Simulated accuracy assessment of small footprint body-mounted probe alignment device for MRI-guided cryotherapy of abdominal lesions

Abstract

Purpose:

Magnetic resonance imaging (MRI)-guided percutaneous cryotherapy of abdominal lesions, an established procedure, uses MRI to guide and monitor the cryoablation of lesions. Methods to precisely guide cryotherapy probes with a minimum amount of trial-and-error are yet to be established. To aid physicians in attaining precise probe alignment without trial-and-error, a body-mounted motorized cryotherapy-probe alignment device (BMCPAD) with motion compensation was clinically tested in this study. The study also compared the contribution of body motion and organ motion compensation to the guidance accuracy of a body-mounted probe alignment device.

Methods:

The accuracy of guidance using the BMCPAD was prospectively measured during MRI-guided percutaneous cryotherapies before insertion of the probes. Clinical parameters including patient age, types of anesthesia, depths of the target, and organ sites of target were collected. By using MR images of the target organs and fiducial markers embedded in the BMCPAD, we retrospectively simulated the guidance accuracy with body motion compensation, organ motion compensation, and no compensation. The collected data were analyzed to test the impact of motion compensation on the guidance accuracy.

Results:

Thirty-seven physical guidance of probes were prospectively recorded for sixteen completed cases. The accuracy of physical guidance using the BMCPAD was 13.4 ± 11.1 mm. The simulated accuracy of guidance with body motion compensation, organ motion compensation, and no compensation was 2.4 ± 2.9 mm, 2.2 ± 1.6 mm, and 3.5 ± 2.9 mm, respectively. Data analysis revealed that the body motion compensation and organ motion compensation individually impacted the improvement in the accuracy of simulated guidance. Moreover, the difference in the accuracy of guidance either by body motion compensation or organ motion compensation was not statistically significant. The major clinical parameters impacting the accuracy of guidance were the body and organ motions. Patient age, types of anesthesia, depths of the target, and organ sites of target did not influence the accuracy of guidance using BMCPAD. The magnitude of body surface movement and organ movement exhibited mutual statistical correlation.

Conclusions:

The BMCPAD demonstrated guidance accuracy comparable to that of previously reported devices for CT-guided procedures. The analysis using simulated motion compensation revealed that body motion compensation and organ motion compensation individually impact the improvement in the accuracy of device-guided cryotherapy probe alignment. Considering the correlation between body and organ movements, we also determined that body motion compensation using the ring fiducial markers in the BMCPAD can be solely used to address both body and organ motions in MRI-guided cryotherapy.

1. INTRODUCTION

Magnetic resonance imaging (MRI)-guided percutaneous cryotherapy is a procedure to eradicate a lesion by creating an ice ball around the needle-like probe tip inserted through the skin under MRI visualization. Magnetic resonance imaging can clearly delineate the ice ball as a signal void, owing to the rapid T2* decay in the frozen tissue. This enables a physician to monitor the growth of the ice ball in real-time for safe and effective treatment. The procedure can be performed in either of open-bore¹ or closed-bore scanners.^2–4 Recently, MRI-guided percutaneous cryotherapy has become a viable clinical option particularly in treating relatively small lesions adjacent to vital anatomical structures in the abdomen.⁵ The current procedural terminology (CPT) codes defined by the American Medical Association covers MRI-guided cryoablation for insurance reimbursement.

Whereas the literature has demonstrated that MRI-guided percutaneous cryotherapy can be performed at least as safely as computed tomography (CT)-guided cryotherapy,⁶ the accuracy of probe alignment needs to be improved to ensure an adequate safety margin of the ablation to obtain efficacious treatment results. Increased accuracy of probe alignment can also minimize trial-and-error in probe placement, potentially resulting in less adverse events. Researchers have identified the importance of tools in facilitating probe placements and proposed navigation software or guidance devices to assist accurate probe placement.^7–9 Researchers have also proposed motorized alignment devices to demonstrate their clinical feasibility. Morikawa et al.¹⁰ reported a floor-mounted device for MRI-guided therapy of liver cancers in a 0.5-tesla open-configuration scanner. It enabled a physician to freely select a probe insertion path while maintaining the remote center of motion (RCM) at the tumor site. Franco et al.¹¹ developed a pneumatically actuated table-mounted robotic system to assist physicians in MRI-guided percutaneous laser therapy for liver tumors with three cases.

In addition to the floor-mounted¹⁰ and table-mounted¹¹ motorized alignment devices, a body-mounted device has been proposed to address alignment errors caused by body motion. This “body-mounting” concept was originally presented by Monfaredi et al.¹² for shoulder arthrography. Here, the goal was to implicitly address misregistration of trajectories caused by body motion. Body motion, which is unavoidable in percutaneous cryotherapy for abdominal lesions,¹³ was addressed in an article from our group¹⁴ as well. Here, a motorized alignment device was used to achieve higher probe alignment accuracy in abdominal percutaneous cryotherapy than that achieved by manual alignments, even in the presence of body motion. However, the study¹⁴ observed that the advantage of the motorized alignment device for accurate probe placement diminishes when organ motion occurs in addition to body motion. This implies that the “body-mounting” concept mitigates body motion by locking the device to the body surface. Nonetheless, organ motion may still have to be addressed independently by tracking organs. However, no literature has been published to investigate if both or either body and organ motion compensation are required for accurate placement of cryotherapy probes in MRI-guided percutaneous therapy.

The objective of this study was to analyze the guidance accuracy of a body-mounted motorized cryotherapy probe alignment device (BMCPAD) and analyze the effectiveness of body motion compensation and organ motion compensation in a BMCPAD operation. In this study, we prospectively enrolled patients in MRI-guided cryotherapy of abdominal organs and performed all the steps up to and including the motorized alignment of the cryotherapy probe using a BMCPAD. From the images collected in these human subject studies, we simulated three scenarios. The first scenario is body motion compensation using fiducial markers embedded in the BMCPAD; the second is organ motion compensation by image-based identification of organ; and the third scenario is neither of them. We then tested the hypothesis that body motion compensation negates the necessity of organ motion compensation for accurate probe alignment in MRI-guided cryotherapy of abdominal organs.

2. MATERIALS AND METHODS

2.A. Body-mounted motorized cryotherapy probe alignment device

This study was approved by the Institutional Review Board. Consent forms were collected prior to the enrollment of each subject. The protocol included the placement of a body-mounted motorized cryotherapy probe alignment device (BMCPAD) on a patient’s body and simulated targeting of a cryotherapy probe. The BMCPAD was designed to maintain the geometrical relationship of the device with the patients’ body by strapping the device in conjunction with MRI surface coils to the patient (Fig. 1). The device is MR Conditional and does not obscure critical anatomy in MR images. Moreover, the operation of the device is not affected by MR imaging.¹⁵

Fig. 1.

Body-mounted motorized cryotherapy probe alignment device (BMCPAD) in use. The magnetic resonance imaging (MRI) surface coil (arrowhead) is first placed on the surface of the patient’s body. The top stage (arrow) is then placed over the coil and fixed with belts. After obtaining an MRI image with the device, the stick-shaped fiducial marker was aligned as a substitute for the cryotherapy probe.

The device consists of two ring-shaped ultrasonic actuators (Canon Inc., Tokyo, Japan) stacked one on top of the other. The top ring holds the probe mount and is marginally angled to realize a tilted cone-like workspace. Because the lower ring rotates in a circular motion, the top ring and its cone-like workspace also do so. This produces a larger cone-like workspace underneath the device. The probe mount can hold the cryotherapy probe as well as the fiducial wand (a stick-shaped fiducial marker). Figure 2 illustrates the mounting of the device on the MRI surface coil, and the probe mount on the top ring. Figure 3 illustrates the fiducial wand in the device.

Fig. 2.

Configuration of Body-mounted motorized cryotherapy probe alignment device (BMCPAD). (Left) (a) Magnetic resonance imaging (MRI) surface coil. (b) Jig with built-in fiducials for stabilizing the MRI surface coil. (Middle)(c) Top stage with two ring-shaped ultrasonic actuators to align cryotherapy probes. (Right)(d) Cryotherapy probe mount with fiducial markers for estimating probe alignment.

Fig. 3.

Images showing the fiducial markers built into the BMCPAD. The arrows indicate a stick-shaped fiducial marker (fiducial wand) used in physical aiming. The arrowheads indicate fiducial markers built into the housing of the device positioned in a shape of a ring. The fiducial wand was used to calculate the simulated position of the probe tip. The ring-shaped fiducial markers were used to co-register the position of the robot from the planning image to the confirmation image. (a) A photo image of ring-shaped fiducial markers embedded in the jig. (b) A design draft of the jig. (c) A reconstructed coronal image of the intraprocedural magnetic resonance (MR) image in Case 10. The fiducial marker is sliced in a cross-sectional direction and five of the ring-shaped markers are visible in this plane. (d) A photo image of the probe mount with a built-in fiducial wand. (e) A design draft of the probe mount. (f) An axial image of the intraprocedural MR image in Case 10. The fiducial wand is sliced in a longitudinal direction and one of the ring-shaped markers is visible in the image.

2.B. Preclinical MR compatibility test

2.B.1. Force and torque

The force induced to the BMCPAD was measured using a protocol derived from ASTM F2052–14¹⁶ and ASTM F2213–06.¹⁷ Specifically, the force was determined by hanging the BMCPAD from a string and measuring the angle of the string when placed at the isocenter, outside but immediately adjacent of the bore, and approximately 1 meter away from bore. The MRI scanner used for this experiment was a 3T scanner (MAGNETOM Verio, Siemens Healthcare, Erlangen, Germany). A translucent box was used as the structure to hold the robot. The string was threaded through a hole in the base of the robot and attached to the box. Protractor scales were attached to sides of the box and the angulation of the robot was measured by projecting the string upon the scale. Using the Gnu Image Manipulation Program, the angles of the strings compared to the horizon of the protractor was determined.

2.B.2. Heat

The heat induction in the MRI scanner due to BMCPAD was measured using methods derived from ASTM F2182–11a.¹⁸ The BMCPAD was placed next to a QC phantom (Siemens Healthcare, Erlangen, Germany) in the MRI with the bottom surface of the robot facing out of the bore. This bottom surface is the surface that will be closest to the patient’s skin during clinical use. The baseline temperature measurement was recorded before any imaging took place. Two clinical sequences HASTE and Volumetric Interpolated Breath-hold Examination (VIBE) were tested. All tests were conducted both with and without power being supplied to the robot. The temperature on the bottom surface of the robot was measured immediately after scans were performed using a infrared thermometer (Milwaukee 2267–20, Milwaukee Electric Tool Corporation, Brookfield, WI) from a distance of approximately three meters. The thermometer has a 10:1 distance to spot ratio, which means that the surface area that was measured was a circle with diameter of 30 cm. The thermometer traced the bottom surface of the robot and the maximum temperature was recorded. The thermometer was also used to measure the maximum temperature of the cables that protrude from the robot since they have the possibility of coming into contact with the patient’s skin.

2.B.3. Image artifacts

To evaluate the impact of BMCPAD on the MRI following ASTM F2119-07,¹⁹ the presence of image artifact and distortion tests were conducted in a 3T scanner (MAGNETOM Verio, Siemens Healthcare, Erlangen, Germany). A saline water bottle phantom (0.375% NiSO4 and 0.5% NaCl, 1900 ml, Model No. 8624186, Siemens Healthcare, Erlangen, Germany) was placed in the center of the scanner’s imaging volume. A Loop Coil (Siemens Healthcare, Erlangen, Germany) was placed between the robot and the imaging volume. Three system configurations that can be situated in the clinical procedure were configured to measure the impact of BMCPAD on MRI: (a) Baseline, where only the phantom is placed in the MRI scanner, (b) Powered off, where BMCPAD is placed and connected to the controller but no power is supplied, (c) Powered On, where BMCPAD is placed, connected to the controller, and powered on. Following ASTM F2119-07,¹⁹ the following two image sequences used in our clinical trial of MRI-guided abdominal ablation therapy was used for the signal-to-noise ratio (SNR) measurement: HASTE (TR/TE = 1000/200 ms; acquisition matrix = 320 × 196; flip angle = 147°; slice thickness = 4 mm; number of slices = 3; receiver bandwidth = 504 Hz/pixel) and VIBE (TR/TE = 4.6/1.6 ms; acquisition matrix = 320 × 240; flip angle = 9°; slice thickness = 4 mm; number of slices = 12; receiver bandwidth = 347 Hz/pixel). A total of 18 image data sets (each imaging sequence was conducted in the three cardinal scan directions under three configurations) were acquired, and artifacts, if any, were visually assessed. If artifacts were present, the size of the artifact was measured.

2.B.4. Image distortion

Image distortion was quantified by calculating the absolute percentage change in the phantom’s diameter from the respective baseline diameter per imaging sequence based on F2119-07.¹⁹ The reported measurements are the absolute maximum percentage difference in each diameter orientation for each configuration setup. As specified in F2119-07, four evenly spaced diameters (D1 to D4) at 45 degree intervals are used to determine distortion. The same imaging parameters for HASTE and VIBE imaging were used to measure while power to the robot was both on and off.

2.C. Early clinical feasibility study

2.C.1. Patient recruitment

The patients requiring MRI-guided percutaneous cryotherapy for abdominal lesions were recruited consecutively between May 2017 and March 2018 for enrollment. The period from January to December 2017 was reserved for maintenance of the BMCPAD. Therefore, no experimental procedure was performed in this period. The inclusion criteria of the study were (a) over 18 yr of age, (b) capable of providing informed consent, and (c) undergoes MRI-guided percutaneous biopsy or cryotherapy of a mass in the abdomen, pelvis or chest. The exclusion criteria of the study were (a) incapable of providing informed consent, (b) contraindicated to MRI, and (c) possibility of being pregnant. All the patients meeting these criteria were enrolled in the study during the period of the study. Case number five was aborted due to device malfunction at the start of the procedure.

2.C.2. Imaging sequences

All the studies were conducted in a wide-bore 3 T MRI scanner (MAGNETOM Verio 3T, Siemens, Erlangen, Germany). Throughout the present study, multi-slice images were acquired using a half-Fourier acquisition with single-shot turbo spin-echo (HASTE) (TR/TE: 1000/200 ms; flip angle: 147°; matrix: 320 × 190; field of view: 289 × 340 mm; slice thickness: 4 mm).The insertion points and the targets of the cryotherapy probe were defined prior to controlling the BMCPAD in the HASTE images, which were acquired in the initial planning phase. Both of these were used in the clinical procedure.

2.C.3. Device-guided probe alignment and image acquisition

A detailed workflow of the actual guiding procedure is provided as Supporting Information. The procedural sites were sterilized, and the physician targeted the lesion (once the patients were removed from the bore) using acquired MR images named “Planning Images.” The interventional radiologist (KT) planned the placement of multiple cryotherapy probes (by using the planning images) and then defined the probe entry points on the skin and target points in the tumors. The locations of the tips of the cryotherapy probes are referred to as “targets” hereafter. The interventional Radiologist also marked skin incision site using a sterile pen. The combined MRI surface coil and the base jig of the BMCPAD were placed coaxially to the marked incision site. See the far left image of the Fig. 2 to appreciate the combined MRI surface coil and the base jig of the BMCPAD. The tip of the aforementioned cone-like work space is designed to coincide with the skin incision site.

Subsequently, the BMCPAD was registered automatically to the image using the ring fiducial markers embedded inside the fixed jig portion of the device, which represented the position of the body surface, to estimate the alignment. An in-house software program based on 3D Slicer²⁰ calculated the angles of the two rings in the top stage to align the cryotherapy probe mount with the target. The MRI-compatible ring-shaped ultrasonic actuators then adjusted the rings to these angles once the physician signaled permission for its actuation. The second set of images was then captured and named “Confirmation Image.” Both planning and confirmation images included, the fiducial wand, the ring fiducial markers and target organs. The fiducial wand attached to the probe mount in the BMCPAD, not the actual placement of the cryotherapy probe, was used to measure the accuracy of guidance.

2.C.4. Simulated body and organ motion compensation

A retrospective simulation study was performed to assess the accuracy of probe guidance by varying the approaches for motion compensation. Specifically, the confirmation images collected from the MRI-guided cryotherapy using BMCPAD were processed to replicate the procedures with body motion compensation, organ motion compensation, and no compensation. The simulation was implemented with 3D Slicer.²⁰

The body motion compensation was performed by registering a virtual model of ring fiducial markers to the markers depicted in the MR images. Because the ring fiducial markers were attached to the patient’s body, this registration functioned effectively as body motion compensation. The output of this registration was a translation matrix indicating the motion of the device from the planning image to the confirmation image.

The organ motion compensation was performed using registered target organs in images. The organ of interest was tracked by co-registering the organs using an intensity-based rigid registration with maximization of mutual information. The outcome of the organ compensation was determined by the image registration, which enabled the estimation of the motion of the target within the organ from the planning image to the confirmation image.

2.C.5. Simulated reaiming using the motion compensation output

In the simulated study, we measured the guidance accuracy of probe alignment with three hypothetical scenarios using the outcome of motion compensation discussed in the previous section. In all the scenarios, we assumed that the true target position after the body and organ motions was at the new location measured by the organ motion compensation.

In the first scenario, the BMCPAD is aware of body motion (translation) by the body motion compensation. It reaims the probe mount to the original target position defined in the planning images.

In the second scenario, the BMCPAD is aware of the organ motion and adjusts the angulation of the probe mount to the new target position measured by the organ motion compensation under the assumption that the device is still in the original position in the planning image.

In the third scenario, the BMCPAD is not aware of either body motion or organ motion. Hence, there is no reaiming.

See Fig. 4 for a graphical representation of the simulated reaiming by BMCPAD.

Fig. 4.

Intra-bore re-aiming simulation schemes. Three simulation schemes with distinct motion compensation scenarios were compared. The ellipses with a solid and dotted line indicate lesions in the confirmation and planning images, respectively. The rectangles with solid and dotted lines indicate the body mounted device in confirmation and planning image, respectively. The dotted line connected to rectangles indicates the simulated probe. (a) Scenario with only body motion compensation, where the simulated probe tip reaims according to the body surface displacement, although the organ has already shifted to a new position. (b) Scenario with only organ motion compensation, where the simulated probe tip reaims according to the organ movement, although the device has already shifted to a new position with the body surface movement. (c) Scenario with no motion compensation, where the device shifts with the body surface while the organ also shifts, although no motion compensation is implemented. Note that the implementation of both the motion compensation yields a zero error in this simulation. Thus, it is not listed.

2.C.6. Data collection

We recorded the patient age, type of anesthesia used (monitored anesthesia care (MAC) or general anesthesia), organ site of the lesions, size of the lesions, and depth of the targets. The lesion size was measured as the largest dimension of a suspicious lesion on an axial image, as is the standard of care in Radiology practices. The depths of the targets were determined as the distance between the designated probe insertion point and the target in Planning Image. Note that target positions in both original Planning Image and in the simulated reaiming did not observe image artifacts due to cryotherapy probes, since the probe was not inserted at the point of planning or in the simulation.

Additionally, two sets of guidance accuracy were measured. First, the physical guidance accuracy of the probe alignment was measured using the fiducial wand attached to the probe mount and delineated in confirmation images. The central axis of the fiducial wand in the image was extrapolated to measure the distance from the targets to the extrapolated line (Metric #1 — physical guidance accuracy). The physical guidance accuracy was measured as the normal distance from the target location in the planning image to the trajectory established by the fiducial wand in the confirmation image.

The second set of guidance accuracy values was collected after simulated reaiming of the guide using the motion compensation. Here, the guidance accuracy was computed as the distance between the reaimed target positions from the trajectory produced by the device after motion compensation and the new target positions produced in the organ motion compensation (Metrics #2 simulated guidance accuracy). In the process of motion compensation, the magnitude of body surface movement and that of organ movement were determined and recorded.

The study data were collected and managed using a data capture tool: Research Electronic Data Capture (REDCap).²¹

2.C.7. Analysis

The primary analysis performed in this study was paired Wilcoxon signed-rank test to analyze the significance of the difference in simulated guidance accuracy among the three motion compensation scenarios (body motion, organ motion, and no compensation).

Second, multiple linear regression analysis was conducted to analyze the contribution of the background parameters (types of anesthesia, depths of the target, organ sites of target, magnitude of body surface movement, and magnitude of organ movement) to the simulated guidance accuracy. The number of parameters incorporated into the multiple linear regression model was reduced by minimizing the Akaike’s information criterion (AIC).²²

Third, Pearson’s correlation analysis was conducted to investigate the correlation between the magnitude of body surface and organ movement as well as the accuracy of the physical and simulated guidance.

In all the analyses, we derived the P-values and considered P-values of 0.05 or less to be statistically significant. Pearson’s correlation coefficient was calculated for the linear regression analyses. All the data analyses were conducted with RStudio Version 1.1.463 (RStudio Inc., Boston, MA).²³

3. RESULTS

3.A. Preclinical MR compatibility test

3.A.1. Force and torque

The results were tabulated in Table I. From the maximum angulation of the string hanging the guiding device (1.45°), it can be concluded that the MRI-induced force applied to the guide is nominal and not harmful to patients and clinical staff.

Table I.

Magnetic resonance imaging (MRI)-induced force measured by angulation of the robot hung by a string.

Location	Horizontal force (°)	Vertical force (°)	Lateral angle (°)
Bore entry	0.43	0.73	0.30
Out of bore	0.28	0.99	1.27
Isocenter	1.45	1.36	0.09

3.A.2. Heat

As indicated in the Table II, there was not a noticeable increase in temperatures in both the cable and the instrument guide, one can conclude that the device and the cable do not induce heat by MR imaging.

Table II.

Temperature measurements after magnetic resonance imaging (MRI) sequencing.

	MR sequence	Bottom surface temp (Deg C)	Cable temp (Deg C)
Robot power off	N/A (baseline)	21.0	20.7
	HASTE	20.9 (−0.1)	20.5 (−0.2)
	VIBE	20.9 (−0.1)	20.6 (−0.1)
Robot power on	N/A (baseline)	21.0	20.7
	HASTE	20.9 (−0.1)	20.7 (0.0)
	VIBE	20.9 (−0.1)	20.7 (0.0)

3.A.3. Image artifact

In all configurations of BMCPAD, namely (a) baseline, (b) powered off, and (c) powered on, no noticeable image artifact was found in any image. Therefore, no measurements of the size of artifacts are reported.

3.A.4. Image distortion

The image distortion induced by BMCPAD in the scanner was a maximum of 0.97% in HASTE imaging (power off), or 1.0 mm. In VIBE sequence, the percentage of distortion was a maximum of 0.74% (power on), or 0.7 mm. The results are tabulated in Table III. It is suggested therefore that VIBE is the preferred method of imaging to digitize targets and compute trajectory. However, HASTE imaging is still acceptable as MRI imaging using BMCPAD.

Table III.

Image distortion.

	HASTE				VIBE
Location*	0°	90°	180°	270°	0°	90°	180°	270°
Robot off	0.55%	−0.75%	1.21%	1.39%	0.33%	−0.01%	0.30%	−0.49%
Robot on	−0.03%	−1.23%	−0.10%	0.84%	−0.28%	0.06%	0.22%	−0.74%

^*Location indicates the angles of a circular function representing the surface of ACR MRI phantom (Verio 3T, Siemens Healthineers, Erlangen, Germany) in images.

3.B. Early clinical feasibility study

3.B.1. Subjects

Seventeen patients with lesion sites on organs including the kidney, liver, retroperitoneum, and diaphragm agreed to be enrolled in the study (Table IV). The mean age of the patients was slightly over 70 yr (range: 56–89+) yr. The study was aborted in one case (Patient #5) owing to a device malfunction. In the remaining 16 cases, 37 cryotherapy probe alignments were successfully attempted using the device. In these 37 sets of image acquisitions, the sizes of the lesions was 26.4 ± 11.7 mm and depths of targets were 100.5 ± 30.8 mm. The average number of probes assessed was 2.3 with a range of 1 to 4.

Table IV.

Patient characteristics.

Patient	Age (y)	Lesion size (mm)	Target depth (mm)	Number of probes	Organ site	Anesthesia
1	69	40.0	131.0	1	Left retroperitoneum	General
2	>89*	32.0	67.0	1	Right kidney	General
3	58	21.0	82.0	4	Left kidney	General
4	57	12.0	88.0	2	Left kidney	General
5**	76	43.0	41.0	1	Right kidney	General
6	70	31.0	65.0	1	Liver	General
7	70	11.0	84.0	2	Left retroperitoneum	General
8	70	30.0	56.0	1	Right diaphragm	General
9	70	16.0	114.0	3	Right kidney	MAC***
10	70	28.0	132.0	3	Left retroperitoneum	General
11	79	14.0	91.2	2	Left kidney	MAC
12	56	48.0	86.0	4	Left kidney	General
13	79	34.9	139.6	1	Left kidney	MAC
14	67	21.0	108.0	3	Right kidney	MAC
15	61	39.5	156.7	3	Liver	General
16	85	27.0	61.0	3	Left kidney	MAC
17	61	16.0	135.4	3	Right kidney	MAC

^*The patient age of over 89 is not displayed to achieve de-identification in accordance with the Health Insurance Portability and Accountability Act of 1996 (HIPAA) Privacy Rule.

^**Excluded owing to device malfunction.

^***MAC, Monitored Anesthesia Care.

3.B.2. Accuracy

The simulated guidance accuracy obtained with body motion compensation, organ motion compensation, and no compensation were 2.4 ± 2.9 mm, 2.2 ± 1.6 mm, and 3.5 ± 2.9 mm, respectively. The physical guidance accuracy was 13.4 ± 11.1 mm. Figures 5, 6 show illustrative cases where discrepancy between body and organ motion compensation was observed.

Fig. 5.

Case illustration (case #17) of a 61-yr-old patient. The site of the lesion was the kidney. The size of the lesion and depth of the target were 1.6 and 13.5 mm, respectively. The simulation based on the intraprocedural MR images is illustrated. The procedure was performed in the prone position under monitored anesthesia care (MAC). The image is rotated 180°. The yellow line indicates the position of the BMCPAD (projected on the plane) and the segmented organ in the planning image. Similarly, the device in the planning image was outlined with white. The blue line indicates the position of the BMCPAD (projected on the plane) and the segmented organ in the confirmation image. The device in the confirmation image was outlined with gray and filled with darker gray. (a) Macroscopic view indicates spatial relationship. The area surrounded by the white box is magnified in (b) and (c). (b) Magnified view of device placed on patient’s body surface. Note that the position of the device has varied significantly from the planning to confirmation image. (c) Magnified view of the organ. Note that the variation in the position is negligible. (d) The white cross, arrow, and arrowhead indicate the new target position in the confirmation image, tip position in the body motion compensation, and tip position in the organ motion compensation, respectively. The green dotted line delineates the lesion in this slice. The tip position of the body motion compensation is nearer (0.27 vs 3.91 mm) to the new target position, presumably because of the relatively large body surface movement compensated by body motion compensation, whereas it is not so in organ motion compensation.

Fig. 6.

Case illustration (case #16) of an 85-yr-old patient. The site of the lesion was the kidney. The size of the lesion and the depth of the target were 2.7 and 6.1 mm, respectively. The simulation based on the intraprocedural magnetic resonance images is shown. The procedure was performed in the supine position under monitored anesthesia care. The yellow line indicates the position of the BMCPAD (projected on the plane) and the segmented organ in the planning image. Similarly, the device in the planning image was outlined with white. The blue line indicates the position of the BMCPAD (projected on the plane) and the segmented organ in the confirmation image. The device in the confirmation image was outlined with gray and filled with darker gray. (a) Macroscopic view indicating spatial relationship. The area surrounded by the white box is magnified in (b) and (c). (b) Magnified view of device placed on patient’s body surface. Note that the variation in the position is negligible. (c) Magnified view of the organ. Note that the position of the organ has altered noticeably from the planning to confirmation image. (d) A white cross, arrow, and arrowhead indicate the new target position in the confirmation image, tip position in the body motion compensation (a projection of the position that is a slice away from this specific slice), and tip position in the organ motion compensation, respectively. The tip position of organ motion compensation is nearer (0.59 vs 4.28 mm) to the new target position, probably as a result of the relatively large organ movement compensated by organ motion compensation, whereas it is not so in body motion compensation.

3.B.3. Data analysis

First, the simulated guidance accuracy among the three motion compensation scenarios (body motion compensation, organ motion compensation, and no compensation) revealed that the body motion compensation aided in producing statistically significantly less guidance error compared to no motion compensation (P < 0.0001). The organ motion compensation also improved the accuracy of targeting, with P = 0.01056. However, body motion compensation and organ motion compensation produced similar guidance accuracies (P = 0.5112) (Fig. 7).

Fig. 7.

Individual value bar graph for simulated guidance errors in each simulation scheme. Left: body motion compensation, Middle: organ motion compensation, and Right: no motion compensation. Each data point depicts simulated guidance accuracy error. The bar indicates the median value for each group. The asterisk indicates a statistically significant difference. The paired Wilcoxon signed rank test revealed a significant difference between body motion compensation and no motion compensation group as well as between organ motion compensation and no motion compensation group.

Second, type of anesthesia, depths of the target, and organ sites of target did not contribute to the simulated guidance accuracy. However, the guidance accuracy exhibited correlation with both the magnitude of body surface movement (P = 0.04488 < 0.05) and organ movement (P < 0.0001) when no motion compensation was included.

Third, Pearson’s correlation test revealed a significant correlation between the magnitude of body surface movement and organ movement. The correlation between physical and simulated guidance did not demonstrate significance (P = 0.04707) (Fig. 8).

Fig. 8.

Scattergrams showing the correlation of each variable. Pearson’s correlation coefficient and the P values are indicated on the left-upper position of each graph. Left: Scatter plot for the magnitude of body surface and organ movements. The regression line is depicted according to the statistically significant correlation. Right: The scatter plot of guidance accuracy with simulated (ring fiducial markers) and physical guidance (fiducial wand). No statistically significant correlation was observed.

4. DISCUSSION

We analyzed the physical as well as simulated guidance accuracy of a BMCPAD and investigated the impact of the motion compensation on intrabore guidance, by utilizing the data sets acquired in the clinical study. Additionally, we investigated further to clarify which of the body and organ motion compensation affect the accuracy of probe guidance.

The accuracy of the cryotherapy probe guidance when using the device in this study was comparable, if not better, than those reported in previous studies that analyzed the accuracy of cryotherapy probe guidance in CT-guided interventions.^24–26 Krucker et al.²⁵ and Engstrand et al.²⁶ reported total errors of 11.9 ± 1.4 mm and 5.8 ± 3.2 mm, respectively, in their studies. Widmann et al.²⁴ separately reported a lateral (in-plane-oblique) error of 3.6 ± 2.5 mm and depth error of 7.4 ± 6.2 mm. In addition to the discussion on the accuracy, patient motion and depth were indicated as significant sources of error by Krucker and Widmann, respectively. As this study is the first attempt to measure the accuracy of MRI-guided percutaneous cryotherapy probe alignment by a body-mounted device, we cannot compare our results with those reported by Morikawa et al.¹⁰ or Franco et al.¹¹ This is because they were floor-mounted and table-mounted rather than body-mounted. However, the physical guidance accuracy errors based on the fiducial wand tended to be disproportionately larger compared to those from simulated guidance. A potential cause for this relatively large error is the low precision in localizing the fiducial wand embedded in the probe mount. The partial volume effect obscuring the fiducial wand at the slice thickness of 4 mm and limited spatial distribution of the fiducial marker may have resulted in suboptimal accuracy. Besides, we experienced certain issues with warping from distortion and/or distortion correction that may have introduced inaccuracy during the procedure. This was apparent when the device was very close to the bore (almost touching). Another possibility is the assembly error. The relatively large standard deviation (11.1 mm) in the accuracy of physical aiming and low correlation with simulated targeting errors may challenge the present methodology’s position as the gold standard in assessing physical guidance.

The simulated guidance accuracy errors in the two individual motion compensation were significantly smaller than that with no motion compensation. This implies that motion compensation improves the accuracy of guidance notwithstanding the absence of either body motion compensation or organ motion compensation. Meanwhile, the difference in guidance errors between the two individual motion compensation was not significant. This implies that the body motion compensation that we have implemented in the present system has a value that is statistically equivalent to that of organ-based motion compensation, which tends to be more complex and resource dependent.^27,28 Studies on mitigating organ motion are relatively common.^29–32 Studies utilizing body surface information to mitigate patient motion is almost de facto standard,³³ although studies comparing body and organ compensation are limited. Among these, Paganelli et al.²⁷ compared errors of motion compensation based on external motion parameters and internal motion parameters in the context of MRI-guided liver radiation therapy. In the study, the authors conclude that the external motion significantly correlated with the internal motion. This concept is consistent with the observation in our study. However, it should be observed that this is not always the case, as shown in the illustrative cases in Figs. 5 and 6.

In this study, the probe mount was designed to hold the fiducial wand for accuracy assessment. In our future study, BMCPAD will also function as a placement guide at the time of percutaneous probe insertion. In such a case, the cryotherapy probe will not be affixed to the BMCPAD.

The experimental protocol did not include the placement of the cryotherapy probe to continue the ablative procedure using BMCPAD as the sole method of probe guidance, as we had not established the accuracy and safety of the BMCPAD before the current study commenced. This is an acceptable cautionary measure for first in human study of medical devices per.³⁴ In fact, the current study helped us to establish guidance accuracy, as opposed to navigating accuracy, allowing us to test the navigating accuracy in the next phase, followed by ablative therapy using the BMCPAD. Here, the definition and difference of “guidance” and “navigation” can be found in³⁵ where “guidance” in Image Guided Therapy is to understand ‘where you go” before tool placements, and “navigation” is to understand “where you are” as you place tools in the surgical site.

The study we conducted with patients was an Early Feasibility Study defined in Ref.³⁴. An Early Feasibility Study is a clinical study involving a device in a relatively early development stage and can be used to optimize the device. As we attempted to evaluate the device design concept in a small number of subjects, the early feasibility study we conducted helped us to mitigate potential risks only found in a real clinical setting. Specifically, the study we conducted with real patients could assess not only the critical assessment of the guidance accuracy of the probe placement, but also safety, case-specific considerations, technical challenges associated with device usage in clinics, whether expected functionality can be achieved in a timely manner, and device failures in the presence of other medical devices. It was also critically important to assess the motion compensation of the patient undergoing anesthesia, as a volunteer without anesthesia would have presented different breathing and motion patterns than the real clinical patient. Finally, using real patients is clearly the most relevant way to reflect a typical patient (lesion) population in clinics. Ultimately, the rationale of the Early Feasibility Study was discussed by comparing the aforementioned benefit and risks to the patients. We concluded, after discussion with our Institutional Review Board, that the benefits would outweigh the risk, thus allowing us to execute the study as we discussed in this article.

To our knowledge, this is the first study to compare the impact of body motion compensation and organ motion compensation in a body-mounted robotic guide, using clinically obtained data. Our study is distinctive also because our data are collected from clinical cases. Moreover, a majority of the studies addressing organ motion addresses the motion of the liver or lung, which deforms intermittently albeit steadily. Abdominal organs situated apart from the diaphragm are different in terms of the mode of movement. Therefore, our study is unique. Moreover, the fact that the magnitude of body surface movement is correlated with the magnitude of organ movement can be leveraged as a safety measure to abort a procedure in case of a large body motion compensation error and to reinitiate the process. The results of the multiple regression analysis illustrate the covariate that influences each targeting error. It is of note that patient motion predicted the guidance accuracy the most.

With respect to the analyses of covariates in the multiple linear regression analyses, the type of anesthesia used has exhibited a significant difference between the use of general anesthesia and MAC in certain previous studies.^36,37 This appears consistent considering the patient motion factor. A feasible interpretation of our analysis result is that by mounting the device on the patients’ bodies, the likely increase in patient motion when not using general anesthesia can be mitigated moderately. Similarly, the depths of the target was considered to be a significant factor in certain studies,^38,39 although not in others.^26,40 The fact that depths of the target influence the accuracy is highly intuitive and therefore, is occasionally considered as a measure of success of study implementation. In our analysis, parameter selection with AIC minimization resulted in the exclusion of depths of the target from the final fitted model. We can interpret this phenomenon in multiple ways. First, erroneous factors related and unrelated to depths: angulation and translation error, implying that angulation can cause an error proportional to the depths of the target, whereas translation does not. Secondly, a mutual cancellation of two conflicting factors: error at the origin and error at the tip of the probe. This implies that the erroneous factor generated around the origin of the probe tends to be proportional to the depths of the target, whereas the erroneous factor generated around the tip of the probe tends to be inversely proportional to the depths of the target, mainly concerning the angulation factor. Moreover, a few previous studies obtained a similar result, which supports this study’s observations. We included the organ site of targets in the analysis considering its relation to organ motion compensation. Most of the studies covering organ motion compensation have limited their organ-of-interest to one, for example, liver.^29–30,41 Because our dataset includes various organ sites of targets in the abdomen, this is a noteworthy exploration. However, owing to the limited number of samples, we could compare only kidney with nonkidney. It could be more effective to compare each organ with a larger number of samples in the future. Engstrand et al. reported the targeting error of liver, lung, and kidney to be 4.7 ± 1.5 mm, 10.4 ± 2.0 mm, and 6.9 ± 3.1 mm, in which significant correlation was absent. The important implication of this study is that patient motion influences accuracy significantly when compared with other factors.

Considering safety, we adopted a conservative approach and did not insert a cryotherapy probe while using the device. The successful implementation of a body-mounted motorized device may also minimize the risks of the procedure, including tract seeding.⁴² by inhibiting unnecessary bouts of cryotherapy probe alignment. Also, this device has a small footprint compared to other floor-mounted¹¹ or table-mounted¹¹ motorized devices, which would facilitate its introduction into the already crowded MRI suite. The first step in the clinical translation of this device is to assess its alignment accuracy without performing cryotherapy probe alignments during patient procedures.

The limitations of this study include the fact that it is a simulation study, even if driven by clinically obtained data. Nevertheless, our study has a distinct advantage over a phantom, animal, volunteer, or cadaver study in terms of at least one of the following aspects: physiology, anatomy, epidemiology, and pathology. Our future goal is to proceed with the clinical use of this device in cryotherapy probe alignment procedures.

The proposed BMCPAD is expected to be useful in MRI-guided percutaneous cryotherapy in closed-bore scanners.^2–4 The device can also be extended to any percutaneous MRI-guided intervention of abdominal organs including, but not limited to: laser interstitial thermal therapy (LITT),⁴³ computer-tomography-guided high-dose-rate-brachytherapy (CT-HDRBT),⁴⁴ irreversible electroporation (IRE),⁴⁵ and MR-guided microwave ablation⁴⁶ that introduces the ablative probes to the lesion. Moreover, the BMCPAD can theoretically be used in nonablative procedures such as biopsy as long as it is using a needle-shaped apparatus, although adaptation to each apparatus should be achieved beforehand.

5. CONCLUSIONS

Our study results verify the feasibility of the BMCPAD in the context of accuracy and provide an insight that favors motion compensation with a statistically equivalent emphasis on body surface and organ component.