Robotic Assistance for Quick and Accurate Image-Guided Needle Placement

Abstract

Computed tomography (CT) image-guided procedures including biopsy, drug delivery, and ablation are gaining increasing application in medicine. Robotic technology holds the promise for allowing surgeons, and other proceduralists, access to such CT guided procedures by potentially shortening training, improving accuracy, decreasing needle passes, and reducing radiation exposure. We evaluated surgeon learning and proficiency for image-guided needle placement with an FDA-cleared robotic arm.

Five out of six surgeons had no prior CT guided procedural experience, while one had prior experience with freehand CT guided needle placement. All surgeons underwent a 60-minute training with the MAXIO robot (Perfint Healthcare, Redmond, WA). The robot was used to place needles into three different pre-specified targets on a spine model. Performance time, procedural errors, and needle placement accuracy were recorded.

All participants successfully placed needles into the targets using the robotic arm. The average time for needle placement was 3:44 ± 1:43 minutes. Time for needle placement decreased with subsequent attempts, with average third placement taking 2:29 ± 1:51 minutes less than the first attempt. The average vector distance from the target was 2.3 ± 1.2 mm. One error resulted in the need for reimaging by CT scan. No errant needle placement occurred.

Surgeons (attending fellows and residents) without previous experience and minimal training could successfully place percutaneous needles under CT guidance quickly, accurately, and reproducibly using a robotic arm. This suggests that robotic technology may be used to facilitate surgeon adoption of CT image-guided needle-based procedures in the future.

Introduction

For many years, image-guided interventions utilizing computed tomography (CT) have been the realm of interventional radiology. Applications for percutaneous needle placement include biopsy, energy ablation, placement of fiducials within a mass for surgical excision, image-guided injection, and drain placement. As CT guided procedures become more common, methods to train and deploy surgeons or other interventionalists for such procedures becomes increasingly important. Computer-aided systems hold the promise of allowing increasing access for surgeons or other proceduralists to CT-guided interventions by shortening the learning curve. Single-arm robotic systems have been developed to direct the trajectory of percutaneous needles by registering previously obtained axial images to robotic computer software [1–3]. While percutaneous needle placement is commonly used on surgical patients, these needles are not often placed by surgeons, but more frequently by radiologists. This may change as improvements in robotic technology lead to increased ease of use. In the current study, we evaluated whether surgeons without prior training in a CT environment could quickly learn procedures in such an environment using MAXIO. This concept could then be expanded to include other proceduralists as well, particularly for trainees. MAXIO is an FDA approved CT-guided, physician controlled stereotactic robotic accessory to a CT system currently marketed for interventional radiology use. We sought to determine what surgeon learning and proficiency would be using this robotic arm and hypothesized that surgeons could quickly be taught to accurately place needles in predefined targets on a model using the MAXIO system.

Materials and methods

The studies described herein were performed with Institutional Review Board approval. Six surgeons, of which five had no prior CT guided procedural experience underwent 60 minutes of training with the MAXIO robotic system (Perfint Healthcare, Redmond, WA) by a product representative. The MAXIO robot is an image-guided, physician controlled stereotactic accessory to a CT system. It consists of a single robotic arm with an attached navigating system that coordinates with CT imaging to assist with preoperative planning, intraoperative guidance, or post-procedural confirmation of location toward a designated target. The MAXIO robot designs and aims needles or probes on trajectories to designated targets for operator insertion. The MAXIO utilizes manual needle or probe advancement, giving the operator ultimate control over placement, but provides direction and assists with positioning the needle or probe on the desired trajectory. The MAXIO is a USFDA approved device.

Following training, each participant then used the robot for three-needle placement trials, aiming for pre-specified needle tip targets within an opaque phantom spine model (CIRS Model 034 Lumbar Training Phantom, Computerized Imaging Reference Systems, Inc, Norfolk, VA). The lumbar model was selected for this project, though the workflow would have been identical with other anatomic radiologic models. The needle tip targets were arbitrarily selected locations within the model designated on the pre-procedural CT scan. The goal of needle placement was to place the tip of the needle as close as possible to this target location in as short a time as possible. Trials were monitored by two different observers, and video recorded for later review.

Workflow for needle placement proceeded as follows. Before testing, the MAXIO robot was placed within the radiology suite at a pre-specified location on the floor relative to the CT scanner (General Electric HD750 64 Slice Computed Tomography Scanner) and connected to power and an ethernet cable to enable communication for loading of CT images. An initial CT scan was obtained with 1.25 × 1.25 mm thick slices of the spine model, which was loaded into the robotic computer interface for registration. Of note, this computer interface is connected to the robotic arm itself. The needle trajectory and target for the needle tip were selected by the surgeon using the robotic computer interface. The CT scanner position and robotic arm were then adjusted as specified by the robotic computer program. The participant then placed a 150 mm 22-gauge needle into the spine model through the robotic arm’s needle holder. A repeat CT scan was performed and compared to the initial scan to assess needle placement accuracy in three dimensions by assessing the shortest vector distance of the needle tip after placement from the intended target on the initial CT scan. Accuracy was measured using the MAXIO’s image analysis tools contained within the MAXIO software after needle placement. Time for scanning, image loading, and interaction with the robotic software/needle placement was recorded, as were the number and type of surgeon errors during the workflow. Computer errors were counted if a technical error arose that disrupted our needle placement workflow such as computer glitches or freezing. Human errors were counted if delay or mistakes in workflow were due to human factors such as steps forgotten or done out of order. Between each investigator, the computer was reset to its starting screen to ensure a consistent starting point. Data were analyzed using Microsoft Excel v 15.0 (Redmond, WA). Values are presented as the mean ± standard deviation and compared using Student’s t-test where appropriate.

Results

Feasibility and Time

All participants successfully placed all the designated needles into the pre-specified targets using the robotic arm (Figure 1). The total scanning time average (initial scan + final scan) was 3:08 ± 0:51 minutes. The average time for the initial robotic computer interface set-up was 2:35 ± 1:02 minutes, and the average time for each needle placement was 3:44 ± 1:43 minutes. The time required for needle placement significantly decreased from the first to the second attempt (p = 0.02), and then stabilized (p = 0.91, Figure 2). The average time for the third needle placement took 2:29 ± 1:51 minutes less than the initial attempt. The average total time for placement of all three needles was 15:50 ± 3:57 minutes.

Fig. 1. Image-guided needle placement with robotic assistance.

a) The robotic arm and computer interface positioned over the spine model as a surgeon passes the needle through the needle holder. b) The robotic arm after it has been released from the needle. c) Imaging from the computer interface showing the previously placed needles (fuchsia, teal) and the anticipated position of the next needle (maroon). d) Imaging after placement comparing planned needle trajectory (dotted maroon line) and actual needle trajectory (green).

Fig. 2. Graph showing time to needle placement by attempt.

Time needed by the surgeon to place the needle into the spine model after the initial scan, by attempt at needle placement. Time for needle placement significantly decreased after the first needle placement, and then was similar for the second and third needle placements.

Accuracy and Errors

Computer and operator errors were assessed during each trial. There were no computer errors and 0 – 4 operator errors per participant. These errors were minor and included issues with image loading and using the robot computer interface. No errors resulted in errant needle placement. One error resulted in the need for an additional CT scan. After needle placement, the average distance from the needle tip to the target was 2.3 ± 1.2 mm (range 0 − 4.4 mm), with similar findings amongst investigators.

Discussion

In this study, we tested the hypothesis that surgeons with no previous experience in CT guided needle placement could rapidly be trained in how to quickly and accurately place percutaneous needles using a physician controlled stereotactic robotic system. We demonstrated that this was possible with only an hour of training. From start to finish, surgeons could place needles percutaneously in pre-selected locations with an accuracy of 2.3 mm in less than 10 minutes on their first attempt, and with increased speed on subsequent attempts. There was almost no learning curve in the use of this system. We believe this accuracy and time frame would be considered acceptable by most surgeons or other proceduralists, for most applications, warranting further evaluation of this technology by non-radiologist proceduralists.

Given our lack of experience with CT guided needle placement, our study was not designed to compare differences in accuracy between freehand and robot-guided needle placement. Additionally, our study was a small, proof of concept study. A larger in vivo study is needed to demonstrate the full effect, however, our study showed that proceduralists with no prior experience could be trained quickly and accurately on a new technology with the potential to help patients.

A key advantage of the MAXIO, however, is its accuracy. Many other robotic stereotactic guidance systems have also been developed for this purpose. These are most commonly applied in neurosurgery and are typically used to assist in directing the trajectory of drills or needles reaching targets within the brain for biopsy, ablation, electrode placement, or catheter placement. Comparative reviews report accuracy similar to the present study, with a 1 – 4 mm deviation from target range [2]. However, the effect these robots have on accuracy and post-operative outcomes remain questionable, due to the heterogeneity of these studies [4].

For the MAXIO, results comparing accuracy have also varied. A previous study of experienced interventional radiologists compared manual fluoroscopic needle placement and robotic guidance with the MAXIO in a porcine model and showed no difference in accuracy [5]. Another report out of a radiology department compared freehand single-pass needle insertions to MAXIO robotic guided needle insertions on a phantom model and showed significant differences in accuracy of needle placement (15 mm vs 7 mm) [6].

It should also be noted that there are other methods of percutaneous imaged guided navigation. These include optical tracking devices, electromagnetic navigation devices, and laser navigational systems [7]. Optical tracking devices use an overhead camera that syncs with axial images and continuously follows instruments so that the operator can see their location relative to the lesion of interest but these systems require a line of sight between the tracking system and the sensor [8]. Electromagnetic navigation systems use an electromagnetic field generator either positioned next to or placed as a plate underneath the patient. These systems can be less accurate, but do not require a line of sight. Further, the tracking devices can be attached to the tip of a tool since it does not always have to be visible to an overhead camera [9,10]. Laser navigational systems have a rotatable laser unit that shows the needle entrance point and angle [11]. All these methods require some element of “free handedness,” however, and there are no studies comparing multiple modalities.

Another benefit to robotic needle guidance is the potential for decreased radiation exposure by decreasing the number of confirmatory scans and eliminating the need for live imaging. In this study, only 1/18 (6%) of needle placements required more than a pre-placement and post-placement scan. In the MAXIO study cited above with experienced interventional radiologists, even though accuracy was similar, the total radiation dose was significantly decreased with robotic assistance [5]. In that study, manual placement required on average 6–7 confirmatory CT scans, compared to robotic assistance, which only required 1–2 scans. Further, during CT guided procedures there may be additional radiation exposure from real-time live imaging while performing the procedure itself, which is not required when using robotic guidance. This radiation exposure during a procedure can account for an additional 15% of radiation to the patient [12]. Typical total dose-length product for percutaneous procedures range from 500 – 2800 mGy, of which 50 – 150 mGy maybe during live imaging [12,13]. This exposure is affected by equipment, with newer equipment delivering lower radiation [12], and also varies by procedure type, with biopsies and aspirations typically requiring less exposure than ablations and drainage procedures [13]. Beyond this, there is also radiation exposure to the operator, which has been measured at 0.2 – 40 uGy/second when 10 – 35 cm from the target [14,15].

This study does have several limitations. This investigation was small and intended to demonstrate a proof of concept, which may limit the generalizability of our results to other surgeons and other non-radiology focused proceduralists. Percutaneous needle placement was also performed on a spinal model, and not on a live patient who may move and breathe. Respiratory variation and other patient movements may affect target to registration accuracy. It can be assumed that this would increase procedural time further. However, the type of model itself was selected for stability, and as the distance measured from the target was linear and identifying the target on the scan was independent of the type of phantom used, the type of phantom does not take away from the broader promise of the study.

Conclusions

In this study, surgeons without previous experience and with minimal training were able to place percutaneous needles into a model under CT guidance quickly, accurately, and reproducibly using a robotic arm. Our findings suggest that robotic technology may enable surgeons or other interventionalists to perform percutaneous needle-based procedures safely and efficiently and that it has the potential to enhance the adoption of CT-image guidance in surgery, thus warranting further evaluation of this technology by non-radiologist proceduralists.