0.16

0.25 0.31

EM 5 electromagnetic; RMS 5 root mean square. The values represent the distances between robotic and EM-measured positions after registration of the EM point set onto the robotic point set.

Table 2 Average relative accuracy of the guidewire with respect to the reference sensor obtained using the robotic system Relative accuracy Translation (mm)

Orientation (degrees)

Configuration

Mean

Maximum

Mean

Maximum

Guidewire vs. reference

0.19

1.45

0.1

0.5

The values represent the standard deviation of the electromagnetic position and orientation measurement of the guidewire relative to the reference sensor for all robotic measurement positions.

and reference sensor was !0.1 mm (Table 3). The out-ofplane accuracy of the flexible guidewire, which represents the ability of the EM system to estimate the location of the tip of a catheter, was 0.61 mm. Figure 6 illustrates the performance of the EM system, as a function of location of the distorting equipment (2000 LCD monitor, in this case). The RMS error fell below 0.5 mm, when the monitor was placed at least 50 cm from the center of the FG. Note that for distances approaching 30 cm, the error was greater than 2 mm, whereas for a distance of 70 cm, EM system performance was similar to that in a ‘‘low distortion’’ environment (without distorting equipment in the vicinity). Table 4 illustrates the in-plane and out-of-plane accuracy of our EM system in phantom experiments for data acquired in the controlled laboratory environment. These results were achieved using a TRUS probe calibration (TUScalib) computed preprocedurally in a 25  20  20 cm water tank. The EM-estimated catheter tracks were in agreement with the TRUS-identified tracks to an in-plane accuracy of !1.5 mm. The out-of-plane error was !0.5 mm. CT provided an additional yardstick for comparison and yielded accuracy values in line with the EMeTRUS comparison (mean EMeCT error !1.5 mm in plane and !0.5 mm out of plane). Note that the comparison with CT is impacted by the registration error of the CT images to the TRUS coordinate system. Care was taken to optimize this registration throughout the prostate phantom volume and to center the Table 3 Temporal noise of individual EM sensors obtained using the robotic system Temporal noise Position (mm)

Orientation (quaternions)

Sensor

X

Y

Z

Qw

Qx

Qy

Qz

Reference Reference (maximum) Guidewire Guidewire (maximum)

0.02 0.09

0.02 0.08

0.02 0.13

0.0001 0.0003

0 0.0001

0 0.0001

0.0001 0.0002

0.04 0.27

0.04 0.28

0.05 0.30

0.0001 0.0004

0.0001 0.0004

0.0001 0.0005

0 0

EM 5 electromagnetic. The values represent the average standard deviation from the mean sensor reading at each robotic measurement position. The maximum deviations are also reported.

S. Bharat et al. / Brachytherapy 13 (2014) 640e650

647

Table 5 Improvement in accuracy afforded by the updated TRUS probe calibration In-plane discrepancy (mm) EM vs. TRUS

Mean

Standard deviation

Maximum

TUSCalib,updated obtained with FG at 14 cm distance from the target (configuration shown in Fig. 2a) FG at 23 cm distance FG above target and angled at 23 cm distance

0.6

0.3

1.6

0.7 0.6

0.3 0.3

1.7 2.6

TRUS 5 transrectal ultrasound; EM 5 electromagnetic; FG 5 field generator.

Fig. 6. Trends in EM system accuracy as a function of monitor location. The blue and green plots represent the absolute RMS error of the reference sensor and guidewire, respectively. The red and cyan plots represent the error in relative position and orientation, respectively, of the reference sensor with respect to the guidewire. EM 5 electromagnetic; RMS 5 root mean square; FG 5 field generator. (For interpretation of references to color in this figure legend, the reader is referred to the web version of this article.)

CT and TRUS representations of the prostate, thus minimizing the impact of potential sound speed mismatches in the TRUS image. Several repeat registrations were carried out and agreed with one another to within 1 mm, suggesting that the uncertainty in the EMeCT error is also less than 1 mm. The time taken to map each catheter using the EM guidewire was ~5e10 s, which is essentially the time required to insert and retract the EM guidewire from the catheter. To avoid errors introduced by variation in the position and orientation of the FG during the calibration and mapping procedures, the calibration was updated during the mapping procedure. A subset of the catheters clearly visible on TRUS images were identified and used with their corresponding EM tracks to compute a correction function. The updated calibration was then calculated using Eq. (4). The validity of the updated calibration for different FG positions was also tested. Table 5 illustrates these findings. Figure 7

illustrates reconstructed EM (blue) and TRUS (green) catheter tracks, using the updated TRUS probe calibration. The concept of updating the TRUS probe calibration to finetune the system for optimal performance during brachytherapy was found to be very useful and further increased the EMeTRUS accuracy to !1 mm. Once updated for a given position of the FG, the probe calibration was found to result in reasonable accuracy (!1 mm) even at other positions of the FG, as evidenced by the results shown in Table 5. Table 6 illustrates the results from experiments in a brachytherapy OR. The TRUS probe calibration (TUScalib) used in these experiments was performed in a CIRS phantom on the treatment table in the OR. The mean EMe TRUS in-plane accuracy in a ‘‘low distortion’’ OR environment was found to be !0.5 mm, when using a calibration performed in a ‘‘low distortion’’ environment. This error increased to ~1 mm when this calibration was used to perform tracking in a distorting environment. Here, the calibration update functionality helped improve the accuracy to !0.5 mm. When a calibration obtained in a distorting environment was tested in a distorting environment, the mean accuracy was !1 mm. The craniocaudal accuracy in all cases was in the range of 0.5e1 mm. We also tested the

Table 4 Performance accuracy of EM catheter mapping system in a controlled ideal environment In-plane discrepancy (mm) Modalities compared

Out-of-plane discrepancy (mm)

Standard Standard Mean deviation Maximum Mean deviation Maximum

EM vs. TRUS 1.3 EM vs. CT 1.2 TRUS vs. CT 0.6

0.5 0.4 0.3

2.4 2.4 1.6

0.1 0.3 0.2

0.1 0.1 0.2

EM 5 electromagnetic; TRUS 5 transrectal ultrasound.

0.2 0.6 0.5

Fig. 7. Reconstructed catheter tracks obtained using TUSCalib, updated. The blue tracks are estimated using EM tracking, and the shorter green tracks are manually identified in TRUS images. The mean in-plane EMeTRUS error is 0.65 mm. EM 5 electromagnetic; TRUS 5 transrectal ultrasound. (For interpretation of references to color in this figure legend, the reader is referred to the web version of this article.)

648

S. Bharat et al. / Brachytherapy 13 (2014) 640e650

Table 6 Performance accuracy of EM catheter mapping system in the OR environment In-plane discrepancy (mm)

Out-of-plane discrepancy (mm)

EM vs. TRUS (in OR)

Mean

Standard deviation

Maximum

Mean

Standard deviation

Maximum

Calibrating and testing in a ‘‘low distortion’’ environment Calibrating in a ‘‘low distortion’’ environment and testing in distorting environment Calibrating in a ‘‘low distortion’’ environment and testing in distorting environment, after updating the calibration in the distorting environment Calibrating and testing in distorting environment

0.4 1 0.3

0.2 0.2 0.2

1 1.5 1.1

0.8 1.1 1.1

0.4 0.8 0.8

1.7 1.9 1.9

0.6

0.3

1.5

0.4

0.2

0.6

EM 5 electromagnetic; OR 5 operating room; TRUS 5 transrectal ultrasound. In the OR, a ‘‘low distortion’’ environment refers to the absence of clinical equipment other than the treatment table and the ultrasound system near the testing area. A distorting environment refers to an environment that includes the presence of other clinically used brachytherapy equipment, such as leg stirrups, stepper holder, treatment planning computer, treatment afterloader, instrument table, and others, in their typical positions.

impact of variation in positions of the FG and the surrounding clinical equipment (leg stirrups, stepper holder, instrument table, etc.). The mean accuracy remained !1 mm for different configurations of the equipment (Table 7).

Discussion CT and TRUS represent the two most common modalities used clinically in HDR brachytherapy for organ delineation and catheter identification. TRUS provides better visualization of the prostate and urethra, whereas CT provides unmatched clarity in visualizing the catheters (5). However, because of the disadvantages associated with CT (such as the need to transport the patient from the OR to the CT scanner, resulting workflow inefficiencies, and possibility of catheter displacement (11)), a TRUS-only workflow is gaining increasing attention in clinical practice. In a TRUS-only workflow, both organ segmentation and catheter identification are performed on TRUS images. Although the clinical efficiency associated with a TRUSonly workflow is higher compared with a CT-TRUS workflow, manual catheter identification based on TRUS images can be time consuming, challenging, and operator dependent because of shadowing artifacts from posterior Table 7 Testing the impact of variation in position of the FG and other clinical equipment (e.g., leg stirrups, stepper holder, instrument table, etc.) In-plane discrepancy (mm) EM vs. TRUS

Mean

Standard deviation

Maximum

Calibrating and testing in distorting environment (Scenario A) Testing above calibration in distorting environment (Scenario B)

0.4

0.2

1.1

0.8

0.3

1.6

FG 5 field generator; EM 5 electromagnetic; TRUS 5 transrectal ultrasound. Scenario A represents a certain configuration of the FG and other equipment, whereas Scenario B represents another configuration of the equipment (with changes in position and orientation). The calibration used in both cases was obtained in Scenario A.

catheters in US images. In this article, we present a solution that maintains the clinical workflow benefits provided by a TRUS-only workflow (i.e., no CT required), while at the same time removing the subjectivity and reducing the time associated with the TRUS-based catheter identification process. The use of EM tracking technology to map the position and shape of catheters has been reported here. Comprehensive robotic validation of the EM system in a distortionfree environment yielded operational system accuracy of ! 0.5 mm. This provided a good baseline for subsequent testing of the EM system on catheters implanted in a phantom. The data for the phantom study were acquired in three configurations: an ideal controlled laboratory environment devoid of any EM-distorting equipment, a relatively ‘‘low distortion’’ OR environment (i.e., on the treatment table without the presence of other potentially distorting clinical equipment), and a typical brachytherapy OR environment (i.e., all clinical equipment in place as in a typical brachytherapy procedure) that can potentially distort the EM measurements. In response to the weak magnetic field generated by the EM FG, eddy currents can be generated in metal that is near the FOV. Because EM position tracking works by analyzing the amplitude of the weak electric current that is generated in the EM tool, the generation of eddy currents in surrounding equipment can potentially lead to an inaccurate reconstruction of the position of the EM tool. Our experiments in the brachytherapy OR show that the equipment configurations tested had a minimal impact on EM system accuracy (accuracy remained !1 mm) and thereby illustrate the robustness of our EM system in a typical brachytherapy environment. The initial OR results are, therefore, encouraging and provide early indications that the use of EM technology in a clinical setting is feasible. Prior work by Zhou et al. (18) has shown that EM system performance is highly dependent on system configuration and the locations of surrounding equipment. To operate close to ideal system accuracy levels, it is essential to achieve a robust system calibration that is valid and applicable in a clinical environment consisting of potential EM distorters. In our case, system calibration primarily refers

S. Bharat et al. / Brachytherapy 13 (2014) 640e650

to the calibration of the TRUS probe. This calibration relates the TRUS image space to the EM tracking space, thus allowing visualization of the EM catheter tracks on the underlying TRUS images. This is important to eventually enable clinical use of EM tracking technology in HDR brachytherapy. We describe an approach to calibration that can account for potential distortions in the clinical environment. The TRUS probe calibration can be performed in a water tank or in a phantom. A water tank offers flexibility and easy access for calibrating multiple points over a large FOV. However, it usually involves a change in relative geometry between the individual EM sensors and the FG as well as between the calibration and the mapping procedure. This leads to reduced accuracy in the mapping procedure. The in-plane accuracy can be restored to higher values by updating the calibration intraprocedurally using a subset of the implanted catheters. The intraprocedural calibration update has been shown to improve the accuracy to !0.5 mm. Alternatively, the initial calibration may be performed in a phantom (instead of a water tank). The advantage of using a phantom is that the calibration geometry matches or is very close to the treatment geometry. In our experiments using a calibration obtained in a phantom, the accuracy was ! 0.5 mm, thereby obviating the need for an intraprocedural calibration update. In addition, calibration phantoms similar to the one introduced by Chen et al. (20) can be used for automatic, easy, and accurate calibration of the probe before the brachytherapy procedure. From our experiments, three factors related to the calibration workflow were found to ensure optimal system performance: (1) maintenance of similar sensor/FG geometry during the tracking procedure as that during the calibration, (2) calibration in an environment similar to the clinical environment, and (3) intraprocedural calibration update to account for any remaining error. If the calibration geometry used is the same as that during the subsequent tracking procedure and the positions of equipment during calibration and tracking are largely unchanged, the need for intraprocedurally updating the calibration should be minimal. Conversely, if previously mentioned factors (1) and/or (2) are not satisfied, the calibration may need to be updated intraprocedurally to yield high system accuracy. Therefore, our workflow can account and correct for different situations that may otherwise reduce system accuracy. Note that adhering to the aforementioned factors can account for systematic error at higher distances from the FG. However, the random error component (noise) will still be a factor. Smoothing techniques are contemplated to deal with random errors. The experiments conducted in the OR showed that typical brachytherapy equipment configurations did not yield significant EM distortions. When aforementioned conditions (1) and (2) were satisfied, the in-plane accuracy was !0.5 mm and the craniocaudal accuracy was !1 mm. Importantly, an intraprocedural calibration update was not required, thereby further simplifying the workflow. Last,

649

we tested the impact of interprocedural changes in clinical equipment positions (i.e., to simulate using a given TRUS probe calibration for tracking procedures on different patients). The accuracy reduced slightly (from !0.5 to ! 1 mm) when the surrounding equipment was moved around. If the initial accuracy in a given procedure is deemed unacceptable by the clinician, the calibration update functionality may be used to improve the accuracy for that procedure. The robotic setup also provided a systematic way to determine safe operating distances for surrounding equipment, from the perspective of fidelity of the recorded EM data. Initial tests using a 2000 LCD monitor showed that distances greater than 60 cm (from the center of the FG) may be sufficient in ensuring high accuracy in EM system performance. LCD monitors are known to contain metal that can distort EM measurements. Hence, it was chosen as a sample distorter for the preliminary experiments reported here. The results presented here provide early evidence supporting the use of EM tracking technology in the brachytherapy framework. Traditionally, spatial accuracy of ~2e3 mm has been found to be clinically acceptable (5, 21). The high spatial accuracy of the EM catheter positions relative to currently used imaging modalities (TRUS and CT) provides the first proof point that EM tracking has the potential to improve the accuracy of HDR brachytherapy procedures. High in-plane spatial accuracy ensures conformity of dose delivery in and around the prostate. This is important because the bladder abuts the prostate anteriorly and the rectal wall borders the prostate posteriorly, with the urethra situated centrally in the prostate gland. Craniocaudal positional accuracy is also critical to avoid suboptimal dose coverage at the prostate base and apex. In addition to being highly accurate, the shortened acquisition time of catheter tracks (~5e10 s per catheter) illustrates the improved workflow efficiency associated with the proposed catheter mapping technique. Assuming an accurate system calibration, this amounts to digitizing all (~10e20) catheters in less than 3 min. If the system calibration needs to be updated intraprocedurally, the time taken to digitize a subset of catheters on TRUS images must also be considered. In both cases, however, significant time savings may be achieved over conventional manual catheter identification. Robotic testing of the EM system showed the EM system to be very accurate (!0.5 mm)dthese tests were independent of other imaging modalities (e.g., US, CT) and therefore, the results do not include errors, such as the variability in identifying catheter tracks on US, registration of CT to EM space, and out-of-plane errors resulting from finite CT slice thickness.

Conclusion An EM system customized for use in HDR brachytherapy is presented here. Robotic validation of the EM system

650

S. Bharat et al. / Brachytherapy 13 (2014) 640e650

found the EM system to be accurate to less than 0.5 mm. The high level of agreement (!1 mm) between EMtracked and US-identified catheter positions further illustrates the potential value of using EM tracking for catheter mapping in HDR brachytherapy by demonstrating equivalence of the proposed EM technique to existing clinical standards of catheter identification. The results of the OR experiments are encouraging and show that typical brachytherapy equipment configurations do not yield significant EM distortions. To summarize, the accuracy obtained during the tracking procedure is directly impacted by the following factors: (1) the inherent accuracy of the TRUS probe calibration, (2) the ability to recreate the calibration geometry (of the tracked devices with respect to the FG) during the tracking procedure, (3) the locations of surrounding clinical equipment, and (4) an intraprocedural calibration update, if necessary. The ability to control the aforementioned factors during brachytherapy procedures will determine the achievable tracking accuracy. The shortened acquisition time of catheter tracks illustrates the improved workflow efficiency associated with the proposed catheter mapping technique. Future work includes retrospective spatial and dosimetric analysis on EM data acquired during clinical brachytherapy procedures.

Acknowledgments We thank our machinists Ray Shoreland and Flavius Rotar for their contributions to the work presented here. We would also like to acknowledge the support of the Ontario Consortium for Adaptive Interventions in Radiation Oncology (OCAIRO). References [1] Pellizzon ACA, Salvajoli JV, Maia MC, et al. Late urinary morbidity with high-dose-rate prostate brachytherapy as a boost to conventional external beam radiation therapy for local and locally advanced prostate cancer. J Urol 2004;171:1105e1108. [2] Hsu IC, Bae K, Shinohara K, et al. Phase II trial of combined high-doserate brachytherapy and external beam radiotherapy for adenocarcinoma of the prostate: Preliminary results of RTOG 0321. Int J Radiat Oncol Biol Phys 2010;78:751e758. [3] Hsu IC, Hunt D, Straube W, et al. Dosimetric analysis of radiation therapy oncology group 0321: The importance of urethral dose. Pract Radiat Oncol 2014;4:27e34. [4] Cunha JA, Pouliot J, Weinberg V, et al. Urethra low-dose tunnels: Validation of and class solution for generating urethra-sparing dose plans using inverse planning simulated annealing for prostate highdose-rate brachytherapy. Brachytherapy 2012;11:348e353.

[5] Schmid M, Crook JM, Batchelar D, et al. A phantom study to assess accuracy of needle identification in real-time planning of ultrasoundguided high-dose-rate prostate implants. Brachytherapy 2013;12: 56e64. [6] Yamada Y, Rogers L, Demanes DJ, et al. American Brachytherapy Society consensus guidelines for high-dose-rate prostate brachytherapy. Brachytherapy 2012;11:20e32. [7] Buzurovic I, Misic V, Yu Y. Needle identification in high-dose-rate prostate brachytherapy using ultrasound imaging modality. Conf Proc IEEE Eng Med Biol Soc 2012;476e479. [8] Kim Y, Hsu J, Pouliot J. Measurement of craniocaudal catheter displacement between fractions in computed tomography-based high-dose-rate brachytherapy of prostate cancer. J Appl Clin Med Phys 2007;8:2415e2427. [9] Menard C, Susil RC, Choyke PL, et al. MRI-guided HDR prostate brachytherapy in standard 1.5T scanner. Int J Radiat Oncol Biol Phys 2004;59:1414e1423. [10] Lessard E, Pouliot J. Inverse planning anatomy-based dose optimization for HDR-brachytherapy of the prostate using fast simulated annealing algorithm and dedicated objective function. Med Phys 2001; 28:773e779. [11] Holly R, Morton GC, Sankreacha R, et al. Use of cone-beam imaging to correct for catheter displacement in high-dose-rate prostate brachytherapy. Brachytherapy 2011;10:299e305. [12] Ghose S, Oliver A, Marti R, et al. A survey of prostate segmentation methodologies in ultrasound, magnetic resonance and computed tomography images. Comput Methods Programs Biomed 2012;108: 262e287. [13] Wood BJ, Zhang H, Durrani A, et al. Navigation with electromagnetic tracking for interventional radiology procedures: A feasibility study. J Vasc Interv Radiol 2005;16:493e505. [14] Kruecker J, Xu S, Glossop ND, et al. Electromagnetic tracking for thermal ablation and biopsy guidance: Clinical evaluation of spatial accuracy. J Vasc Interv Radiol 2007;18:1141e1150. [15] Kruecker J, Xu S, Venkatesan A, et al. Clinical utility of real-time fusion guidance for biopsy and ablation. J Vasc Interv Radiol 2011; 22:515e524. [16] Pinto P, Chung PH, Rastinehad AR, et al. Magnetic resonance imaging/ultrasound fusion guided prostate biopsy improves cancer detection following transrectal ultrasound biopsy and correlates with multiparametric magnetic resonance imaging. Urology 2011;186: 1281e1285. [17] Zhu M, Bharat S, Michalski JM, et al. Adaptive radiation therapy for postprostatectomy patients using real-time electromagnetic target motion tracking during external beam radiation therapy. Int J Radiat Oncol Biol Phys 2013;85:1038e1044. [18] Zhou J, Sebastian E, Mangona V, et al. Real-time catheter tracking for high-dose-rate prostate brachytherapy using an electromagnetic 3D-guidance device: A preliminary performance study. Med Phys 2013;40:0217161e0217168. [19] Zhang H, Banovac F, White A, et al. Freehand 3D ultrasound calibration using an electromagnetically tracked needle. SPIE Proceedings, Volume 6141, San Diego, CA, 2006. [20] Chen TK, Heffter T, Lasso A, et al. Automated intraoperative calibration for prostate cancer brachytherapy. Med Phys 2011;38:6285e6299. [21] Tiong A, Bydder S, Ebert M, et al. A small tolerance for catheter displacement in high-dose rate prostate brachytherapy is necessary and feasible. Int J Radiat Oncol Biol Phys 2010;76:1066e1072.