Electromagnetic navigation improves minimally invasive robot-assisted lung brachytherapy

Abstract

Objective: Recent advances in minimally invasive thoracic surgery have renewed an interest in the role of interstitial brachytherapy for lung cancer. Our previous work has demonstrated that a minimally invasive robot-assisted (MIRA) lung brachytherapy system produced results that were equal to or better than those obtained with standard video-assisted thoracic surgery (VATS) and comparable to results with open surgery. The purpose of this project was to evaluate the performance of an integrated system for MIRA lung brachytherapy that incorporated modified electromagnetic navigation and ultrasound image guidance with robotic assistance.

Methods: The experimental test-bed consisted of a VATS box, ZEUS® and AESOP® surgical robotic arms, a seed injector, an ultrasound machine, video monitors, a computer, and an endoscope. Our previous custom-designed electromagnetic navigational software and the robotic controller were modified and incorporated into the MIRA III system to become the next-generation MIRA IV. Inactive brachytherapy seeds were injected as close as possible to a small metal ball target embedded in an opaque agar cube. The completion time, the number of attempts, and the accuracy of seed deployment were compared for manual placement, standard VATS, MIRA III, and the new MIRA IV system.

Results: The MIRA IV system significantly reduced the median procedure time by 61% (104 s to 41 s), tissue trauma by 75% (4 attempts to 1 attempt), and mean seed placement error by 64% (2.5 mm to 0.9 mm) when compared to a standard VATS. MIRA IV also reduced the mean procedure time by 48% (85 s to 44 s) and the seed placement error by 68% (2.8 mm to 0.9 mm) compared to the MIRA III system.

Conclusions: A modified integrated system for performing minimally invasive robot-assisted lung brachytherapy was developed that incorporated electromagnetic navigation and an improved robotic controller. The MIRA IV system performed significantly better than standard VATS and better than MIRA III.

• Minimally invasive surgery
• robot-assisted surgery
• lung brachytherapy
• image guidance
• electromagnetic tracking
• electromagnetic navigation

Introduction

Surgical resection is the treatment of choice for early stage non-small cell lung cancer, however, only a third of patients who present are eligible for a curative resection due to limited pulmonary reserve and/or other medical contraindications [1]. Interstitial brachytherapy is an alternative form of radiation treatment that places a radioactive source near or into a tumor. The delivery of radiation is more precise, and larger doses can be concentrated into the tumor while minimizing damage to normal tissue [2], [3].

Minimally invasive surgery (MIS) employs instruments inserted through small incisions in the patient's chest or abdomen to remove cancers. The procedure relies on video images for guidance and on the surgeon for instrument manipulation. The minimally invasive approach reduces postoperative pain and length of hospital stay, and facilitates an earlier return to normal activity. There are, however, some limitations with this technique, such as impaired vision due to the two-dimensional (2D) image display, restricted maneuverability of the instruments, and a long learning curve. Surgical robots have been shown to vastly improve MIS efficiency by providing superior three-dimensional (3D) magnification, superior dexterity with 7 degrees of freedom, and improved precision of dissection with tremor filtration and motion scaling [4–7]. Furthermore, robots allow the surgeon to take advantage of improved ergonomics and a shorter learning curve. Surgical robots facilitate not only the surgical removal of cancer using MIS techniques, but potentially may be integrated with image guidance for more precise, consistent and safe implantation of low-dose radiation brachytherapy seeds with reduced patient morbidity. Only robots can allow the healthcare team to implant brachytherapy interstitial seeds at a safe distance from the radioactive source. In order to perform lung brachytherapy minimally invasively when the tumor is not visible with the video thoracoscope, accurate image guidance is needed [8]. Intra-operative ultrasound is our imaging modality of choice because of its real-time imaging capabilities, wide availability, portability, and safety.

An integrated surgical guidance system for minimally invasive robot-assisted (MIRA) lung brachytherapy procedures is currently being developed at Canadian Surgical Technologies & Advanced Robotics (CSTAR) [9–11]. The initial system (MIRA I) consisted of the ZEUS® system, the ATL HDI 5000 ultrasound system with the ATL Lap L9-5 transducer, and a hydraulic seed injection system [9]. The MIRA I system performed well in in vivo animal testing and was able to remotely manipulate the ultrasound transducer and the needle to allow deployment of the seeds. Ultrasound images of the lung were of good quality. A customized seed injection instrument that holds the needle and deploys the seeds was added and tested in vitro using transparent agar cubes with visible steel spheres as targets [10], [12]. These experiments with the MIRA II system suffered from a complete lack of depth perception due to the 2D video images.

The next-generation MIRA III system included the ZEUS® system, the seed injector, the Phillips iU-22 ultrasound system with the C-9 transducer, and the microBIRD™ electromagnetic tracking system [11]. Image guidance was achieved using custom software for “hidden” targets called InterNAV™ [13], [14]. This software calculates the trajectory of the needle and displays it relative to the ultrasound image. The user can then select a target on the ultrasound image and adjust the orientation of the needle according to navigational instructions given by the software. Although significant improvements were possible, the system lacked intuitiveness, and the accuracy with the robotic system was limited by the AESOP® robot controls.

We continued with the development of the MIRA system to produce the MIRA IV. We hypothesized that improvements in the EM tracking system, the robotic controller, and the guidance software would improve performance by reducing tissue trauma, shortening surgical time and improving the accuracy of seed placement when compared to the VATS system and the older MIRA III system.

Materials

The hardware components of the MIRA IV system consisted of an ultrasound probe (C9-4 probe on a Phillips iU22 ultrasound system); a standard brachytherapy needle and the aforementioned brachytherapy seed injector [12]; two 6-degree-of-freedom (6-DOF) electromagnetic sensors and an electromagnetic field transmitter (microBIRD™, Ascension Technology) used to track the positions and orientations of the ultrasound plane and the needle tip; an endoscopic camera; a computer; and two surgical robotic arms, one for maneuvering the ultrasound probe (AESOP®), and the other for holding the brachytherapy seed injector (ZEUS®) (Figure 1) [14].

Figure 1. Experimental test-bed of MIRA IV at CSTAR.

The original software interface, InterNAV™ 1.0, used ultrasound images and position information of the instruments to recreate a virtual workspace that provided intuitive guidance [13]. Electromagnetic trackers were placed on the needle and the ultrasound probe to provide real-time position and orientation feedback to the software. The software calculated the trajectory of the needle and displayed it relative to the ultrasound image. The user then selected a target on the ultrasound image and adjusted the orientation of the needle according to navigational cues. It was found that the discrete step motions of the AESOP® robot using the original controller limited the achievable positioning accuracy. To overcome this problem, controller hardware was modified to receive analog input signals directly from a personal computer instead of via manual key presses from a pendant [15]. These modifications allowed superior control of the robot motions by controlling the pulse magnitude and width of the analog command signals. The InterNAV™ navigational software was also modified to accommodate these changes and the user was now able to command the robot from control keys integrated into the navigational software. An additional modification to the InterNAV™ software incorporated the direct control of the seed insertion instrument from the same interface.

The original InterNAV™ contained five views – US, World, 3D Position and two 2D Position views (Figure 2a). To obtain 3D information, the user re-orients the camera to different vantage points. With this interface, the user was required to repeatedly switch views in order to re-orient the needle, and three other views were needed to depict distance from the target. This increased task completion time. Task completion time was also increased by having to use both the mouse and a pendant for robotic control, as well as having to change programs to drop the seed. Placement accuracy was significantly affected by the discrete steps in AESOP® arm motion: During the experiments, the desired position of the instrument often fell between two consecutive steps of the arm.

Figure 2. Original InterNAV™ graphical user interface (a) and the modified InterNAV2.0™ (b). The main modifications include the following: The camera is oriented in line with the needle axis (for easier usage of the software); the control panel provides direct control of the robot motion and dropping of the seeds (for finer control over robot motions); a depth display (for increased penetration accuracy) has been introduced; and arrow guides are provided (for increased orientation accuracy). [Color version available online.]

The needle, ultrasound probe, and camera were inserted through openings in the VATS box containing agar-gelatine cubes to simulate lung tissue. Two ZEUS® arms held the ultrasound probe and the camera. Our seed injector [12] that held the needle and deployed the seeds was attached to the AESOP® arm. Reduced direct visibility hampered seed placement accuracy. To address this, electromagnetic sensors were attached to the needle and the ultrasound probe. InterNAV™ [13] incorporated information from the electromagnetic trackers and ultrasound probe to position the needle at the targeted location.

To improve software intuitiveness, the navigation model was modified such that the viewpoint aligned with the needle axis, thereby shifting from a third-person to a first-person view. This collapsed what were four views into just one (Figure 2 b). If the target could not be seen in the world view due to the needle pointing in the wrong direction, the arrow guides displayed at the edges of the screen indicated the direction in which to move in order to align with the target. Since a 6-DOF sensor was used to track needle tip position, the orientation about the needle axis also had an effect on the camera view. To eliminate this effect and always keep the world plane horizontal, the camera view needed to be made spin-invariant. To achieve this, the first step consisted of recording the position of the needle, moving the needle up, re-recording the position of the needle, and obtaining the normalized difference between the two recorded positions. This step was then repeated moving the needle sideways. To overcome restricted camera control, two features were implemented. The first performed a bounding box ray trace to determine if the target was in the current view; if not, an arrow guide pointing to the direction in which the needle ought to move was displayed. The second consisted of zooming the view out during the orientation phase and in during the penetration phase.

Figure 3 shows a flow chart of component interaction in the MIRA IV system with InterNAV2.0™ software.

Figure 3. Flow chart of component interaction in the MIRA IV with InterNAV2.0™ system. [Color version available online.]

Methods

A single 1.6-mm diameter steel ball was placed in a random position inside an opaque phantom cube (1.7% gellan gum powder and 0.2% cellulose particles, Sigma-Aldrich) to serve as the target for seed injection (Figure 4) [16]. Inactive brachytherapy seeds that measured 0.8 × 5 mm were injected into the cubes using an 18-gauge brachytherapy needle (Cook Urological, Canada). The five different experimental setups were: (1) manual, a setup simulating open thoracic deposition of brachytherapy seeds using ultrasound guidance; (2) standard VATS – video-assisted thoracic surgery using ultrasound guidance; (3) MIRA II, a setup similar to standard VATS except that manual needle insertion is replaced with robotic needle insertion; (4) MIRA III, an integrated system guided by InterNAV™; and (5) MIRA IV, an integrated system guided by InterNAV2.0™ (Table I).

Figure 4. The opaque agar cube (4.5 × 4.5 × 4.5 cm) used to simulate lung tissues beside the brachytherapy seed cylinder (0.8 × 5 mm) and the stainless steel bead (1.6 mm in diameter) used as a target.

Table I.  The five experimental setups.

Setup	Simulated operating environment	Guidance	Deposition
Manual	Open thoracotomy	Ultrasound	Manual
Standard VATS	MIS operation	Ultrasound	Manual
MIRA II	MIS operation	Ultrasound	Robot-assisted
MIRA III	MIS operation	Ultrasound + InterNAV™	Robot-assisted
MIRA IV	MIS operation	Ultrasound + InterNAV2.0™	Robot-assisted

The manual setup simulated an open thoracotomy for brachytherapy seed deposition using a needle and was performed as a reference of comparison for the four minimally invasive procedures. Although this type of surgery is very invasive for the patient, it is the ideal condition for implanting brachytherapy seeds from the surgeon's point of view because it allows direct visualization and completely unrestricted access to the lung tissue. This surgical environment was simulated by removing the walls of the VATS box. In true minimally invasive surgeries, such as VATS, the brachytherapy seeds are delivered through an access port on the chest wall, either directly with the operator holding the needle, or indirectly with the robotic seed injector. In both the standard VATS and the robot-assisted VATS surgeries, the operator relied on ultrasound images for pre-operative and intra-operative guidance of needle insertion. The user approximated the position of the ultrasound plane by observing the location of the probe from the video display and then mentally estimated the position of the target within the image from knowledge of the depth of the ultrasound. The MIRA surgeries integrated navigational guidance and robotic assistance in the setting of a minimally invasive procedure.

Four novice subjects each performed approximately 20 trials for each of the experimental setups. They received initial training before performing the actual experiments, until they felt confident about the task, and each user performed 6 trials with each setup prior to the start of the experiment. The experiments were done in random order to preclude the effect of training. The users were instructed to inject the brachytherapy seed as close as possible to the target, with no limit on the amount of time or the number of attempts allowed. If the seed was within 2 mm of the target it was felt to be acceptable. The accuracy was computed from three orthogonal radiographs of the agar phantom by measuring the 2D center-to-center distance from the target to the seed on each image and then calculating the true distance using the Pythagorean Theorem as previously described [14]. Since this distance would include the radii of the target and the seed, a perfect outcome would give a value of 1.18 mm. This number was subtracted from the distances because it was felt to be an unnecessary distraction from the true meaning of error. The error rate would be the smallest in the case where the seed lay adjacent to the target bead. However, in cases where the needle was inserted such that a head-on collision with the bead occurred, it became physically impossible to achieve the best-case scenario. Hence, our accuracy measure gives a conservative measure of the error in needle placement.

Statistical analysis

A preliminary sample size was calculated after 12 pilot experiments and showed that it would require 100 trials for each setup to discern a 30-second difference in the time needed to insert a single seed with 80% power and a type I error rate of 0.05 [17]. After 80 trials, our calculations found sufficient power to detect this difference. The Statistical Package for Social Sciences software (SPSS, Chicago, IL), version 15.0 for Windows, was used for statistical analysis. A one-way analysis of variance (ANOVA) was performed to establish differences among the accuracies of the different methods. Unpaired t-tests with unequal variances were then used for comparison between individual setups. Due to the skewed nature of the total time and the number of attempts, the non-parametric Kruskal-Wallis test was used to ascertain significant differences among the setups, followed by the Mann-Whitney test for individual comparisons.

Results

A summary of the performance (measured with respect to time, number of attempts and accuracy) is provided in Table II. Approximately 80 trials (4 users each performing 20 trials) were undertaken with each of the four setups. In these in vitro experiments, tissue trauma caused by surgery is quantified as the number of needle insertions into the agar phantom before an acceptable needle placement was obtained for seed deployment.

Table II.  Summary of experimental results.

	Setup
Analysis	Manual	Standard VATS	MIRA II	MIRA III	MIRA IV	P value: MIRA IV vs. VATS	P value: MIRA IV vs. MIRA III
Time (s)	N = 78	N = 83	N = 78	N = 75	N = 80
Mean	36	151	120	85	44	<0.001	<0.001
Median	29	104	107	62	41
Range	7–150	26–665	39–301	28–267	22–112
No. of attempts
Mean	2.3	5.2	3.1	1.8	1.1	<0.001	<0.001
Median	2	4	3	1	1
Range	1–9	1–18	1–10	1–7	1–3
Accuracy (mm)
Mean	2.7	2.5	2.7	2.8	0.9	<0.001	<0.001
Std dev	1.3	1.5	1.6	1.5	0.7
Range	0.4–6.0	0–7.0	0.2–7.3	0.1–6.9	0–3.0

Significant improvements in seed placement accuracy, procedure time and number of attempts (p < 0.001) were obtained with the MIRA IV system when compared to the standard VATS setup. The mean accuracy was improved by 1.6 mm (64% reduction in error, p < 0.001) (Figure 5), the median required time was reduced by 63 seconds (61% reduction) (Figure 6), and tissue trauma (number of attempts) was reduced by 3 (75% reduction, p < 0.001) (Figure 7). MIRA IV also reduced the mean procedure time by 41 seconds (48% reduction, p < 0.001) and the seed placement error by 1.9 mm (68% reduction, p < 0.001) compared to the MIRA III system. When compared to open chest (manual) surgery, the median procedure time was increased by only 12 seconds (41%, p = 0.013), the tissue trauma (number of attempts) was reduced by 1 (50%, p < 0.001), and the seed deposition accuracy improved by 1.8 mm (67%, p < 0.001) with the MIRA IV system. The improvements with the MIRA IV were even more pronounced compared to the MIRA II system (Table II).

Figure 5. The accuracy of each method is displayed as the mean error. The MIRA IV system significantly improved the target accuracy by 64% (2.5 mm to 0.9 mm) compared to a standard VATS and by 68% (2.8 mm to 0.9 mm) compared to the MIRA III system without navigational guidance.

Figure 6. The MIRA IV system significantly reduced the median procedure time by 61% (104 s to 41 s) compared to a standard VATS and by 48% (85 s to 44 s) compared to the MIRA III system without navigational guidance.

Figure 7. The estimate of tissue trauma is based on the number of needle insertion attempts. The MIRA IV system significantly reduced tissue trauma by 75% (from 4 attempts to 1 attempt) when compared to a standard VATS.

Discussion

The new MIRA IV system, which consists of robotic arms, real-time ultrasound imaging, electromagnetic tracking, and the new InterNAV2.0™ navigation software, allowed for very accurate placement of brachytherapy seeds and significant reduction in procedure time and tissue trauma, when compared to standard VATS, MIRA II and the MIRA III techniques. Clinical expert opinion from radiation oncologists and medical physicists estimates that approximately 20 seeds are needed for each 10-mm-diameter tumor. This translates into 21 min and 7 min, respectively, in time saving. Greater time saving can be expected for larger tumors.

When compared to the performance of the reference manual open thoracic brachytherapy procedure, there was an increase in the median time of 12 seconds, although the maximum time was decreased by 38 seconds. If 20 seeds are to be deposited into each tumor as above, this would represent an increase in total surgery time of only 4 min, which is not clinically significant.

The median number of needle insertions is greatly reduced by the use of the MIRA IV system. This is felt to be a very encouraging clinical measure because it represents decreased surgical trauma to the already fragile pulmonary tissue, translating into a decreased likelihood of injury to proximal mediastinal structures, bleeding, and lung collapse.

It was interesting to note that the addition of robotics alone without navigational guidance (MIRA II), while removing the surgeon from the radioactive source and providing better ergonomics, only resulted in minimal changes in performance compared to VATS procedures [10]. A major obstacle in the performance for lung brachytherapy procedures involves the intra-operative needle insertion planning. Without a guidance system, the surgeon would be required to do mental 3D visualizations to plan the insertions, resulting in inconsistencies and inaccuracies. The robotic assistance and electromagnetic navigation complement one another by providing intuitive control and improved dexterity.

Expert clinical opinion stated that an accuracy of 2 mm is biologically and clinically acceptable. The 0.9-mm median accuracy achieved by the MIRA IV system shows its promise in clinical applications. Respiratory movements were not considered in this set of experiments; however, navigational guidance and robotic needle insertion can be expected to make their compensation easier. Future work will address this challenge.

Due to the length and flexibility of the brachytherapy needles, the electromagnetic sensor needs to be located near the tip of the needle. The current sensor was attached on the outside of the barrel, and the resulting added bulk at the needle tip increases trauma to the lung tissue with every insertion. Work is presently underway to incorporate the electromagnetic sensor into the barrel of the needle.

A limitation in any electromagnetic tracking system is the possible magnetic field distortion caused by the presence of nearby metallic objects. Improvements in technology have made the current generation of sensors less susceptible to such interference, however, and with minor reductions in the amount of steel in the surgical setup we found it feasible to use the InterNAV2.0™ system with minimal distortion. Design modification of the devices using only non-magnetic materials will not only eliminate any field distortions, but also enable their use in an MRI scanner, making multi-modality image guidance a possibility. Furthermore, to improve control over radiation dose to tumors, 3D ultrasound image reconstruction [18] and live dosimetry planning capabilities [19] are currently being integrated into the software. Ex vivo and in vivo animal testing will be conducted to assess their performance. Attachments for other treatment modalities (such as HDR brachytherapy, biopsies, radiofrequency ablation, cryotherapy and gene therapy) can be developed to broaden the system's clinical application. Electromagnetic sensors and connecting pieces for robotic arms can be easily added to commercially available applicators for their adoption into the existing system.

In conclusion, a new integrated system for performing minimally invasive robot-assisted lung brachytherapy was developed that incorporated modified electromagnetic navigation software (InterNAV2.0™) and an improved robotic controller. These additions to the MIRA IV system allowed it to perform significantly better than standard VATS and better than both MIRA II without electromagnetic navigation and MIRA III with our earlier electromagnetic navigational system InterNAV™.

Acknowledgments

This research was supported by the Summer Research Training Program from the Schulich School of Medicine, The University of Western Ontario; the Ontario Research and Development Challenge Fund under grant 00-May-0709; the Natural Sciences and Engineering Research Council (NSERC) of Canada under grants RGPIN-1345 and 26 2583-2003; and by infrastructure grants from the Canada Foundation for Innovation awarded to the London Health Sciences Centre (Canadian Surgical Technologies & Advanced Robotics) and to the University of Western Ontario.