Robotic ultrasound-guided prostate intervention device: system description and results from phantom studies

Transcription

1 THE INTERNATIONAL JOURNAL OF MEDICAL ROBOTICS AND COMPUTER ASSISTED SURGERY Int J Med Robotics Comput Assist Surg 2009; 5: Published online 14 January 2009 in Wiley InterScience ( ORIGINAL ARTICLE Robotic ultrasound-guided prostate intervention device: system description and results from phantom studies H.S.S. Ho 1 * P. Mohan 1 E.D. Lim 1 D.L. Li 2 J.S.P. Yuen 1 W.S. Ng 2 W.K.O. Lau 1 C.W.S. Cheng 1 1 Department of Urology, Singapore General Hospital, Singapore 2 Department of Computer Integrated Medical Intervention Laboratory, Nanyang Technological University, Singapore *Correspondence to: H.S.S. Ho, Department of Urology, Singapore General Hospital, Outram Road, Singapore ho.henry.s.s@gmail.com Abstract Background We introduce the first robotic ultrasound-guided prostate intervention device and evaluate its safety, accuracy and repeatability. Methods The robotic positioning system (RPS) determines a target s x, y and z axes. It is situated with a biplane ultrasound probe on a mobile horizontal platform. The integrated software acquires ultrasound images for three-dimensional (3D) modelling, coordinates target planning and directs the RPS. Results The egg phantom evaluates the software s safety and workflow protocol. Two random targets are planned in each quadrant and biopsy needles are inserted. All were within three separate eggs. Metal wire tips are targeted and their distances from the biopsy needle tips are measured. With 20 wires, <1 mm accuracy is obtained. Repeatability is demonstrated when previous positions are returned to with similar accuracy. Conclusion Our device demonstrates safety in a defined boundary with a repeatable accuracy of <1 mm. It can be used for accurate prostate biopsy and treatment delivery. Copyright 2009 John Wiley & Sons, Ltd. Keywords prostate; robotic; ultrasound; experiments; biopsy; accuracy Introduction Accepted: 12 November 2008 Prostate cancer (PCa) has been the most common cancer in American and European men for the last 5 years. In 2007, the American Cancer Society estimated new cases of PCa and cancer-specific deaths in the USA (1). In Asia, its incidence is increasing (2). The standard means of diagnosis are abnormal digital rectal examination of the prostate gland or high serum level of prostate-specific antigen. However, prostate biopsy remains the only means of histological confirmation. Prostate biopsy (PB) can be obtained via the transrectal (TR) or transperineal (TP) approach. Since the introduction of transrectal ultrasoundguided (TRUS) sextant prostate biopsy by Hodge et al. (3), its short learning curve has ensured that TRPB is the tool preferred by the urologist (4). However, it has false-negative rate of 31% and potential morbidities of life-threatening sepsis and TR bleeding (5). In addition to negating such morbidities, the TP approach can perform thorough biopsy at the apical and anterior parts of the prostate, where PCa is commonly found (6). Copyright 2009 John Wiley & Sons, Ltd.

2 52 H. S. S. Ho et al. In localized PCa, surgery (radical prostatectomy) and radiotherapy are acceptable treatment options (7). The latter includes external beam radiation and brachytherapy, in which radioactive seeds are placed into all parts of the prostate gland under TRUS guidance. The current technique of seed placements is via a template of holes for needle insertion. For sterility, the seeds are delivered via multiple perineal skin punctures, instead of the transrectal route. A prostate intervention system that performs both roles while eliminating the current shortcomings is highly desirable. In addition, an accurate system could also perform targeted prostate biopsy and precise treatment delivery. With improvements in prostate cancer imaging, the ability to identify cancer tissue and its location is rapidly coming within our grasp (8). It will be ideal for obtaining prostate tissue for histological diagnosis with the minimal number of biopsies to reduce its associated morbidity. Potentially, it can also deliver treatment to the appropriate region. In this article, we introduce our device and describe the principles behind its design. We also aim to demonstrate its safety capability and document its accuracy and repeatability. Technical principles There are two fundamental technical principles upon which our device is built the transperineal approach and the dual-cone concept. Transperineal approach TP prostate intervention maintains sterillity, which is an important surgical principle. In addition, TPPB has efficient access to the apical and anterior part of the peripheral zones of the prostate, where cancers are commonly found (9). A contemporary comparison between the two approaches shows that TP detects 95% of the PCa vs. 79% in the TR (10). However, brachytherapy is strictly TP, as sterile radioactive seeds are inserted into the prostate, while TR insertion would contaminate them. Thus, a device that adopts the TP approach will permit biopsy and treatment delivery. Dual-cone concept Our novel dual-cone concept ensures that the any part of the prostate can be accessed with minimal skin puncture. In this concept, one cone refers to an imaginary threedimensional (3D) range of possible interventions in the prostate gland and the other cone is its mirror image at theexterior.thesitewherethesetwoconesmeetisat the skin and is known as the pivot point (Figure 1). Although a single dual-cone design may also cover the entire prostate with one pivot point, it has the following disadvantages. The coverage angle to the outermost parts of the gland is wider (Figure 1A). This will lead to its physical obstruction with the patients thigh (left and right sides) and the ultrasound probe platform (lower side). In addition, there is imminent risk of urethral injury from multiple passages of the biopsy needle. If the pivot point is in the midline, it will inevitably pass through the urethra, which is also in the midline. If it is placed on either side Figure 1. (A) Single dual-cone coverage. (B) Double dual-cone coverage

3 Robotic ultrasound-guided prostate intervention device 53 of the midline, the needle will traverse the urethra on its way to the opposite side. Our double dual-cone concept is a non-crossing design (Figure 1B). The outer parts of the prostate can be accessed by narrower angles with no external physical obstruction. The left pivot point is meant for left-sided prostate biopsy and vice versa. In this manner, there will be no needle passage that will cross the mid-portion of the prostate, negating the risk of urethral injury. Materials and methods Hardware The device consists of a mobile cart that houses a computer, a power lift that supports the robot and a touch screen for the user interface (Figure 2). The robot is situated upon a horizontal platform at the top of a balljoint that gives it six rotational degrees of freedom. This is pivotal for intimate contact between the ultrasound probe and the prostate gland for optimal ultrasound images. The robot consists of three coordinated components; a gantry, a gun-holder and the ultrasound probe holder. The gantry holds part of the automated robotic positioning system (RPS) and the two pivot points (Figure 3). The gantry RPS determines the x and y axes (left/right and up/down) of a point in 3D space. The z axis (depth of insertion) is determined by the gun-holder, which holds the biopsy gun and its needle. At each position, a stopper automatically shifts along the needle length, limiting the depth of its insertion (Figure 4). Thus, the complete RPS consists of the gantry RPS and the gunholder. The ultrasound probe holder consists of the ultrasound transducer holder and the rectal sheath holder (Figure 3). The ultrasound probe is placed in the rectal sheath, before they are inserted into the rectum. It is made of plastic material that does not interfere with the ultrasound waves. It maintains the anteroposterior position of the prostate gland during ultrasound probe Figure 3. Gantry and ultrasound probe holder Figure 4. Gun-holder and biopsy gun Prostate Probe sheath Probe Gel Air goes in and out to the rectum Figure 5. Schematic diagram of ultrasound rectal sheath Figure 2. BioXbot G1 movements in and out of the rectum (Figure 5). This ensures that the prostate position remains constant. Digital image acquisition is via a video cable from the output of the ultrasound machine to the device. The ultrasound probe is driven by the motion controllers,

4 54 H. S. S. Ho et al. which stops at every 0.5 mm distance within the userdefined range. At each stop, a two-dimensional (2D) transverse plane ultrasound image is captured by the frame-grabber software and digitally archived. We use the Aloka ultrasound machine (SSD-1700, Aloka, Tokyo, Japan) and multi-frequency bi-planar convex and linear transducer probe (Aloka, UST-672-5/7.5). The biopsy device is a Magnum biopsy gun, Magnum needle MN1620 (Cook) and needle guide C1610B. Software Our device integrated an in-house software development that marries the engineering essentials and clinical requirements. Its development is in collaboration with urologists, with clinical application of prostate biopsy in mind. Here, we shall describe the operationally important and clinically relevant components of the software. The third-party software programmes are detailed in the Supporting information. The following clinical safety and requirements are incorporated: 1. Targets can only be placed within the defined prostate boundaries, particularly at its base. This is to prevent extra-prostatic interventions, leading to bladder injuries. 2. Biopsy always starts at the apex of the prostate. The specimen trough of the biopsy needle is taken into consideration for this part of the software design. This is important as the apex is not thoroughly biopsied in TRPB, despite its high incidence of PCa. 3. In large prostate glands, there will be overlapping biopsies for each trajection. While this ensures that the entire length of the prostate is biopsied, it also demonstrates that any part of the prostate can be accessed. This is an important requirement for eventual treatment delivery. Software workflow protocol Image acquisition (Figure 6A). The automated scanning range is set by the user. It is defined by the length of the prostate, from base to apex or vice versa. When the scanning begins, ultrasound images are digitally acquired. Figure 6. Software screen shots: (A) 2D images with boundaries drawn; (B) 3D model creation; (C) target planning; (D) 3D model with trajectories

5 Robotic ultrasound-guided prostate intervention device 55 3D surface modelling (Figure 6B). The series of 2D ultrasound images of the prostate obtained by the frame-grabber is presented. Next, the user selects 5 6 images from the base to the apex of the prostate and demarcates its boundary in each of them. This information is rendered into a interactive 3D model for verification. Target planning (Figure 6C). Atthisinterface,theuser plans the target points. Although it is constructed on a single transverse ultrasound image, any point within the 3D prostate model boundary can be denoted as a target point (biopsy). The device will not permit a target point to be planned beyond the 3D model boundary. Our biopsy scheme would be at least 20 cores biopsy, with emphasis on the peripheral zone and the anterior part of the prostate. Increased number of biopsy cores would be taken with a larger prostate gland. Pivot point preparation. The software computes the target trajectory and which pivot point should be used from its location. This feature ensures safety and accuracy, such that unless the pivot point is on the correct side, the RPS will not move to the first biopsy position. Thus, to biopsy the left side of the prostate, the pivot point must be on the left side and vice versa. Biopsy can begin from either side of the prostate. Execution (Figure 6d) The interactive 3D prostate model with the proposed trajectories is presented. The user will verify that each trajectory with proposed biopsy(s) passes through the desired location. When the user approves the plan, the RPS moves to its first planned position. At each planned position, when the user has verified the completion of intervention, RPS will move automatically to the next position until the plan is fulfilled. In an unfavourable situation, where the prostate gland is very mobile, the target can be shifted accordingly to accommodate these changes. Calibration Device calibration is an important pre-use procedure that brings the robotic and imaging systems into a single common reference frame. Every ultrasound image pixel has an unique 3D position with respect to the physical coordinate. Calibration ensures that this relationship remains constant. In other words, an ultrasound image change in any axis will result in a similar change in the RPS. As such, the 2D image data can be used for modelling and planning while executing the plan in the physical coordinate precisely. The ultrasound calibration is mathematically formulated as follows: qi = R (S pi) + t + t. t where pi is the 2D coordinates of a pixel in ith the image and qi is the corresponding point in the 3D robot space. The aim of the calibration procedure is to compute the Figure 7. Calibration concept and ultrasound image rotation matrix R, the translation vector t and the scaling factor S. A calibration box with three edges at predetermined distances has been specially designed. These edges are identified on the ultrasound image and the transformation is calculated (Figure 7). Design of the phantom experiments Our objectives were to determine the device s safety, accuracy and its repeatability. The latter refers to its ability to return to the same pre-determined point after it has been directed away to another point. This is clinically important if targeted treatment delivery is a goal for this device. We tested the above features with different types of phantom tests. Egg phantom test In this test, we aimed to determine whether the device is able to direct needles to targets within a defined boundary. This was to test the software s safety in terms of image acquisition, 3D modelling and pre-defined clinical boundaries. For this purpose, we suspended a shelled hard-boiled chicken egg into gelatin. The container was a clear plastic box, measuring cm. The gelatin was prepared by mixing 100 g gelatin powder with 400 ml warm water (80 C). (Figure 8). The egg was suspended in the gelatin, which was cooled to 4 C in the refrigerator for 1 h. As the gelatin was transparent, it allowed visual inspection and the passage of ultrasound waves, which was the cornerstone of our device. Moreover, it was cheap and easily prepared. We could not use free space (air), due

6 56 H. S. S. Ho et al. Figure 8. (A) Egg phantom test and (B) ultrasound screen shot Figure 9. Metal wire phantom test and ultrasound screen shot to its poor conductance to ultrasound waves. Moreover, we needed the egg to be well-supported to provide counter-resistance for needle penetration. After the egg phantom box was secured to our device the biplanar linear probe was placed in contact with its underside. The standard protocol began with the acquisition of the ultrasound images of the suspended egg. After we had delineated its margins, the generated 3D model was verified (Figure 8). In target planning, we identified eight random target points, with two in each quadrant within the egg boundary. The software did not allow us to place any target points outside the verified 3D model. This was consistent with the clinical safety requirements defined by the urologists. The endpoint was whether any of the directed needles was outside the egg. Through the transparent gelatin, we were able to observe the needles passage en route to each target point. All the needle tips were within the egg and none went beyond its boundary. At the end of each experiment, we sliced the egg phantom box to verify the number of trajections planned. We repeated the egg phantom tests three times with no change in our findings. While this test documented the feasibility and safety of our software protocol, its accuracy needed to be established. Accuracy and repeatability phantom experiments results In the next phase of the phantom tests, we determined the device s accuracy and repeatability. We prepared the gelatin phantom box as described previously without the egg. We insert five metal wires of 1 mm diameter from the side of the box opposite to the ultrasound probe. They were randomly placed within 4 7 cm radius of the ultrasound probe. We chose this area because this was where the prostate gland was normally located and the ultrasound wave was still useful. After the ultrasound images were collected, the tip of each metal wire was identified, starting from the leftmost (Figure 9A). On the planning screen, a marker was placed directing the RPS to that position (Figure 9B). Next, the biopsy needle was inserted through the needle sheath to the metal wire. The procedure was repeated for the remaining metal wires. Using the electronic vernier callipers, we measured the distance (mm) between the tip of the biopsy needle to the metal wire. To document repeatability, we returned to each metal wire location in a random fashion for similar measurements. We were able to return to each target location after the RPS had been directed to another position within 1 mm consistency. Discussion At present, prostate interventions are not ideal. TRPB has significant morbidities and a false-negative rate of 30%. As the TP approach is limited by its steeper learning curve, TRPB is the preferred diagnostic tool for PCa. Moreover, current brachytherapy delivery is template-based, which leads to multiple perineal skin punctures. Thus, a device that addresses these issues and perform both roles safely and efficiently is desirable. Our device is the first ultrasound-guided robotic system for transperineal prostate intervention. A device that attempts mechanical systematic prostate biopsy is the TargetScan (Envisioneering Medical Technologies, St. Louis, MO, USA) (11). Our device is only similar in the ultrasound image acquisition and reconstruction to a 3D model. Their positioning system is manual and limited to specific planes of the prostate. Moreover, their transrectal approach will probably restrict its applications to biopsy. In contrast, our intervention is not limited to specific plares of the prostate and our transperineal approach ensures dual roles biopsy and treatment delivery. One key principle upon which our device is built is an economical development with potential for widespread application. It utilizes equipment commonly found in

7 Robotic ultrasound-guided prostate intervention device 57 an urology office, such as ultrasound and standard biopsy needles. There are other groups that are working on a magnetic resonance imaging (MRI)-compatible prostate intervention system (12,13), made of non-ferrous material, such as ceramics and plastics, using titanium biopsy needles. The high cost of both MRI and the device may restrict its availability. Another challenge is the constrained space within the MRI tunnel. This will limit the space required for prostate intervention, particularly in the transperineal approach. Our dual-cone concept has been in clinical use; Emiliozzi et al. were the first to utilize this approach. In their description, their perineal needle entry points were placed on each side about 1.5 cm above the anus at 45 to the midline. Subsequent biopsy needle passage into theprostateglandwasguidedbya7.5mhzlinearprobe in the longitudinal plane (14). While it was planned to fan out into all parts of the peripheral zones of the prostate, the operator s experience would contribute significantly in accessing the desired areas. When they compared this approach with the TRPB, the TP approach detects 95% of the cancers vs. 79% in the TR approach (10). We use dual-cone and pivot points as engineering descriptions for these clinical techniques. Our device had materialized them into a consistent and accurate system. The clinical advantage of the dual-cone, approach ensures that the every part of the prostate gland can be reached by the two pivot points. With this design, only two perineal skin punctures are needed. Current perineal prostate intervention (biopsy/brachytherapy) uses the template to localize the exact position for intervention. This leads to multiple skin punctures. Moreover, the anterior parts of a large prostate gland may be obscured by the overlying pubic bone in this transperineal approach. The platform of our device permits a certain degree of angulation, which enables these hidden parts of a large prostate gland to be reached. With image guidance on the prostate cancer location, we believe that we will be able to target any part of the prostate gland. In the phantom studies, our device documents a repeatable accuracy of <1 mm. In these accuracy tests, gelatin represents the most ideal condition for passage of ultrasound waves while capable of holding targets in the 3D space. This is crucial, as we are testing both mechanical and software precision. The latter s safety is further verified with the egg phantom test. Given a defined 3D model, no target planning is permitted beyond it and in the actual testing; none of the trajectories went beyond the physical boundary of the egg. There are other factors affecting accuracy not considered in our phantom tests. Needle deflection secondary to tissue elasticity or stiffness is not simulated. Gelatin does not simulate human perineum or prostate tissue; it has minimal resistance to deflect needle passage. We also do not vary the type of needle tested. It is possible that the in stiffer tissue, the needle may deflect in the bevel direction (15). We need to consider this potential source of needle deflection if we are to have clinical inaccuracy. We may have to test accuracy with different bevel directions or symmetrical needles. Further tissue (non-homogeneous) phantom tests are needed to verify whether our observed accuracy can be maintained. They will involve animal organs such as liver or kidney, or even surgically excised prostate gland suspended in gelatin. Although they may provide variable tissue consistency to test needle deflection, they can never mimic the interaction of the prostate gland in the human perineum. We acknowledge that this is a flaw of our current homogeneous phantom experiments. In a clinical scenario, prostate gland movement may also contribute to inaccuracy. This is clearly demonstrated to be a significant problem in kidney puncture (16). As the kidneys are in direct contact with the diaphragm, breathing affects their position tremendously. However, the prostate is in the pelvis and the effect of shallow breathing on its position is not evident. Although we attempt to minimize the dorso-ventral movement with the ultrasound rectal sheath, cranio-caudal movement secondary to deep breathing or straining have to considered. Clinical trials with ethics committee approval are on the horizon to shed light on this issue. Our device has been developed with several clinical applications in mind. While systematic transperineal prostate biopsy will be the first clinical endeavour, treatment delivery for prostate cancer is next. It will be developed to deliver radioactive seeds for brachytherapy or cryotherapy. For the former purpose, we are working with the brachytherapy team to identify clinical considerations, such as placement of fiducial markers. We are constructing customized needles for delivery of the radioactive seeds with minimal prostate movement. With an accurate device, targeted intervention (biopsy/treatment) will also be possible. Imaging modalities that detect prostate cancer are improving rapidly (10). Software adaptations will allow integration of digital information from different modalities to direct our device to the cancer (17). With focal therapy of prostate cancer playing an increasingly active role, our device may further reinforce this treatment option. Conclusions Our device is the first ultrasound-based transperineal prostate intervention system. In these homogenous phantom tests, it has been demonstrated to be safe and accurate. It has potential application for TPPB, treatment delivery and, under imaging guidance, targeted prostate interventions. Acknowledgements We warmly thank our past colleagues, Wu Ruoyun and Liu Feng, for their contributions in this project. Thus study was supported by the National Medical Research Council, Singapore (Grant No.

8 58 H. S. S. Ho et al. NMRC/0859/2004) and the Enterprise Challenge Grant from the Prime Minister s Office, Singapore. Supporting information Supporting information may be found in the online version of this article. References 1. Jemal A, Siegel R, Ward E et al. Cancer statistics, Ca Cancer J Clin 2007; 57: Sim HG, Cheng CWS. Changing demography of prostate cancer in Asia. Eur J Cancer 2004; 41: Hodge KK, McNeal JE, Terris MK et al. Random systematic versus ultrasound guided transrectal core biopsies of the prostate. JUrol1989; 142: Shandera KC, Thibault GP, Deshon JE Jr. Variability in patients preparation for prostate biopsy amongst American urologists. Urology 1998; 52: Epstein JI, Walsh PC, Akingba G et al. The significance of prior benign needle biopsies in men subsequently diagnosed with prostate cancer. JUrol1999; 162: Satoh T, Matsumoto K, Fujita T et al. Cancer core distribution in patients diagnosed by extended transperineal prostate biopsy. Urology 2005; 66: Heidenreich A, Aus G, Bolla M et al.; European Association of Urology. EAU guideline on prostate cancer. Eur Urol 2008; 53: Oehr P, Bouchelouche K. Imaging of prostate cancer. Curr Opin Oncol 2007; 19: MohanP,HoH,YuenJet al. A 3D computer simulation to study the efficacy of transperineal versus transrectal biopsy of the prostate. Int J CARS 2007; 1: Emiliozzi P, Corsetti A, Tassi B et al. Best approach for prostate cancer detection: a prospective study on transperineal versus transrectal six-core prostate biopsy. Urology 2003; 61: Andriole GL, Bullock TL, Belani JS et al. Is there a better way to biopsy the prostate? Prospects for a novel transrectal systemic biopsy approach. Urology 2007; 70: Fichtinger G, DeWeese TL, Patriciu A et al. System for robotically assisted prostate biopsy and therapy with intraoperative CT guidance. Acad Radio 2002; 9: Muntener M, Patriciu A, Petrisor D et al. Magnetic resonance imaging compatible robotic system for fully automated brachytherapy seed placement. Urology 2006; 68: Emiliozzi P, Longhi S, Scarpone P et al. The value of a single biopsy with 12 transperineal cores for detecting prostate cancer in patients with elevated prostate specific antigen. JUrol2001; 166: Blumenfeld P, Hata N, Dimaio S et al. Transperineal prostate biopsy under magnetic resonance image guidance: a needle placement accuracy study. J Magn Reson Imaging 2007; 26: Mozer P, Leroy A, Payan Y et al. Computer-assisted access to the kidney. Int J Medical Robotics Comput Assist Surg 2005; 1: Daanen V, Gastaldo J, Giraud JY et al. MRI/TRUS data fusion for brachytherapy. Int J Med Robotics Comput Assist Surg 2006; 2: