Patient-Specific Needle Guidance Templates Drilled Intra-procedurally for Image Guided Intervention: Feasibility Study in Swine

Abstract

Purpose:

Thermal ablation of large tumors may benefit from simultaneous placement of multiple needles, but accurate placement becomes challenging as the number of needles increases. The aim of this work was to evaluate use of personalized needle guidance grid templates based on intraprocedural CT and fabricated at the point-of-care to implement ablation treatment plans with multiple needles in vivo.

Methods:

A plastic frame was designed to hold two parallel plastic guide plates in a rigid relationship, fixed over the abdomen by a mounting arm. Steel ball targets (1.5mm) were implanted under ultrasound in the livers of two domestic swine under general anesthesia. Following breath-hold CT of the subject and frame, the targets and frame were identified using customized 3D-Slicer-based planning software. Multiple needle trajectories targeting the balls were planned, including complex off-plane trajectories. A machining program for drilling the hole pattern corresponding to the planned needle trajectories was generated. The pattern was drilled in the two plates with a numerical-controlled milling machine in the suite. The plates were attached to the frame and needles passed through the paired holes to the calculated depth. Placement accuracy was defined as needle tip-to-target distance on post-placement CT.

Results:

The planning process and manufacturing required approximately 6 and 15 minutes, respectively. Needles were rapidly inserted (n=11) to the target points without complications or traversing non-target anatomy. The mean needle tip-to-target distance error was 3.4±2.2, range 0–7mm.

Conclusion:

Rapid and accurate needle placement was feasible using a subject-specific, custom drilled, needle guidance grid template fabricated intra-procedurally. Targeting accuracy and performance were similar to more complex and expensive tracking systems which may enable accurate intraprocedural implementation of treatment plans in the liver or other organs. This may be of value in complex ablation cases or in areas where more advanced guidance systems are not available.

INTRODUCTION

Some large tumors may be more effectively treated with ablation using multiple needles [1]. However, as the number of needles increases, the complexity also increases. Accurate placement of multiple needles can be difficult, time consuming, and prone to inaccuracies. Multiple needles enabling composite ablations may be facilitated with optical and/or electromagnetic tracking, which can result in effective outcomes [2, 3]. Patients undergoing this kind of ablative therapy may be immobilized using a vacuum beanbag cushion while being scanned in an interventional CT. Needle paths to the tumor may be planned in a variety of fashions. A targeting device may then be used together with a tracking system to position introducers at pre-planned locations and orientations. Cannulas are sequentially introduced and the positions are confirmed using CT prior to inserting ablation probes. A recent study in 96 patients undergoing stereotactic radiofrequency ablation of hepatocellular carcinomas showed pathology-proven success in 50/52 (96%) of the nodules larger than 3cm [2]. Various freehand computer assisted image guidance systems are available but may not be widely used due in part to the time required to plan and position needles in an iterative fashion. Time, ergonomics, equipment investments, training, and expertise remain barriers to adoption for standard commercial tools.

A novel approach is implemented for placement of multiple needles for a composite ablation. This method uses patient-specific needle guiding grid templates that are fabricated at the bedside, during the procedure. This method has been previously validated in vitro using phantoms [4]. The purpose of this study was to evaluate the in vivo accuracy of the template in replicating treatment plans for multiple ablation needles.

Materials and Methods

Device and procedure overview

The fully assembled patient specific needle guidance system consisted of two parallel plastic plates into which needle guidance holes were custom drilled. The plates were held 25 mm apart in a cassette which was itself mounted in a cassette holder and held over the patient using a mechanical positioning arm. To create the guide holes in the plates, a CT scan of the abdomen was acquired that included the cassette holder in its fixed position (Fig. 1). The CT scan was used in special planning software to define multiple needle trajectories to intrahepatic targets. The intersection of these trajectories with the virtual location of the two parallel plates defined the required locations of needle guide holes (Fig. 2). The holes were drilled in the plates using a commercial 3-axis computer numerical control (CNC) milling machine. The plates were affixed to the cassette and the assembly inserted into the cassette holder. Insertion of the needles through the paired holes constrained each needle to follow its predefined trajectory.

Fig. 1.

Planning abdominal CT. a. The cassette holder attached to the positioning arm (asterisk) is shown in fixed position over the abdomen. The cassette holder contained integral fiducials (white arrows) for registration in the planning software. Skin fiducials (black arrows) were used to monitor the accuracy of the system. b. An axial CT section shows the cross section of the cassette holder with one of the integrated fiducials (white arrow) and one of the 1.5 mm steel balls implanted in the liver as a target (white arrowhead)

Fig. 2.

Slicer view of the planning trajectories and computer numerical control (CNC) milling machine and planning computer. The virtual representation of the cassette holder (blue) and two parallel plates (green) is shown. The green virtual plates were not in the actual scan but were calculated based on the fiducials in the frame holder. The thin yellow and red cylinders denote planned trajectories to the targeted 1.5 mm steel balls or skin fiducials. The lower viewports in the image indicate 2D reformats in a bull’s-eye view (left) and along the needle path (center and right) for target 7 (red cylinder) and illustrate the intersection of the trajectory with the target ball and the upper (blue) and lower (green) parallel plates. The computer numerical control (CNC) milling machine and planning computer are shown in the inset

Cassette holder

The polycarbonate cassette holder (150×220 mm) was fabricated with four embedded radiopaque fiducial markers that were used during treatment planning to register its position and define the future position of the plates relative to the subject. The holder was rigidly suspended approximately 2–7 cm above the subject’s abdomen, supported by a custom flexible locking arm. The arm was mounted on a carbon fiber table overlay (Civco Inc., Coralville, Iowa) that rested on top of the CT table, enabling a rigid mechanical co-registration of the holder and arm to the table and CT (Fig. 1A).

Needle targeting planning and design

A planning abdominal CT (Brilliance MX8000 IDT; Philips, Andover, MA, USA) was acquired with a sufficient field of view to include the abdominal targets and the cassette holder with its integrated fiducial markers (Fig. 1B). This was sent to a custom planning module based on 3D Slicer [5]. The software enabled identification of fiducials in the cassette holder and selection of paired needle targets and skin entry points for each needle insertion (Fig. 2). The anatomy along the needle path was previewed in 3D Slicer as two-dimensional oblique, tri-planar reformats, and a three-dimensional reconstruction, allowing adjustment of the needle paths so they did not traverse critical structures, other needles, or ribs. Anatomic structures could also be segmented for ease of visualization. Identification of the fiducial markers on the cassette holder within the 3D Slicer module allowed a registration to be performed between the fiducials in image space and the fiducials in a coordinate system fixed to the cassette holder. This enabled us to display the location of the virtual plates even though they did not appear in the scan. The intersections between each needle path and the virtual plates were calculated along with the insertion depths of the needles and transformed to the coordinate system of the cassette holder. The coordinates of the needle intersections with the target, front and back plates were then exported. A separate custom program written in C++ used the intersection data to calculate the required hole geometry and generated the G-code instructions to control the milling of the plates. The user interface to the program included editable presets for plate size, thickness, and spacing as well as drilling parameters. The required user inputs were needle size selected from a drop-down menu, as that affected hole size, and the identification text to be etched on the plates. G-code is a standard scripting language used for CNC machines that includes information such as cutter speed and movements of the cutting head, optimized to produce the cleanest holes in the shortest time. In practice, milling the elongated holes in the polycarbonate sheets required several passes to create each hole. The best feed rate, spindle speed, and cutter type were determined by multiple bench trials to minimize the chipping and melting of the plate plastic.

Parallel plates and cassette

A commercial three axis numerical controlled (CNC) desktop milling machine (SRM-20, Roland Inc., Shizuoka-ken, Japan) (Fig. 2) was used to mill and drill the guide holes in the two 5 mm thick clear polycarbonate plates. Two plate sizes were available: 75 × 90 mm and 120 × 200 mm. Blank plates were secured in a custom mounting fixture in the CNC milling machine with thumbscrews that ensured correct positioning for drilling. The fixture accommodated two of the smaller plates side-by-side or a single large plate, the latter requiring a plate change during milling. The trajectory holes were drilled and engraved with labels for each hole and patient information. Since the CNC milling machine only drilled vertical holes, the holes for angled trajectories were elongated into correctly sized ovals so the needle would pass through the plate following the planned trajectory. The plates were brushed, reamed with a pipe cleaner, and washed to remove machining residue and were compatible with steam immediate-use (“flash”) sterilization (132°C for 3 min).

System assembly

The drilled and engraved plates were attached to the cassette using thumbscrews and the cassette was mounted into the cassette holder. Care was taken to ensure the cassette holder and subject remained in the same position as at the time of the scan, since changes after the planning CT scan would change the needle paths.

Needle design

While any straight rigid needle or ablation probe could be used, a coaxial introducer needle with a stylet and two-piece hub was developed (Fig. 3). Once inserted, the proximal hub could be unscrewed and removed along with the stylet, leaving the distal end of the cannula in place, yet unattached to the rigid stationary plates. This would prevent pinning of the needle shaft to the parallel plates and mounting mechanism during respirations or motion, with the goal of reducing risk from shear stress or capsule laceration.

Fig. 3.

Coaxial introducer needle with detachable hub. A fully assembled 13G coaxial cannula and inserted stylet are shown. The canula is a two-piece construction wherein the proximal cannula and hub can be unscrewed from the distal cannula that remains in the subject after placement. The 12G coaxial introducer needle is shown with the distal cannula unscrewed and removed from the proximal cannula and stylet

In vivo study

All procedures were performed with approval of the Institutional Animal Care and Use Committee. All applicable international, national, and institutional guidelines for the care and use of animals were followed. Two castrated male Yorkshire domestic swine (78 and 55 kg, Oak Hill Genetics, Ewing, IL) were sedated with intramuscular ketamine (25 mg/kg), midazolam (0.5 mg/kg), and glycopyrrolate (0.01 mg/kg) and anesthetized with propofol (1 mg/kg IV). Animals were intubated and maintained under general anesthesia with isoflurane (1–5%) throughout the procedure, without use of paralytic agents. Steel balls, 1.5 mm in diameter (McMaster-Carr, Cleveland, OH) were implanted in the liver through a coaxial introducer needle under ultrasound guidance to serve as targets. Four markers (X-spots, Beekley Medical, Bristol, CT) placed on the skin served as verification fiducials for confirming positions. With the cassette holder positioned over the animal near the intervention site, a CT scan was performed during breath-hold using 120 kVp and 275 mA reconstructed as 3.0 mm thick sections at 1.5 mm intervals with a 350 mm field (Fig. 1). After transferring the CT data to the planning laptop PC, the cassette holder fiducials were identified and the implanted targets and paired skin entry points were selected, defining needle trajectories and insertion depths (Fig. 2). Needle trajectories to the skin fiducials were also defined for use as position checks. The machining program was generated and the plates were drilled and inserted in the cassette, which was then affixed to the cassette holder. For all needle placements, each needle was marked with the placement depth as measured from the top surface of the top plate for its corresponding hole. Needles were advanced to touch the verification skin fiducials to confirm correct positioning prior to subsequent needle insertion into liver (Fig. 4). Needle insertion into the liver targeting the steel balls was rapidly performed during breath-hold matching the positioning in the planning scan (Fig 5). Following the insertions, CT was performed with the needles in place, reconstructed with 1.0 mm slice thickness at 0.5 mm intervals. Accuracy was defined as Euclidian needle tip-to-target distance as measured on the images via standard coordinate geometry.

Fig 4.

Use of the skin fiducials to confirm accuracy and registration of the targeting system. Skin fiducials were used to assess the accuracy of the system before needle insertion into the subject and to monitor movement during respiration and breath hold. a. Needle inserted through the guide holes for a trajectory designed to place the needle tip on the skin fiducial. b. Needle touching the skin fiducial during breath hold demonstrating fidelity to the treatment planning

Fig. 5.

Assembled template with three needles placed in the swine liver

RESULTS

Performing the CT scan, selecting trajectories and generating the instruction file for milling required approximately 6 minutes. Milling the templates (one pair of plates per swine) with eight pairs of holes, engraving the labels in the plates, and the required plate change were completed within approximately 15 minutes.

Needles were inserted to the planned location of the skin fiducials. The skin fiducials moved approximately 1 cm relative to the corresponding needle tips during mechanical ventilation with normal tidal volume. During breath hold and needle insertion, the needle tips touched the skin fiducials (Fig. 4) confirming stability of positioning and registration. Insertion of each needle into the liver required a few seconds permitting placement of several needles in a single breath hold, without correction of respiratory drift.

The Euclidian tip-to-target distance for internal needles was 3.4 ± 2.2 mm, range 0–7.0 mm for n=11 needles. A representative post procedure CT scan showing the proximity of three needles to the targeted fiducials is depicted (Fig 6).

Fig. 6.

CT imaging following needle insertion. a. Surface rendering of the swine abdomen, needle guidance grid template, and multiple inserted needles. b. Multiplanar reconstruction showing the proximity of the needle tips to two of the target spheres (arrows). Inset shows needle tip in contact with the targeted sphere

DISCUSSION

Successful targeting of liver tumors for ablation is limited by imaging, monitoring, motion, human performance, and potentially large tumor size which necessitates precise and coordinated placement of multiple needles. Without specialized equipment, it is challenging to manually place multiple perfectly positioned needles at one time. This translates into reduced success for larger tumors that require accurate needle placement to ensure complete ablations with adequate margins. In this study, a patient-specific needle guidance grid template technique was demonstrated in vivo, which facilitated accurate, rapid insertion of multiple needles in an automated and prescriptive manner. The method of needle trajectory planning with selection of entry points and targets is comparable to currently available systems. Once trajectories were defined, the system generated the drilling program with minimal input, i.e., needle size and labels, and the plates were milled without additional intervention by the user.

This method physically “encodes” the needle trajectories in a guiding template, simplifying the procedure and eliminating the need for intra-procedural re-positioning or iterative adjustments. This technique expands on a previously described method, that used a template manufactured before the procedure for the implantation of brachytherapy seeds in the prostate and other tumors of the pelvis and neck [6]. However, that technique required a manual alignment and calibration process at the time of the procedure by calculating and drawing overlapping plans on orthogonal AP and lateral plain X-ray images.

Prior bench studies with this system yielded results comparable or superior to conventional techniques [4]. The mean Euclidean distance to the target was 2.8 mm (SD ± 1.0), as measured on CT with 3 mm slice thickness. This was comparable to standard commercial tools based upon electromagnetic and optical navigation (PercuNav, Philips, Bothell, WA; AxiEM and Treon, Medtronic Inc., Louisville, KY) in both phantom and in vivo accuracy reports [7, 8].

Even in a mobile, highly deformable organ such as the liver, this novel system achieved an accuracy in the range of more costly and complex systems. For example, an in vivo study of 19 patients using a prototypical version of the PercuNav system demonstrated a basic needle tip-to-target error of 5.8 mm ± 2.6 [9]. That error improved to 3.5 mm ± 1.9 with use of non-rigid registrations that used previous internal needle positions as additional fiducials. The results with this new system compare favorably, with needle tip-to-target error of 3.4 mm ± 2.2 mm. The skin fiducial technique, where needles were guided to the external skin fiducials, enabled direct visual verification that the needles and trajectories were accurate and enabled easy detection of patient or template motion. Skin fiducials also assisted in coordinating needle advancement with respiratory motion. Patient movement between the planning scan and treatment is a concern. However, a vacuum mattress fixation system could be used to limit patient movement. These devices have been described to aid co-registration of images for single-photon emission tomography/CT and for fractionated stereotactic body radiation therapy with accurate intrasession and intersession positioning [10–13]. Additionally, Widmann et al and Schullian et al have demonstrated that intervention following a CT scan and planning for stereotactic placement of multiple needles for ablation can be performed accurately in a similar setting [3, 14].

The system has the potential to reduce x-ray exposure to both physician and patient. While conventional CT guidance requires multiple incremental scans, the intra-procedural template method could be performed with a single planning scan and a single verification scan prior to ablation, although additional scans may of course be done to verify each needle placement. A two-part needle with a removeable hub was used that would allow detachment from and removal of the parallel plates. This may reduce the risk of organ laceration by allowing the segment of the coaxial needle within the patient to have unrestricted motion during breathing, free from rigid constraint by the parallel plates used to guide placement.

Reduction of plate manufacturing time would be requisite for any broad adoption of this technology. The time to manufacture the drilled plates was determined in part by the number of holes and plate size. For example, eight paths required approximately 15 minutes, including 2 minutes to switch the plates. The machine used was one of the smallest (<20 kg) and the least expensive CNC machines on the market (US$4,000). This three-axis machine could be replaced with a five-axis machine, for more facile drilled hole angulation. CNC drilling machines are available that could drill these plates in under 20 seconds. Two separate parallel plates were engineered instead of a single thicker plate due to cost and weight considerations. Compared to the alternative of 3D printing or additive manufacturing, the subtractive manufacturing and drilling used here is more accurate, less costly, and faster, with a broader choice of materials.

Accuracy and production time may be reduced with more sophisticated optimization. A balance between hole size and needle gauge could be considered, with more manual fine adjustment possible with a size mismatch or offset. Fusion of prior imaging with the software, perhaps with preliminary needle insertion plans that could be edited, may aid in intraprocedural treatment planning. Integration of an ultrasound transducer in an imaging window in the plates might enable real-time feedback and verification.

The study had several limitations. Needle insertion was performed without the physician observing needle advancement with imaging, nor correcting for misalignment. Adjunctive use of CT or ultrasound might improve performance and accuracy and add a measure of safety. The limited ability to correct for patient motion might reduce accuracy or increase risk. However, the use of verification skin fiducials may be used to roughly judge respiration, effectively permitting respiratory gating. Sterility validation remains unproven as does washing out of any non-biocompatible materials shavings. While milling machines employ methods to eliminate particulate generation, the milling machine may need to be placed in an adjacent room to avoid generation of particulates in the surgical suite.

Intra-procedural, custom drilled, patient-specific needle guides enabled accurate insertion of multiple needles to predefined locations. The system was low-cost and required no real-time navigation feedback, yet still achieved in vivo accuracies close to those reported for commercial navigation systems with real-time feedback. Fabrication of templates at the bedside enabled a complex multi-needle ablation treatment plan which could be adapted for needle-based therapies in different organs. The technology may also benefit patient care in medically underserved areas where advanced guidance equipment is not widely available. Patient-specific precision ablation treatment plans may be enabled by this novel bedside drilled system, but further validation and failure mode analysis are required prior to clinical translation.