An MR-Conditional Needle Driver for Robot-Assisted Spinal Injections: Design Modifications and Evaluations

Abstract

This paper introduces design modifications to our MR-Conditional, 2-degree-of-freedom (DOF), remotely-actuated needle driver for MRI-guided spinal injections. The new needle driver should better meet cleaning and sterilization guidelines needed for regulatory approval, preserve the sterile field during intraoperative needle attachment, and offer better ergonomics and intuitiveness when handling the device. Dynamic and static force and torque required to properly install the needle driver onto our 4-DOF robot base are analyzed, which provide insight into the risks of intraoperative tool attachment in the setting of robot-assisted spinal injections under MRI guidance.

I. INTRODUCTION

Low back pain is a common condition that affects the lower portion of the spine, and is a leading cause of activity limitation and work absence globally [1]. Aside from physical therapy and other rehabilitative methods, spinal injections remain a common tool to either diagnose the source of pain, or deliver medication in order to relieve pain. However, these types of injections are commonly performed manually by the clinician under computed tomography (CT) or X-ray fluoroscopy guidance [2]–[4], subjecting both the clinician and the patient to ionizing radiation during the process. Efforts have been made to utilize ultrasound (US) as an alternative guidance method for spinal injections [5]–[7], and although clinical studies have reported similar efficacy as compared to fluoroscopy-guided injections [8], its limitations such as acoustic shadowing artifact and inability to visualize deep structures make US-guided injections a difficult skill to master—often times, fluoroscopy is recommended to confirm needle placement, thus some radiation is still required [9].

Magnetic resonance imaging (MRI), which provides superior soft tissue contrast and produces no ionizing radiation, has gained great attention in the medical robotics community as an image guidance modality. Although challenges exist due to the strong magnetic field of magnetic resonance (MR) environment, a number of robotic systems for MRI-guided interventions have been developed, and their safe operations in the MR environment have been investigated. For instance, Krieger et al. demonstrated the needle placement capabilities of a robot for MRI-guided prostate interventions in canine experiments and clinical trials [10]. Fischer et al. developed a pneumatically actuated needle placement device for real-time MRI-guided transperineal prostate interventions [11]. Patel et al. developed a patient-mounted robotic system for MRI-guided shoulder arthrography, and has reported targeting accuracy of 2.07mm from cadaver experiments [12]. Our lab developed an MR-Conditional 6-DOF robot for lumbar injections, and the device was able to achieve targeting error of 1.3mm for facet joint injection and 1.9mm for epidural injection in cadaver experiments [13]–[16].

While many systems for MRI-guided interventions have been prototyped, few have accounted for the requirement to keep the surgical field sterile during actual clinical procedures [17]–[21]. Among those, separating the device into sterile and non-sterile groups, and draping the non-sterile group with a sterile drape remains a popular option. For example, in [21], the authors describe a MRI-guided robot for percutaneous interventions, where the non-sterilizable robot base can be covered by a sterile plastic drape, but requires a sterile screwdriver in order to install the other sterile components. An exception is introduced in [19], where the authors showcase a robot actuated by a combination of ultrasonic motors and pneumatics, and the needle goes through the center of the parallelogram-based robot. Since the complexity of the robot mechanical design prevents the introduction of sterile drapes, the authors demonstrate the sterilizability of the entire robot.

One system that has been approved by the US Food and Drug Administration (FDA) is an MR-Safe, pneumatically actuated robot for transperineal prostate percutaneous interventions. Developed by Stoianovici and colleagues, this robot also features the separate-and-cover method to separate sterile and non-sterile components [22].

Although our robotic system has demonstrated great potential for MRI-guided lumbar injections, previous studies have revealed several flaws in our needle driver design that hinder its necessary post-operative device cleaning and sterilization. Patient variability as well as target selection could prevent the 4-DOF robot base from carrying the needle driver to the correct pose. In addition, the needle driver is not intuitive to use for someone unfamiliar with the device, albeit its ingenuity from an engineering perspective.

As a step to address these shortcomings and move forward in our regulatory approval process, we propose modifications of the original needle driver design and suggest changes to our clinical workflow. The new needle driver design (1) provides a clear delimitation between sterile and non-sterile components to preserve the sterile field, thus offering the possibility of attaching the needle driver intraoperatively, (2) enables easy sterilization process of sterilizable components, and (3) features better ergonomics and a more intuitive process for device installation and needle attachment. The rest of the paper is organized as follows: Section II reviews the previous design and clinical workflow, presents the new needle driver design, and suggests a modification to the workflow. In Section III, force and torque experiments are performed on the new design, and an evaluation of the results are presented in Section IV.

II. MECHANICAL DESIGN AND MODIFICATIONS

Design of the needle driver closely relates to the clinical workflow, especially considering the sterile field that must be preserved in the operating room. Post-operative cleaning and sterilization of the surgical device must also be considered. Thus, components of the needle driver must be modular, and a clear categorization of the sterility of each modular component is required.

A. Review of Needle Driver Design and Clinical Workflow

As detailed in [15], [16] and shown in Figure 1, the MR-Conditional 6-DOF robot consists of a 4-DOF robot base and a 2-DOF needle driver. Power to the needle driver is transmitted by a beaded chain mechanism and is provided either manually by turning the crank and knob, or automatically by two motors of the actuation unit. A detailed explanation of the construction of the remote actuation mechanism, as well as the design of the beaded chain sprockets, is provided in [13], [14]. Figure 2 illustrates the needle driver used in previous studies. The needle driver’s translational DOF is provided by the bottom chain sprocket, which turns a lead screw and actuates the top sprocket assembly in a linear fashion. The rotational DOF is controlled directly by the top chain sprocket. The needle driver is primarily composed of 3D printed ABS plastic, with some non-ferrous metal components made from brass and aluminum.

Fig. 1:

Overview of the 6-DOF MRI-Conditional robotic system for lumbar injection. The 2-DOF needle driver is attached on top of the 4-DOF robot base, which will be mounted on the low back region of the patient. The actuation unit transmits power to the needle driver via a beaded chain transmission.

Fig. 2:

A close-up view of the previous needle driver design, with the needle attached. The needle goes through the top beaded chain sprocket and the bottom chain sprocket housing before reaching the robot base and patient. The needle is fixed or released by turning the thumbscrew near the top.

Briefly, to attach the needle onto the needle driver, the clinician needs to pass the needle through a series of components of the driver before the needle can reach the patient; to secure the needle in place, the thumbscrew needs to be turned to pin the plastic needle base in place. To detach the needle, the clinician simply reverses the steps, and pulls out the needle from the driver; however, the series of components that have physical contact with the contaminated needle cannot be independently removed without some strenuous effort by the engineering team. Notably, the beaded chain transmission for each needle driver DOF is installed in a one-and-done fashion: to remove any component in the transmission, one must either cut the beaded chain, or break some other connected component, making it practically impossible to isolate the contaminated parts for post-operative cleaning and sterilization. The needle attachment process itself might also be difficult to untrained clinicians, as the components that the end users will interact with are rather unrefined.

The needle driver does not carry any fiducial markers, therefore previous clinical workflow suggests immediate needle driver attachment onto the 4-DOF robot base after the base is mounted onto the patient. Homing, registration, and targeting procedures are carried out with the needle driver attached [16]. However, cadaver studies (with needle insertions) and volunteer studies (without any needle insertion) have revealed that, depending on the mounting position of the robot on the patient and selected target needle orientation, the robot base could not always carry the needle driver to the desired pose due to the weight of the needle driver and the reaction force created by the beaded chain tubing. These findings suggest the needle driver to be attached intraoperatively, presumably after the robot base has moved into position designated by the surgical planning software.

B. Modified Needle Driver Design

The aim for modifying the previous needle driver design is precisely to address the aforementioned shortcomings. A new design is proposed, and its distinct groups of components, as well as needle attachment process, are shown in Figure 3. A few new components are introduced: a needle adapter (green) to hold the needle and provide connection to the top chain sprocket (purple) of the main driver body (white), a driver-robot interface (red) and locking ring (cyan) to facilitate driver attachment to the 4-DOF robot base, and a detachable needle guide (blue) to contain the needle and prevent any contact between needle and other components of the robot.

Fig. 3:

New needle driver design and its attachment process. The driver-robot interface is non-sterile, while the rest will be sterile upon entering the operating room.

To attach the needle and install the needle driver, the first step is to insert the needle to the needle adapter. The second step involves putting the main needle driver body onto the driver-robot interface. This interface consists of a long aluminum rod, which is part of the 4-DOF robot base [15], and a slotted cylindrical component, which will engage the locking ring mechanism on the driver body. To engage the lock in step three, the clinician can use one hand to hold the needle driver body, and the other to rotate the locking ring with its two handles. Steps four and five involve inserting the needle guide and the needle adapter onto the interface and driver, respectively. The driver’s top chain sprocket is redesigned, and the needle adapter has one channel on each side to guide its path during the attachment process. Finally, in step six, turn the needle adapter counterclockwise slightly to lock the adapter to the chain sprocket. A detailed view of the needle adapter design and assembly steps 5 and 6 is shown in Figure 4.

Fig. 4:

Detailed view of the needle adapter design and the attachment steps 5 and 6. The needle adapter’s compliant handle and leaf springs allow easy needle attachment and detachment, and its rigid channels and tabs provide secure engagement with the top chain sprocket.

This design involves more moving pieces, which might seem unnecessarily complex, yet the advantages it offers outweigh its appearance.

First, the new design is intuitive to install with the addition of the locking ring mechanism and the needle adapter. The cylindrical interface has narrowing slots with large initial openings, which can help self-correct the needle driver body position during the locking process. The needle adapter is designed particularly to be compliant and to match the convex profile of the plastic needle base, allowing easy and secure needle attachment.

More importantly, a clear distinction between sterile and non-sterile groups is established. The driver-robot interface, which remains with the 4-DOF robot, is non-sterile; all the other components will be sterile, and can be installed without the clinician touching the non-sterile surface, thus preserving the sterile field during needle attachment, and making it possible to perform needle attachment intraoperatively after the robot base has moved into position. The components that physically interact with the needle, i.e. the needle adapter and needle guide, can be detached, cleaned and sterilized, or simply disposed after each use. Currently, the needle guide assembly consists of a Delrin handle and a brass tube, and the needle adapter is 3D printed using Tough 1500 resin (Formlabs, Somerville, MA, USA) but could be replaced with medical grade nylon. These components could be sterilized with low-temperature methods with gas or steam. The main needle driver body is not sterilizable in its current state, but work is underway to make it sterilizable as well.

III. FORCE AND TORQUE EXPERIMENTS

A series of experiments is carried out to measure the force and torque required to attach the needle and driver, in order to evaluate the risk of invalidating the 4-DOF robot base position by the attachment process. While it is impossible to fully simulate the clinical situation where the 4-DOF robot base is strapped onto a patient along with the imaging coil, the force and torque information collected is used to provide insight into the attachment process of the new needle driver, and offer feedback on the design itself.

The setup for data collection is shown in Figure 5. A 6-axis force and torque sensor (Nano25, ATI Industrial Automation, NC, USA) is attached onto a vise (Model 350, PanaVise Products Inc, NV, USA). The force sensor holds the driver-robot interface by the aluminum rod, and the vise is tilted into different poses to simulate different mounting scenarios that could arise during a clinical procedure.

Fig. 5:

Experiment setup for force and torque measurements. A vise holds the force and torque sensor, which supports the aluminum rod and the driver-robot interface. Five different poses as well as directions of positive sensor measurements are labeled.

Two sets of experiments are conducted. For the first experiment, data are collected for each distinct action that could generate force and torque to the robot. Although a total of six steps are required to complete the needle attachment process as shown in Figure 3, combination of steps is required to fully complete an action. To this end, we group steps 2 and 3 as an action that attaches the needle driver body onto the interface (Driver); group steps 1, 5 and 6 as an action that attaches the needle onto the driver (Needle); the only step left is step 4, which represents an action that attaches the needle guide onto the interface (Guide). Each action is repeated 15 times in each pose. For the second experiment, the three actions are performed in sequence for 10 times in each pose, and complete force and torque histories are recorded.

For both types of experiments described above, the needle driver is positioned into five different poses: (V)ertical, 30° (E)ast, 30° (S)outh, 30° (W)est, and 30° (N)orth with respect to the vertical position. The angles are checked using a laser level (DLT-675, Checkpoint Levels, CA, USA).

Since the new workflow suggests attaching the needle driver intraoperatively, the actual position of the actuation unit could vary based on the level of tension created by the beaded chain tubing, allowing a more natural configuration to be assumed; therefore in the experiments, after the needle driver is moved to a new pose, the actuation unit is also moved to alleviate stress in the tubes.

IV. RESULTS AND DISCUSSION

To evaluate the feasibility of attaching the needle driver intraoperatively, the maximum force and torque values for each step during the dynamic attachment process must be considered, since they are important factors determining whether such disturbance could potentially move the robot relative to the patient, thus invalidating the 4-DOF robot position. For this reason, components of the maximum values for each action are reported when the needle driver is attached at different poses, as well as the maximum resultants $F_{max}^{res}$, ${\tau }_{max}^{res}$, and the mean resultants after the system reaches steady state $F_{\mathit{\text{mean}}}^{ss}$, ${\tau }_{\mathit{\text{mean}}}^{ss}$, during which the main contributor is the weight of the components and the hanging beaded chain tubing. The needle guide weighs approximately 0.03N, and the assembled needle and adapter 0.04N. The weight of the needle driver is difficult to measure due to the contribution to the hanging tubes.

A summary of data collected for the first experiment is presented in Tables I and II. As made evident from the two tables, significant differences exist between the maximum resultants and the steady-state resultants. This observation suggests that dynamic forces application during the attachment process could have a much greater impact than the static forces, and by considering only the steady-state resultants, one could potentially underestimate the potential risk of moving the robot base position relative to the patient body.

Table I:

Force Maximum and Steady-State Mean for Individual Actions (Unit: [N])

Pose	Action	$F_{max}^x$	$F_{max}^y$	$F_{max}^z$	$F_{max}^{res}$	$F_{\mathit{\text{mean}}}^{ss}$
V	Driver	−2.62	−1.54	−6.98	7.61	1.80
	Guide	−1.33	0.79	−4.76	5.00	0.04
	Needle	−0.61	1.15	−4.44	4.63	0.05
E	Driver	−1.84	3.11	−4.53	5.80	2.64
	Guide	−1.72	0.90	−4.19	4.61	0.04
	Needle	−0.65	1.12	−3.55	3.78	0.03
S	Driver	−3.26	6.36	−6.04	9.36	1.84
	Guide	−1.25	0.74	−3.87	4.13	0.04
	Needle	0.83	0.40	−2.50	2.67	0.04
W	Driver	2.18	−3.13	−6.32	7.38	1.82
	Guide	−1.10	−1.11	−4.78	5.03	0.04
	Needle	−0.90	0.94	−4.71	4.89	0.05
N	Driver	2.06	1.48	−8.04	8.43	1.81
	Guide	−1.27	−0.47	−4.13	4.34	0.05
	Needle	−0.67	0.99	−2.79	3.03	0.04

Table II:

Torque Maximum and Steady-State Mean for Individual Actions (Unit: [Nm])

Pose	Action	${\tau }_{max}^x$	${\tau }_{max}^y$	${\tau }_{max}^z$	${\tau }_{max}^{res}$	${\tau }_{\mathit{\text{mean}}}^{ss}$
V	Driver	−0.322	−0.357	0.214	0.526	0.290
	Guide	0.076	−0.106	−0.008	0.130	0.001
	Needle	−0.214	−0.137	−0.077	0.266	0.001
E	Driver	−0.402	−0.228	0.214	0.509	0.346
	Guide	−0.079	−0.133	0.008	0.155	0.001
	Needle	−0.227	−0.134	−0.053	0.269	0.003
S	Driver	−0.537	−0.425	−0.483	0.838	0.311
	Guide	−0.062	−0.109	0.007	0.125	0.002
	Needle	−0.080	0.163	−0.034	0.185	0.004
W	Driver	0.201	−0.175	−0.199	0.333	0.205
	Guide	0.099	−0.093	−0.007	0.136	0.002
	Needle	−0.175	−0.176	−0.066	0.257	0.003
N	Driver	−0.266	−0.253	−0.265	0.453	0.297
	Guide	0.043	−0.102	0.006	0.111	0.002
	Needle	−0.189	−0.140	−0.042	0.239	0.004

Regardless of pose, the first action—attaching the main needle driver body onto the driver-robot interface—has the largest resultant than the other two actions. This is expected, since the driver body outweighs the other components and requires larger force to manipulate. However, the introduction of needle driver brings a large torque to the system, even when it is in the steady-state vertical position. This highlights the amount of resistance from the stiff beaded chain tubing as it keeps trying to return to its more natural configurations. A better solution could be needed to further alleviate the stress in the tubes.

As mentioned before, previous clinical workflow requires the needle driver and the actuation box to be installed before any homing or targeting. If the actuation unit is not properly positioned, it is likely that certain pose of the needle driver would create significant reaction force due to the stiff tubing. Since the new design opens up the possibility of attaching the needle driver intraoperatively after the 4-DOF robot base has moved into position, the position of the actuation unit could be adjusted according to the final robot configuration, and could result in much more consistent behavior of the robot after the needle driver is attached. This is verified from benchtop experiments, by comparing the standard deviations of the steady-state force and torque resultants of the new workflow to those reported previously in [14], since both studies use the same experiment setup. For the new workflow, the standard deviations for force and torque resultants across all five poses are 0.22N and 0.049Nm, respectively; the standard deviations of the values reported in [14] are 0.66N and 0.182Nm for a 110cm initial transmission length, and 0.43N and 0.079Nm for a 80cm initial transmission length: all higher than the current values.

For the second experiment, entire force and torque histories for the needle and driver attachment process are recorded, and a typical trial is shown in Figure 6. The final steady-state values are listed in Table III. A cross-comparison of the steady-state resultants in Table III and those in Tables I and II reveals that, for each pose, the sum of individual action’s steady-state resultants is approximately the same as the total resultant at the end of the process; small variations exist due to the fact that in the second experiment, the components are put in random locations before they are picked up and installed, therefore the beaded chain tubing could have assumed a slightly different configuration than the other trials.

Fig. 6:

Force and torque history of a typical needle attachment process. The beginning and end of each action is marked by dashed lines.

Table III:

Steady-State Mean on Completion of Needle Attachment (Units: [N], [Nm])

Pose	$F_{\mathit{\text{mean}}}^x$	$F_{\mathit{\text{mean}}}^y$	$F_{\mathit{\text{mean}}}^z$	$F_{\mathit{\text{mean}}}^{ss}$	${\tau }_{\mathit{\text{mean}}}^x$	${\tau }_{\mathit{\text{mean}}}^y$	${\tau }_{\mathit{\text{mean}}}^z$	${\tau }_{\mathit{\text{mean}}}^{ss}$
V	0.09	0.72	−1.64	1.79	−0.192	−0.027	−0.184	0.267
E	−0.31	2.29	−0.69	2.42	−0.297	−0.011	−0.147	0.331
S	0.78	0.97	−1.53	1.97	−0.191	0.072	−0.203	0.288
W	0.10	−0.19	−1.91	1.92	−0.130	−0.027	−0.130	0.186
N	−0.76	0.80	−1.51	1.87	−0.249	−0.124	−0.114	0.301

Figure 6 shows that the reported peak force and torque values are only transient, and will diminish in less than a second. Although previously in [15], a finite element analysis is conducted using forces and torques of the same order of magnitude, and reveals that the deformation to the 4-DOF base robot is submillimeter, the actual consequence of these loads when device is attached to the patient is unknown—device mounting position, tissue compression of the patient, etc. can all play into how much the robot could move relative to the patient under different loading conditions. Subsequent studies are necessary to interpret the physical meaning of these force and torque values under a more realistic setting.

A common theme that is strictly inherent in the design aspect of the needle driver manifests itself in the form of small jiggles right before the force peaks, shown in Figure 6. For attaching the main driver body onto the interface (Driver), the jiggle comes from aligning the bottom sprocket housing and the locking ring tabs to the interface; for attaching the needle guide (Guide), it comes from the needle guide hitting the top of the aluminum rod, which happens quite often since the rod does not feature a conic surface to make it easy for the thinner tube to go in; for attaching the needle and adapter onto the driver (Needle), it comes from passing the needle through the opening of the needle guide. Although the magnitude of these jiggles are small, the duration is rather long. Modifications of the current design, especially the addition of more conic surfaces to help align tubular structures, are necessary to further streamline the needle attachment process.

The viability of attaching the needle driver intraoperatively needs to be further verified by a clinician. To test if the reported force and torque values could create relative motion between the robot and patient and invalidate the targeting procedure, further studies need to be carried out in a more realistic setting that is closer to the actual clinical situation.

In addition to implementing the suggested design changes and verifying the modified workflow, future work also includes searching for lightweight materials to reduce the static load of the needle driver, and more compliant tubes to further reduce the torque transmitted to the needle driver.

V. CONCLUSIONS

This paper presents design modifications to our 2-DOF MR-Conditional needle driver for spinal injections. The new design is better in line with device post-operative cleaning and sterilization regulations, more intuitive to work with, and offers the possibility of attaching the needle driver intraoperatively. Force and torque data suggest that the new design and workflow could generate more consistent behavior to the 4-DOF robot base in steady state, but underlines the possibility of large dynamic forces during the attachment process.