Robotic system for prostate brachytherapy

Abstract

In contemporary brachytherapy procedures, needle placement at the desired target is challenging for a variety of reasons. A robot-assisted brachytherapy system can potentially improve needle placement and seed delivery, resulting in enhanced therapeutic outcome. In this paper we present a robotic system with 16 degrees of freedom (DOF) (9 DOF for the positioning module and 7 DOF for the surgery module) that has been developed and fabricated for prostate brachytherapy. Strategies to reduce needle deflection and target movement were incorporated after extensive experimental validation. Provision for needle motion and force feedback was included in the system to improve robot control and seed delivery. Preliminary experimental results reveal that the prototype system is sufficiently accurate in placing brachytherapy needles.

• Robot
• prostate brachytherapy
• needle insertion
• transrectal ultrasound

Introduction

In modern brachytherapy procedures, the needles are inserted transperineally under the guidance of transrectal ultrasound (TRUS) images (Figure 1). Both the needle and TRUS are operated manually, with the seeds being deposited using a manual applicator. The needles are inserted through fixed holes in a physical template, severely limiting their flexibility and maneuverability during insertion. This sometimes makes it difficult to avoid the pubic arch, especially in patients with large prostates. The consistency and efficiency of the treatment procedure are highly dependent on the clinicians involved. To assist the clinicians, it was proposed to develop a motorized semi-automated robotic system for prostate brachytherapy. The robotic system would not only place the needle and deliver the seeds with accuracy, but could also help less skilled or inexperienced surgeons perform the procedures with higher consistency and efficiency.

Figure 1. Conventional brachytherapy procedure (Seattle method).

Several research groups have developed robotic systems for prostate brachytherapy procedures [1–8]. The needle placement robot system developed by Fichtinger et al. [1] and Stoianovici et al. [2] comprises a Cartesian bridge with 3 degrees of freedom (DOF) over the patient, a 2-DOF remote center of motion (RCM), and a 1-DOF needle inserter with a motorized driver using an axially loaded friction transmission. A 7-DOF passive arm is employed between the Cartesian stage and the other two modules (the RCM and needle inserter) to position and orient the needle in imaging instruments. Although the stages have sufficient encoders, the needle driver lacks precise encoding of the depth of needle insertion, which is performed by a friction drive. In this system, seeds can only be deposited manually. Wei and colleagues [3–5] have developed a robotic system for prostate brachytherapy in which an industrial robot is used to position and orient a single-hole template through which a needle can be inserted manually. A separate motorized device was developed to rotate the needle, and another module is used to operate the US probe. The needle placement robot system designed by Kettenbach et al. [6], [7] consists of two offset x-y stages which allow positioning and orientation of the needle over the perineum with 4 DOF (two translational and two rotational). The needle can be inserted manually through the needle guide by the clinician, and the depth of insertion is monitored on US images.

When the patient is in the lithotomy position for transperineal prostate brachytherapy (Figure 2), the available workspace for the robot is quite limited. Thus, most industrial robots may lose dexterity (or lose DOF/encounter singularity) when working in the severely constrained workspace in the operating room (OR). Currently, hardly any robotic/mechanized system is available that can provide the full functionalities required during actual brachytherapy in the OR. Recently, our group described a robotic system for brachytherapy that is better equipped to perform most of the required procedures automatically (or semi-automatically) [8]. The current paper extends the previously reported research and development efforts: It presents the detailed design and development of a prototype robotic system called Euclidean, the sequence of operations when using Euclidean in clinical procedures, and the results of preliminary experiments to evaluate Euclidean's accuracy and repeatability in prostate seed implantation.

Figure 2. Workspace for robotic insertion of brachytherapy needle: (i) front view, (ii) top view. [Color version available online.]

System design and development

At this time, there is no complete mechanized/motorized system available for performing prostate brachytherapy semi-automatically (i.e., with a motorized system working under the supervision of the surgeon). Euclidean was designed and fabricated as a more complete robotic system for assisting clinicians during prostate brachytherapy procedures. Measurements of the available workspace for a robot were made in the OR during actual brachytherapy procedures (Figure 2). The knowledge gleaned from these measurements was useful in designing a compact robotic system that would work efficiently in the severely constrained workspace in the OR.

Objectives

The main objects of this robotic system are to

• improve accuracy of needle placement and seed delivery;

• improve avoidance of critical structures (urethra, pubic arch, bladder, etc.);

• update dosimetry after each needle is implanted (automatic seed localization);

• detect tissue heterogeneities and deformation via force sensing and imaging feedback;

• reduce trauma and edema;

• reduce radiation exposure;

• reduce the learning curve; and

• reduce OR time.

Functional requirements

The functional requirements of the system are as follows:

• Provision for quick and easy disengagement in case of emergency.

• Provision for reverting to conventional manual brachytherapy method at any time.

• Improved prostate immobilization techniques.

• Provision for periodic quality assurance checking.

• Provision for updating the implantation plan after implanting the periphery of the prostate, or after most of the needles have been placed and the seeds implanted, or at any other time.

• A method to enable the clinician to review and approve the motion plan before needle placement commences.

• Ability to modulate velocity and needle rotation by automatic feedback control.

• Ability to provide visual confirmation by the chosen imaging technique of each seed deposition or of the needle tip at the resting position.

• Ability to steer the needle by automatic feedback control.

• Visual/haptic force feedback during needle insertion.

• A teach mode to simulate the force/velocity patterns of expert practitioners.

• Ease of operation and safety for the patient and OR staff.

• Ease of cleaning and decontamination.

• Compatible with sterilization procedures.

Prototype robotic system: Positioning module

The prototype robotic system comprises a 9-DOF positioning module and a 7-DOF surgery module. The positioning module in turn comprises a 3-DOF cart and a 6-DOF platform.

1. 3-DOF cart. This can move in the x-y horizontal plane and rotate about a vertical axis. It consists of a base with four wheels capable of rotating about two of its own axes (a horizontal axis and a vertical axis) and a floor locking mechanism; during locking, four legs drop to the floor and lift the whole system off the wheels. This ensures rigid locking on any floor and provides greater rigidity to the system. Above the base, all the electronic and electrical components, including an industrial computer, are housed in an enclosure. The cart is brought to a relatively close operating position near the patient and locked on the floor by pressing a lever, after which the positioning platform is suitably adjusted.

2. 6-DOF platform. This connects the surgery module to the cart. The platform has 3 DOF in translational motions and 3 DOF in rotational motions. Thus, it can position and orient the surgery module at any location in the 3D space so that the US probe can be positioned and oriented in the patient's rectum easily and comfortably; at the same time the needling mechanism should be suitably aligned with the patient's perineum. The vertical lift (y-motion) of the platform is motorized for ease of operation and to avoid gravitational effects on the braking system. The translational motions in the horizontal plane (x- and z-motions) are manual, but these joints are unlocked using two solenoids; for safety and stability they are locked by default. The 3-DOF rotational motions (roll-pitch-yaw) are achieved by using a spherical joint which can be mechanically locked at a desired orientation.

Prototype robotic system: Surgery module

This module comprises a 2-DOF ultrasound probe driver, a 3-DOF gantry, a 2-DOF needle driver, a seed pusher, and a teach-pendant

1. 2-DOF ultrasound probe driver. The US probe (typically having two transducers for the axial and longitudinal planes) can be translated and rotated separately by two DC servo motors fitted with encoders and gearboxes. This enables imaging in the transverse as well as fan-beam (sagittal) directions, providing the capability to improve the 3D prostate model for dosimetric planning. The working range of motion for the US probe is 0–185 mm and −91° to +91° in translation and rotation, respectively. The clinician can also drive the US probe manually using the knobs; while in this mode the motors will be automatically disengaged by the electrical clutches (Figure 3). Provision for an optional template holder at the end of the US probe driver will permit manual takeover, if required. The prostate stabilization needle guide can orient the needle at any desired angle in both the horizontal and vertical planes, resulting in improved stabilization of the prostate and thereby enhanced accuracy of seed placement.

2. 3-DOF Gantry. This connects the needle-driving module to the positioning platform. The gantry has two translational motions (in the x- and y-directions) and one rotational motion (pitch). The motions are achieved by DC servo motors and optical encoders fitted to the motors (Figure 4). The working range of motion is 0–62 mm in the x-direction and 0–67 mm in the y-direction, which is sufficient to cover the projected treatment area of a conventional template grid (60 mm × 60 mm). The rotational range for angling the needle to avoid the pubic arch is −5° to +5°. The 3-DOF motions of the gantry position can orient the needle at any desired location on the patient's perineum with greater freedom due to the absence of a physical template as used in conventional brachytherapy. Once the needle is positioned at the desired location close to the perineum, the needle driver inserts the needle into the patient. The motions of the US probe driver and the rest of the surgery module (the gantry and needle driver) are decoupled by making two separate open kinematic chains attached to the same positioning platform.

3. 2-DOF needle driver. The needle components, comprising a hollow cannula and a solid stylet, are driven separately by two DC servo motors. Since the stylet and cannula motions are concentric, they essentially provide a single DOF. However, the cannula can be rotated continuously or partially using another tiny DC motor. Thus, with 3 motorized motions the needle driver actually has 2 DOF. Both the stylet and cannula are driven from the back (i.e., they are pushed), so there is no chance of slipping. During actual brachytherapy procedures, the needle traverses different types of tissues to reach the target point in the prostate, and these tissues have different types of boundary conditions that cannot be assessed from experimental data obtained with ex-vivo tissue/organ samples. Therefore, in-vivo measurement of needle insertion force (Figure 5) [9] is very useful in designing and controlling any robotic system that will work in such a constrained space. In our own design process, the collection and detailed analysis of this in-vivo information were found to be very useful. It was demonstrated that continuous rotation improved targeting accuracy and reduced insertion force [10], whereas partial rotation increased needle placement accuracy [5]. We therefore incorporated the provision for needle rotation in the Euclidean system to improve accuracy and seed delivery. To measure and monitor force profiles during the operational procedures, the design incorporated two single-axis force sensors (Model 13, Honeywell Sensotech, Columbus, OH), one at each of the proximal ends of the stylet and cannula, and one 6-axis force-torque sensor (Nano17, ATI Industrial Automation, Apex, NC) at the distal end of the cannula (Figure 6). Monitoring of these forces is useful in detecting pubic arch interference (PAI) and will help in assessing needle bending. The travel range of both the cannula and the stylet is 0–312 mm. The cannula does not need to move so far in normal operation except in the initial installation; however, the stylet has to move much further than the cannula in order to remove seeds from the cartridge and deposit them. Unlike commercially available seed cartridges (which hold 15 seeds), this cartridge can accommodate 35 seeds at once, resulting in less frequent replacement of the cartridge and thereby reducing potential radiation exposure as well as OR time.

4. Seed pusher. A flat-ended stylet (for a beveled-tip brachytherapy needle) is used to push the seed out of the cartridge and deposit it at the planned location in the prostate; the stylet's motor is deployed to push the stylet from the proximal end (Figure 6). The entire sequence of motions during seed delivery is fully automatic; however, the clinician can interrupt the motion at any time. The force sensor at the proximal end of the stylet monitors the force profile on the stylet and thereby confirms the removal of the seed from the cartridge to the cannula; experiments showed that an average of approximately 2.5 N of force is required to push the seed out of the cartridge. This is also checked by monitoring the stylet's motor current.

5. Teach-pendant. The teach-pendant with 10 buttons enables the surgeon to take over control of the surgery module at any time. From the teach-pendant, the surgeon can control needle insertion and rotation, seed deposition, the x-y movement of the gantry and the system abort.

At the distal end of the US probe driver is a provision for holding a conventional template, which will be useful following manual takeover (if requested by the clinician). Thus, the patient treatment will not be affected if the clinicians wish to switch to manual/conventional mode. In manual mode the needle-driving mechanism will be taken off the gantry to make room for the clinician, and the US probe will be operated by handwheels (knobs 1 and 2, see Figures 3 and 7). An emergency button on top of the cart can be used to stop the whole system in case of emergency (Figure 8).

Figure 3. Ultrasound probe driver.

Figure 4. Gantry robot.

Figure 5. In-vivo needle insertion force measured during actual (conventional) prostate brachytherapy procedure in the operating room.

Figure 6. Needle driver.

Figure 7. Assembled surgery module: (top) drawing and (bottom) fabricated prototype.

Figure 8. Prototype robotic system.

The surgery module is quite compact to facilitate working in the constrained workspace available during prostate brachytherapy procedures (Figures 7 and 8). The system's rigidity (stress-strain), deformation and safety characteristics have been analyzed using COSMOS, a commercial finite element analysis software. The maximum dimensions (length × width × height) of the surgery module are 510 mm × 290 mm × 185 mm for the US probe driver; 500 mm × 140 mm × 70 mm for the needle driver; and 510 mm × 290 mm × 235 mm overall. The width at the patient (distal) end of the US probe driver is 100 mm, and the equivalent measurement for the needle driver is 60 mm. The gross weight of the surgery module is approximately 9 kg, which can be reduced significantly by using plastic/nylon instead of the current surgical-grade stainless steel and aluminum.

System software and control

The system architecture is shown in Figure 9. The surgery module is fully motorized. An industrial computer with a Pentium4 processor, 2.8 GHz, 1 GB RAM, and 8 PCI slots (Chassis Plans, San Diego, CA) is used to control the system. We used two Galil control cards (Model DMC-1842; Galil Motion Control, Inc., Rocklin, CA): one card to control the US probe driver and gantry motions and the other to control the needle driver and seed pusher. A robust PID (proportional, integral and derivative) controller has been developed for the system control. We have tuned the PID gains in such a manner that the system's stability is maintained when the needle changes its state from merely position control while in the air to both position and force control mode upon entering the patient's body. The needle can achieve a velocity of up to approximately 100 mm/s, which is sufficient for brachytherapy procedures.

Figure 9. Schematic diagram of system architecture.

We used one frame-grabber (FalshBus, Integrated Technologies, Indianapolis, IN) for US image capturing, and three force-torque sensors (ATI Nano 17 and Honeywell M13) for needle insertion force monitoring and robot control feedback; all the motors are fitted with optical encoders (MicroMo Electronics, Inc., Faulhaber Group, Clearwater, FL) which provide final motion resolutions (considering gear ratios and screw leads) of 0.0007 mm for gantry x-y translations, 0.004 mm for stylet and cannula motions, and 0.005 mm and 0.06° for US probe translation and rotation, respectively.

We have developed software in C++ that is capable of dosimetric planning, 3D visualization, needle tracking and seed detection for dynamic planning. The optimized dosimetric plan is achieved using PIPER (developed at the University of Rochester), which uses a genetic algorithm. Intraoperative isodose delineations are displayed on transverse, sagittal and coronal views of the prostate model built from US images.

Clinical procedures

In the setup state, the Euclidean is initialized and patient information is entered into the computer by the user. The TRUS is then moved to scan the prostate in the transversal plane and the images are saved at desired intervals. In the next step, the modeling state, the TRUS images are used in delineating the prostate boundary, urethra, pubic bone, rectum, and seminal vesicle. A 3D model of the prostate is then generated automatically.

This 3D model is used for dosimetric planning to obtain the desired coordinates of the radioactive seed distribution. The designed software can display the planned iso-dose contours, needle position and seed locations in three dimensions. This provides the clinicians with a useful visualization of the whole treatment plan which can be edited if required.

Once the plan is approved by the clinicians, a single needle is inserted into the patient according to the plan. At this stage the TRUS is employed in the sagittal plane to track the needle location. To ensure the patient's safety, this needle insertion is performed in a sequential order: First, the gantry moves in the x-y direction to bring the tip of the needle close to the perineum. The gantry is then stopped and the needling mechanism pushes the needle (stylet and cannula together) into the patient to a predefined depth. Next, the clinician uses the user's pendant to insert the needle to the final depth. By taking over control, the operator ensures the patient's safety and is able to accommodate any change in planned depth that may be required due to tissue/organ deformation or needle deflection. After the needle has been inserted, the system goes into the implanting state: The seed is loaded from the cartridge and implanted according to the plan, whereupon the needle is withdrawn.

While seeds are being delivered, the user may request that the system go into the validation state to validate the latest dosimetry, or skip the validation and return to the needling state. In the validation state, the prostate is scanned and the seeds located in order to update the dosimetry for validation purposes. If the clinicians find that the dosimetry is incorrect, the system can return to the planning state to make adjustments for the remaining seeds that are yet to be inserted.

Results and discussion

Experiments were performed to evaluate the system's accuracy and repeatability. The system was run in a pattern (as shown in Figure 10) for 2 hours at a time. An 18-gauge beveled-tip brachytherapy needle was inserted into a graph paper pasted on the front of a foam block, which provided adequate and uniform support to the paper. There were a total of 16 penetration locations in a 60 mm × 60 mm area. The gantry moved the needle driver in a sequence (1-2-3- ··· -16), as shown in Figure 10a, so that the mechanical systems underwent back-and-forth motions in order to experience backlash/slack errors in the motion-transfer trains/linkages. After a single penetration at each location, the needle was moved to the next location, and this process was repeated 100 times. The x and y speed of the gantry was 10 mm/s, and the needle speed was 50 mm/s. To assess the error and deviation, we also performed the same procedure for a single insertion at each location (Figure 10b). It was very difficult to distinguish the deviation of the penetration holes after 100 runs. Optical images of the holes were magnified by a factor of 5 to enable measurement of the errors. The average error in x and y motions that the system (needle tip) encountered after 100 insertions was ± 0.2 mm with a standard deviation of 0.045 mm.

Figure 10. Position and size of holes (a) after one hundred penetrations at each location by moving the needle from one location to the next after each penetration; and (b) after a single penetration at each location. [Color version available online.]

We have also performed some preliminary seed implant experiments. We used a tissue-equivalent soft material phantom prepared from polyvinylchloride (PVC) by mixing supersoft plastic and softener in the ratio 70% to 30%, respectively (MF Manufacturing, TX). As seen in Figure 11, the brachytherapy seeds are deposited in the PVC phantom 1 cm apart (in the x, y and z directions). Assessment of the deposited seeds revealed that the accuracy (rms error) of seed placement is 0.67 mm (SD = 0.58 mm) in the x direction, 0.13 mm (SD = 0.12 mm) in the y direction, and 0.11 mm (SD = 0.11 mm) in the z direction. The 3D (Euclidean) rms error is 0.69 mm, which is quite small in comparison to the clinically acceptable (or desirable) value of 2.0 mm. It has been observed that the seed placement inaccuracy in the x direction is much larger than that in the y or z directions; we are investigating this issue with rigorous experimentation.

Figure 11. Seed delivery in soft material (PVC) phantom.

Conclusions

This paper has presented the design and development of a compact robotic system for prostate brachytherapy treatment with radioactive permanent seed deposition. Numerous techniques and sensors were incorporated into the prototype to improve the needle insertion accuracy, and thereby reduce seed delivery errors. The system can be operated in both automated and semi-automated modes. In case of emergency, the system can be stopped immediately and the clinicians can switch to the conventional manual mode. To achieve patient safety, a variety of hardware (sensors and stops) and software checks are incorporated into the system design and development. Needle and seed placement accuracy and repeatability of the prototype system were tested and found to be sufficient for clinical requirements. Additional more rigorous experiments are being conducted in phantoms to evaluate the overall performance of the system. Following full characterization of the prototype robotic system in phantoms, a clinical study will be carried out; the clinical protocol is currently undergoing IRB review. In addition to US guidance, we are also considering the use of computed tomography (CT) as an alternative imaging modality.