ProBot: Robot-Ultrasound Probe for Transrectal and Transperineal Prostate Biopsy

Abstract

Objective:

Over the past decade the incidence of prostate cancer (PCa) has been on a steady increase. Efforts to improve PCa outcome include targeted biopsy with magnetic resonance imaging (MRI) and precision sampling. Several biopsy devices are currently available to guide the biopsy with MRI-ultrasound fusion. They commonly use generic handheld ultrasound probes retrofitted for fusion biopsy. Robotic probe manipulation has the potential to reduce the required training, skills, and outcome variability among urologists. Rather than retrofitting an ultrasound probe, we developed a novel cohesive ultrasound-robotic probe (ProBot) that enables a novel needle insertion path.

Methods:

Tissue-probe contact frequently deforms the prostate contributing to fusion errors. To minimize deformations, ProBot uses a side-fire probe and its only motion is a robotic rotation about its axis, which can’t force the gland. An additional robotic angulation of the needle allows targeting any gland location. As such, only 2 degrees-of-freedom are required for prostate biopsy. However, the probe and robot must allow clearance to angle the needle, thus a special probe and Remote-Center-of-Motion (RCM) robot kinematics were developed. We present their design, ProBot prototype, and adapters for transrectal and transperineal needle access.

Results:

Pre-clinical tests showed sub-millimeter targeting errors and validated the sterilization process. With IRB approval, transrectal prostate biopsy was successfully performed in two patients.

Conclusion:

ProBot is a novel device for prostate biopsy (robotic, simple, compact, precise) with a successful pilot clinical evaluation.

Significance:

A precision, skill independent biopsy device can impact the management of PCa. Additional trials are required.

I. INTRODUCTION

The Cancer Statistics reveal a resurgence of prostate cancer (PCa) incidence in the United States (US) with a steady 3% yearly increase since 2015, and with 313,780 new cases and 35,770 deaths last year [1]. PCa is the most commonly diagnosed type of cancer and the second leading cause of cancer related death among US men. Efforts to address the rise in PCa diagnosis without over detection and overtreatment include more targeted biopsy techniques for clinically significant PCa (csPCa) using magnetic resonance imaging (MRI) and precision targeted biopsy [2].

The best estimate of PCa aggressiveness is the Gleason score obtained from core needle biopsy [3], commonly obtained under transrectal ultrasound (TRUS) guidance. But ultrasound suffers from poor sensitivity and specificity for PCa. As such, systematic biopsy (SB) intends to obtain tissues that provide a representative sample of the overall gland. Moreover, freehand biopsy is highly inconsistent and subjective [4], leaving large regions of the prostate unsampled, often missing csPCa. Targeted biopsy (TB) emerged in response. Multiparametric or biparametric MRI [5] may show cancer suspicious regions of interest (ROI) for biopsy. MRI-TRUS registration (fusion) biopsy has shown improved csPCa detection [6].

Still, in combined TB and SB procedures, TB cores often miss csPCa detected by SB cores [7]. Errors include manual targeting and MRI-ultrasound fusion errors. In addition, csPCa may also evade MRI detection, therefore SB cores are still sampled alongside the TB cores.

Guided by TRUS imaging, the needle path may be transrectal (TR) or transperineal (TP) with the latter potentially reducing infectious complications [8] and gaining popularity especially in Europe [9]. While studies have not yet shown a difference in infection and cancer detection rates between TR and TP [10, 11], some suggested the potential omission of antibiotic prophylaxis in TP [12]. Both biopsy approaches remain clinically viable and safe [13].

Ultrasound probes are commonly operated manually and known to deform the prostate gland due to uneven pressure. Deformations cause ultrasound artifacts that complicate the fusion with MRI. Manual probe handling also results in significant inter- and intra-urologist variability [4].

Currently available fusion biopsy devices include the Artemis (Eigen) [14], BioJet (DK Technologies), BiopSee (MedCom), RVS (Hitachi/FujiFilm), UroNav (Philips), bkFusion (BK Medical), Navigo (UC Care), FocalBx (Focal Healthcare), and UroStation (Koelis). Large inter- and intra-urologist cancer detection rates (CDR) variability (12%–57% [15]) exists, suggesting that urologist’s training and skills remain critical even with fusion biopsy devices. Robotic hands-free probe handling renders a skill-independent biopsy potentially reducing variability among urologists. Only one commercial biopsy robot exists, Mona Lisa (BioBot Surgical, Singapore) [16]. Mona Lisa biopsy predominantly requires general anesthesia and an operating room setting [17].

We have previously developed a robot for prostate biopsy, the TRUS-Robot [18]. This device currently undergoes a randomized clinical trial [19] and was licensed by industry [20]. However, it uses a standard TRUS probe and is limited to the TR approach.

ProBot includes new mechanisms that aim to improve prostate biopsy technology with a simple and compact, multifunctional, robotic, skill independent device.

II. METHODS

A. Prostate Deformations Caused by the Ultrasound Probe

TR prostate biopsy is typically performed with an end-fire TRUS probe and needle parallel to the probe (Figure 1a). To image, the probe must maintain acoustic coupling in contact with the rectal wall (air gaps prevent imaging). In a human hand, the contact pressure is variable, resulting in variable prostate deformations that are captured in each gland section image. The deformed sections skew the rendered 3D gland shape and complicate fusion to MRI.

Figure 1:

Schematic of prostate biopsy with end-fire probe

The problem persists if the needle-guide has an angle relative to the probe. Since the angle is fixed, the probe must still be tilted to aim various prostate targets [21], thus variably pressing against the prostate. With end or side-fire probes, studies have found no significant difference between CDR [22].

Elastic registration methods attempt to correct deformation errors, but the models used are not patient specific [23]. Meta-analyses comparing rigid vs. elastic registration identified no significant difference at detecting csPCa [24]. Moreover, elastic registration is often performed at the initial registration only, missing to account for deformations at biopsy sampling.

Prostate deformations impact fusion, TB, and SB targeting, and so minimizing deformations is critical for accurate biopsy.

B. ProBot Degrees of Freedom, Needle Path, and Prostate Deformations

ProBot uses a linear side-fire array and the needle path is variably angled (Figure 2). To scan in 3D, the probe is rotated about its axis (R1) to acquire multiple TRUS slices throughout the gland. Then, any 3D point can be targeted by rotating the probe (R₁) and angulating the needle (R₂). The rotation of the probe (R₁) does not change the compression of the gland, preserving its shape throughout scanning and targeting. Even though patient motion may still deform the gland, the ProBot approach mitigates deformations caused by the probe motion, and may improve the overall accuracy.

Figure 2:

ProBot side-fire linear array and oblique needle path: a) Transrectal and b) Transperineal needle paths.

The same kinematic principle applies to the TR and TP needle paths. Figure 2a shows the TR path, where the pivot of the needle is close to the probe. For the TP path, the pivot point is shifted close to the perineal site, as shown in Figure 2b. The same R₁, R₂ rotations allow TR or TP biopsy, and the configuration can be simply selected with needle-guide adapters (later shown in Figure 13ab).

Figure 13:

Needle guides for a) TR and b) TP biopsy, and c) ProBot probe in a Trophon HDL machine

To angle the needle, however, a special probe body is needed to clear the space for the needle (with a regular probe the dotted part of the needle in Figure 2 interferes with the body).

C. The ProBot Probe

The ProBot probe is built around a long side-fire linear ultrasound array. Its body clears the space to angle the needle by having an offset shaft (Figure 3a). A metallic rib is included to augment its structural strength over the offset half-shaft (Figure 3a). In operation, needle adapters are placed over the half-shaft to refill it to a round full-shaft (later in Figure 10a).

Figure 3:

ProBot probe: a) Geometry clears space to angle the needle, b) Metallic rib included within the probe for structural strength

Figure 10:

ProBot a) prototype with TR needle-guide adapter, b) Control box

Handheld probes have ergonomic geometry and are difficult to mount precisely and repeatedly in a robot. Instead, the ProBot probe includes a magnetic metallic mount and two positioning pins for quick and precise attachment (Figure 3), even over a sterile bag.

D. Probe Rotation with Remote Center of Motion Mechanism

Rotary mechanisms are commonly centered on the rotation axis. To clear space for the needle, such mechanism would need to be placed distally, further away on the probe tail (i.e. Artemis, Eigen [14]). To reduce size and improve stiffness, ProBot uses a Remote Center of Motion (RCM) mechanism (Figure 4). We used our cable-driven compact RCM mechanism [25–27] with a new harmonic drive actuator [28]. Figure 4a also shows the magnetic mount base of the probe. In Figure 4b, 7 superimposed positions of the mechanism (−90° to 90° in 30° steps) show how the probe is rotated about its axis by the RCM. The motion is shown in the supplement movie.

Figure 4:

ProBot uses a compact RCM mechanism to rotate the probe: a) RCM and Probe, b) Superimposed positions at 7 RCM angles

E. Needle-Guide Angulation Mechanism

The angulation of the needle (R₂) is implemented with a 4-bar 1 degree-of-freedom (DoF) mechanism comprising the needle-guide, a connecting link, and an arm connected to a driver, as shown in Figure 5a. To minimize size at this close to patient location, we oriented the motor along the probe axis and used a cable transmission over pulleys and a spool driven by the motor (see driver detail is shown in Figure 5b). The driver also includes a ball type lock for simple connection/release of the needle-guide arm (sterile components). The same driver is used in both the TR and TP applications.

Figure 5:

ProBot needle-guide actuation mechanism (R₂): a) Overall and b) Detail of the Needle-guide driver mechanism

F. Transrectal and Transperineal Adapters

As shown schematically in Figure 2, choosing a TR or TP application can be simply achieved by shifting the pivot point of the needle-guide. To use the same driver, we shift the entire needle-guided mechanism consisting of the 4-bar linkage and driver mechanism as shown in Figure 6. For the TR approach, the driver is fixed in place (Figure 6a). For the TP approach, the needle-guide mechanism is slightly lifted and positioned adjustably along the probe axis to account for patient variability in the location of the perineum (Figure 6b). The needle-guide pivot is connected to the driver with a link. Adjusting the location of the driver also adjusts the pivot, and both are locked in place with the lever lock (see 4 red arrows).

Figure 6:

RCM hidden for brevity. ProBot configuration for: a) Transrectal and b) Transperineal biopsy

Both needle adapters include a semicircular body that fills the opposite side of the probe shaft, which must be round about the rectal sphincter (shown later in Figure 13ab).

G. 3D Image Scan and Ultrasound to Robot Calibration

In image-guided interventions, image-to-robot registrations are normally performed for every case. An advantage of coupling an ultrasound probe with a robot (or position tracker) is that the registration of their spaces is invariant, and is set by a calibration procedure [29]. A simple calibration rig with holes that guide a string (0.36mm) was made for calibration (Figure 7a). The rig attaches to the probe. In a water tank the strings appear as dots in ultrasound. Their image and known positions are used to measure the scale and position of the linear array relative to the probe and therefore robot.

Figure 7:

a) Calibration rig on probe, and b) Detailed view

If the holes and string have the same diameter, the string is difficult to pass through, and so the holes should be made larger. But then the string is no longer centered on the hole making it inaccurate as a marker. Figure 7b shows how larger holes were offset by design in the opposite direction of winding the string, so that the string is at the correct location when set tangent to the left or right of the holes by tightening it.

With ProBot, a 3D scan is performed by sweeping the prostate gland side-to-side with parasagittal 2D ultrasound from the side-fire array. Image-position pairs are recorded while rotating the probe about its axis (R₁) and recording images and their respective angles from the axis encoder. Ultrasound images from the side-fire probe are oriented radially on the axis of the rotary scan. Therefore, image pixels distal from the axis are less dense in 3D. The speed of the rotary scan is calculated based on the desired 3D resolution $\rho [\text{mm}]$ at a radius $R[\text{mm}]$ measured from the probe axis, as:

$$V_1=180\rho f/\left(\pi R\right)\left[°/s\right]$$ (Eq. 1)

where $f$ is the framerate of the image acquisition [frames/s] and is set and typically limited by the ultrasound machine. The pixel-size calibrated images are represented in place in 3D based on the angles where they were acquired, kinematic robot parameters, and then rendered volumetrically. As such, a point selected in ultrasound is already in robot coordinates, ready to target.

H. Prostate Coordinate System, Prostate Segmentation, and MRI-Ultrasound Fusion

ProBot inherits image navigation methods and software from our TRUS-Robot [18] that is currently used in a randomized clinical trial [19]. The fusion uses a novel anatomic landmark approach, the Prostate Coordinate System (PCS) [18]. The landmarks are the apex and base of the prostate. Figure 8a shows the apex (A) and base (B) in a central sagittal view of the gland. A and B selection follows several refinement steps: 1) Select A and B points in original ultrasound slices; 2) Refine their locations in reconstructed axial slices and orient the Anterior-Posterior direction; 3) Refine the A and B location in a coronal view; 4) and then in sagittal. The origin of the PCS is located at the center of the AB segment and the direction of the PCS follows the standard LPS (Left, Posterior, Superior) anatomic system of the DICOM standard.

Figure 8:

Prostate Coordinate System (PCS)

The PCS enables segmentation in axial, sagittal, and coronal views thus circumventing the common difficulty of segmenting the ends of the prostate in just one view [30]. We developed a wireframe prostate model alike Earth’s longitude and latitude coordinates, with the North and South poles at the apex (A) and base (B), as shown in Figure 8. The model includes 26 (2+3*8) control points that are automatically placed and manually refined in all 3 views.

Ahead of the biopsy procedure, the PCS and segmentation are completed on the MRI of the patient, together with defining the TB targets. At biopsy, the PCS is set in 3D ultrasound. MRI-ultrasound fusion automatically superimposes the two PCS and may be manually adjusted. This also fuses the segmentation and TB targets. For the SB plan we use an optimization method that personalizes the plan for the patient based on a “capsule model” [31, 32].

I. ProBot Positioning and Clinical Setup

The ProBot robot and probe are supported by an adjustable passive arm (Figure 9). The schematic in Figure 9 shows a man in the common left lateral decubitus position for TR biopsy. The probe is placed manually, positioned to show a central sagittal image of the prostate, and then the Support Arm is locked in place. Manual axial adjustment of the probe can be made from a slider and lock mechanism. All subsequent operation of the probe is robotic. A 3D scan is acquired (Sec. II G). The TB and SB plan is made (Sec. II H). For each biopsy site, the robot automatically rotates the probe (R₁) and angulates the needle-guide (R₂) to align the needle-guide on target. The physician inserts the needle through the needle-guide under live ultrasound feedback and takes the biopsy. A microphone included within the base of RCM listens for the biopsy needle firing sound and automatically saves the ultrasound.

Figure 9:

Setup for biopsy with patient in left-lateral decubitus position

A TP biopsy follows similarly with the exception that in the initial positioning the TP adapter is manually advanced towards the perineum. The common TP patient position is lithotomy, however ProBot may facilitate a lateral decubitus approach.

III. RESULTS

A. ProBot Prototype

The materials presented resulted from years of research. The initial invention of a cohesive robot-ultrasound probe [33] was derived from an MRI-guided robot [34, 35]. Then, a first ProBot prototype was built and preclinically tested. The Food and Drug Administration (FDA) determined its clinical investigation as a nonsignificant risk (NSR) device study. However, probe sterilization was challenging and updates were made into the second ProBot prototype that we present herein. The IRB approved the study.

Figure 10a shows the current ProBot prototype. Parts were manufactured at our Urology Robotics Laboratory with Computer Numerically Controlled (CNC) machines and 3D printing (FormLabs Inc., Somerville, MA). The ProBot probe was built with a commercial ultrasound linear array, the side-fire array of the EUP-U533 probe of Hitachi Medical Systems. Other long side-fire arrays may be used. The robot has a small size and together with the probe weighs only 1.3Kg.

B. Robot System, Control, and Software

Dedicated robot control electronics were built using EPOS controllers (Maxon Motor Ltd.) and talk to a PC over USB. Two Maxon 329041 EC motors are used. Their 3 hall-effect sensors are fail-safe (unlike quadrature encoders [36]) mitigating the need of redundant encoding.

The robot control box is shown in Figure 10b. Safety features include a software-hardware watchdog [37], emergency stop buttons, and visual alerts. The software runs on a Windows 11 PC, was developed in C++ (Visual Studio, Microsoft Corp.) with open-source Visualization Toolkit (VTK), Insight Toolkit (ITK), Grassroots Digital Imaging and Communications in Medicine (GDCM, DICOM), and EPOS libraries. The software includes robot control, ultrasound scanning, 3D image rendering, biopsy planning, and ultrasound-guided needle targeting components. The user interface shows a virtual clinical environment that includes the robot following in real-time the actual robot together with volumetric and real-time ultrasound (see movie).

C. 3D Ultrasound Scan, and Targeting Tests in a Water Tank

The speed of the rotary scan is automatically set in software (Eq. 1). For example, to achieve 1mm 3D resolution at 50mm over the probe radius (11mm, R=50+11) with 10 [fps], the scan speed is 9.4 [°/s], completing a 90° scan takes approximately 10s, and records 95 images. The ultrasound screen is captured at 1280 × 1024 [pixels]. The scale depends on the ultrasound machine zoom factor, for example 0.13 [mm/pixel].

A rectangular grid of strings 0.36mm in diameter, equally spaced 10 mm apart, and submersed in a water tank was 3D scanned with the robot (Figure 11a). A re-slice of the rendered volume through the grid plane is shown in Figure 11b. This shows accurate grid spacing, measured at 10.024mm. Then, the 25 crossings of the grid (green dots) were targeted robotically, as in Figure 11c. The robot automatically aligned the need-guide on each target. A ceramic rod (1.22mm diameter, close to a 1.27mm 18Ga needle, less ultrasound artifacts than metal) was inserted manually through the guide, and observed visually if it touched/pushed the strings. The experiment was performed with the TR and TP adapters and repeated 10 times at different depths. In all experiments the needle point touched the strings (see movie). As such, errors were no more than the sum of the sting and needle radii, (0.36+1.22)/2=0.79 mm, with either the TR or TP adapter.

Figure 11:

3D scan and needle targeting in a water tank with TR adapter: a) Setup, b) 3D scanned grid of strings, c) Needle point at a grid crossing

D. ProBot Targeting Tests in Prostate Mockups

Needle-targeting experiments were carried out with ProBot on prostate mockups (Model 066, CIRS, VA). The mockup was 3D scanned, the PCS was assigned, the prostate was segmented, and a biopsy plan was defined to target ultrasound visible lesions of the mockup (Figure 12). Sequentially, the robot oriented the needle-guide on each target and an 18Ga needle was inserted manually. Targeting error was measured as the shortest distance between the biopsy needle direction and the target point, as shown in Figure 12c. At TR and TP simulated biopsies over 27 and 30 targets, average errors were 0.30mm (SD 0.19mm) and 0.41mm (SD 0.22mm), and maximum errors were 0.67 and 0.92mm, respectively.

Figure 12:

Needle targeting in a prostate mockup with TP adapter: a) Test setup, b) 3D virtual environment t, c) Ultrasound image

E. Probe Disinfection and Sterile Single-Use Components

In operation, the robot and supporting arm are covered with a sterile bag (EZ-28, Preferred Medical Products, LLC). The probe attaches to the robot over the bag with the magnetic mount. Then, a needle-guide adapter attaches to the probe. According to FDA guidance [38] the ultrasound probe must be processed with high-level disinfection (HLD) prior to each case. We use Hydrogen Peroxide HLD with a Trophon2 (Nanosonics Inc.) equipment (Figure 13c).

Needle-guide adapters (Figure 13ab) are single use components with autoclave sterilization (Hot Steam, Pre-Vacuum cycle, temperature 132°C, exposure time 4 min, dry time 15 min). The materials used are:

• FormLabs resin Surgical Guide: Class I biocompatible (ISO 10993-1:2018, EN ISO 10993-5:2009 Not Cytotoxic, ISO 10993-10:2010/(R)2014 Non Irritation, ISO 10993-10:2010/(R)2014 Not a sensitizer)

• FormLabs resin Tough 1500: Biological evaluation of medical devices - Part 5: In vitro cytotoxicity evaluation - no observed cytotoxicity (NAMSA ISO 10993-5).

• Stainless steel 304/316 tubes with certificate of ASTM A908, Fed. Spec. GG-N-196 compliance;

FormLabs provided autoclave sterilization data for both resins, and the IRB Cleaning, Disinfection and Sterilization Oversight Committee (CDS) approved processing. However, HLD of the probe involved massive additional testing.

The probe is also made Surgical Guide resin. While the manufacturer tested and certified this resin for autoclave sterilization, they did not for hydrogen-peroxide HLD, so we had to conduct extensive testing. In short, we:

1. Made material specimens according to ISO 527-2 (Plastics - Determination of tensile properties)

2. Shipped specimens to Nanosonics Inc (Sydney, Australia). NanoSonics processed the specimens (Hydrogen Peroxide 35% w/w; Temperature of the Disinfection Chamber 60°C; Nebulising Function Time 30s ON / 30s OFF; 30s drive stop; Hydrogen Peroxide Delivery Rate 1.8 g/cycle) for 410 and 1053 cycles.

3. Nanosonics mailed samples backs, and we conducted dimensional and mechanical property testing on a Criterion Electromechanical Test Systems, Model C43-504 (MTS Systems).

4. Sent new specimens to Australia and Nanosonics performed HLD efficacy testing of Trophon2, ISO 527-2. The efficacy was validated against the Mycobacterium terrae microorganism (ATCC15755).

The tests results showed that:

1. A material discoloration was observed after 410 cycles.

2. No appreciable deformations or cracking were observed after all 1053 cycles.

3. Loss of mechanical properties of the test material in 410 hydrogen peroxide cycles are equivalent to those of 5 autoclave cycles.

4. However, between 410–1053 cycles the material becomes brittle.

5. The worst-case Trophon disinfection process (lowest dosage, minimum concentration of disinfectant) reduced the inoculated Mycobacterium terrae ATCC 15755 bioburden on the test item (specimen ISO 527-2) by ≥ 6.0 log10 cfu. The specimen ISO 527-2 has passed the simulated use test for Mycobacterium terrae ATCC 15755.

Overall, Hydrogen peroxide HLD is effective on the Surgical Guide material and is less impactful on mechanical properties than autoclaving. However, it is recommended that the number of sterilization cycles should not exceed 410, which was sufficient for a prototype.

Based on these results, and that the Hitachi array was already cleared for HLD, the CDS approved processing the ProBot probe in the Trophon HLD machine and the IRB approved the study.

F. First-In-Man Clinical Trial

Two TR biopsies were recently performed at the urology clinic under local anesthesia. Figure 14a shows ProBot with the patient in the left-lateral decubitus position. Figure 14bc shows ultrasound images with the needle inserted on the novel oblique path that ProBot enabled. ProBot’s kinematics maintained the needle within the imaging plane making it visible in its length, as shown in Figure 14cb.

Figure 14:

a) ProBot during biopsy procedure, b) and c) Side-fire images with angled needle path unique to ProBot

Ultrasound images were recorded when the biopsy gun was fired, triggered by the microphone. Inadvertent captures occurred, such as at the time when the urologist demonstrated the sound of the biopsy gun to the patient.

Prostate volumes were 45.7 and 55.1cm³, the total duration of the cases were 20.07 & 15.68min of which 0.47 & 0.57min for the 3D scan, 4.35 & 6.13min for biopsy planning, and 8.9 & 5.85min for the actual biopsies, and 18 and 15 samples were collected with the robot for the two patients, respectively. No prostate deformations nor device-patient interference were observed, and there were no complications.

IV. DISCUSSION

We report a novel biopsy device and development to the stage of initial clinical trial. The device includes several novel features including fully robotic handling with minimal DoF, simplicity and compactness, potential application to TR and TP biopsies, and mitigated probe-induced prostate deformations.

Robots were made in academia for ultrasound guided prostate biopsy [39]. With few exceptions, these are in preclinical validation stages [18, 40]. The ProBot approach uses only 2 DoF which is the minimum possible, the other robots use between 4 and 9 DoF [40–49]. Systems with end-fire probes require at least 4DoF, such as our TRUS-Robot [18] and Artemis system (Eigen) [14]. Reducing DoF offers simplicity and, possibly, increased precision. Moreover, probe induced deformations between 3D imaging and biopsy are mitigated by design by restricting probe’s motion to rotation about its axis. Indeed, patient motion may still deform the gland, which is to be determined clinically.

Kinematically, to the best of our knowledge, the use of an RCM mechanism to rotate a slender instrument about its axis is novel. RCM mechanisms were developed for laparoscopy [36] and are commonly used to orient the instrument laterally about a fulcrum point on their axes [25], such as the laparoscopy port. Another example is the Artemis (Eigen) [14] device that uses an RCM to pivot the probe laterally about the rectal sphincter. The new use of a RCM may extrapolate to other devices, such as catheter manipulation.

Several prostate biopsy systems are commercially available (Table 1). Mona Lisa [16] is the only commercial robot for prostate biopsy. Its fundamental deficiency is that it angles the needle relative to the imaging planes. As such, the needle is typically invisible in ultrasound during insertion. Typically, the needle appears only after firing the biopsy needle (a dot in the transversal plane) raising targeting and safety concerns. In contrast, ProBot always keeps the needle within the image plane (Figure 14bc).

Table 1:

ProBot and current commercial prostate biopsy devices features table

Name	Country	Needle Path	TRUS Probe				Needle visible in Ultrasound	Fusion method	Minimize prostate deformations		SB Plan		Skill independent	In Clinic (not OR)
			Type	Support	Tracking	Handling			At 3D scan	At Biopsy	Generic	Optimized
ProBot	US	TR & TP	Side-fire 2D	Robotic			Y	PCS	Y	Y	Y	Y	Y	Y
Artemis, Eigen	US		End-fire 2D	Passivearm	Encoders	Manual		Segmented MR to Segmented ultrasound	Y	N	N	N	N
UroNav, Philips	Netherlands		End-fire 2D	Handheld	Magnetic				N		Y
RVS, Fujifilm	Japan		Bi-planar 2D								N
bkFusion, BK Medical	Denmark		Bi-planar 2D					Manual alignment
Trinity, Koelis	France		End-fire 3D		Image based			Segmented MR to Segmented ultrasound
		TP	Side-Fire 2D	Fixture					Y
Mona Lisa, BioBot	Singapore		End-Fire & Side-Fire 2D	Robotic			N			Y	Y		Y	N

The next closest biopsy device is Artemis [14] because it uses a mechanical arm (not motorized, therefore not a robot). The arm is encoded and offers essential support for the probe. Like ProBot, Artemis can maintain uniform pressure over the gland at the time of the 3D scan to reduce prostate deformation. At biopsy, however, maintaining the pressure is more difficult, requiring skill, as with the other manual devices.

Native 3D ultrasound probes use a multitude of 2D arrays or move a 2D array internally to scan. To fit more components within the limited space of the probe, the size of their ultrasound array is typically smaller than that of 2D probes, reducing in-slice image quality. Instead, most biopsy devices scan with external motion of a 2D probe. The probe is handheld and position tracked. Scanning robotically can make the scan more uniform than manually and improve 3D imaging. Future clinical trials may reveal if improved imaging correlates with biopsy outcomes.

The probe-attached needle-guide with a variable angle is novel. At TR or TP biopsies, probes commonly use a needle-guide that is fixed to the probe (i.e. TP adapter, Perineologic, Cumberland, MD). As such, the probe needs to tilt between biopsy targets and is likely to deform the gland. ProBot’s controlled needle angle can target without tilting the probe.

Other side-fire probes with angled needle-guide exist but their angle is fixed (i.e. BK [50]). In this regard, the ultrasound images in Figure 14bc are original, depicting the side-fire / para sagittal ultrasound and variable needle angulations.

Previous ultrasound guided prostate biopsy robots apply to either TR [18, 49] or TP [16, 41, 44, 47, 48] biopsy. Unique to ProBot is its applicability to handle both approaches by simply changing a needle-guide. To date, only the TR biopsy has been clinically tested. Future research will expand to TP biopsy.

We have also extensively tested the 3D printing resin Surgical Guide of FormLabs. Having 3D printing capabilities of ultrasound probes is essential for prototyping, for us and possibly other researchers in the field.

Finally, the ProBot robot uses manual segmentation methods. In the two procedures performed, biopsy planning took 4.35min and 6.13min. These could be shortened and improved by automation, using recent developments in artificial intelligence (AI) [51]. Ultimately, AI driven robots could lead to fully automated procedures.

V. CONCLUSIONS

We report the development of a new side-fire ultrasound probe and simple robot that enables hands-free operation on a novel controlled angle path for TR and TP biopsies. We show that prostate biopsy can be performed with only 2 DoF, the minimum number, never used before. An essential feature is that the only motion of the probe is a rotation about its axis, which mitigates by design the deformations of the gland caused by probe motion. Also, the use of an RCM to spin about the slender axis of an instrument is novel.

Side-fire ultrasound provides high image quality and ProBot always keeps the needle in the image plane. Hands-free robotic probe operation gives additional potential to improve targeting accuracy, reduce the skill required and variability of outcomes among physicians.

TR biopsy was successfully completed in two patients. Safety and feasibility trials for TR and TP biopsies will follow. Clinical efficacy trials are needed to establish if the technology features improve clinical outcomes.