Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

A ROS2-based Testbed Environment for Endovascular Robotic Systems

A ROS2-based Testbed Environment for Endovascular Robotic Systems DE GRUYTER Current Directions in Biomedical Engineering 2022;8(1): 8 9 - 92 Christoph Eyberg*, Lennart Karstensen, Tim Pusch, Johannes Horsch, and Jens Langejürgen A ROS2-based Testbed Environment for Endovascular Robotic Systems https://doi.org/10.1515/cdbme-2022-0023 instruments, i.e. guidewire and catheter, are navigated through the patient’s vascular system to the site of the lesion under Abstract: Developing autonomous endovascular robotic medical imaging, where the treatment is performed. systems requires physical testbeds to test control algorithms. Navigating through the vascular system is a complex task that Typically, such testbeds comprise of several hard- and exposes the surgeon to radiation and requires the usage of software components along with a way of having these contrast agent to guide the surgeon through the vascular components communicate with each other. Building such a system. testbed is a multidisciplinary task which can be beyond the Current research aims to automate the navigation task in scope of expertise for research groups. The goal of this work order to enhance patient safety and allow the physicians to is to facilitate setting up such testbeds in two ways: First, we focus on the actual treatment while reducing the required propose a testbed architecture that allows to develop tracking, dosage of contrast agent and radiation [2–4]. Yet, the control and instrument manipulation systems separately by development and testing of automated guidewire and catheter utilizing the ROS2 communication protocol. Secondly, we navigation in physical testbeds is challenging as it requires present a reliable yet straightforward to implement tracking solving multiple tasks at once: Receiving feedback about the algorithm for endovascular instruments that is built using only position of the instruments, e.g. by using the unfiltered open-source software packages. The tracking algorithm is medical image or tracking the instrument position, evaluated using both video camera and x-ray imaging and is determining an appropriate navigation manoeuvre to move the found to meet the requirements for real time control instruments closer to the target position and executing this algorithms. Furthermore, we show an example of the proposed manoeuvre. modular testbed architecture as it is used in our lab. Both the A frequently used approach to tracking guidewires is to modular testbed architecture and the open-source tracking represent the guidewire with a B-spline and then update the algorithm may serve as helpful building blocks for other position of its control points at every step by optimizing an researchers in the field seeking to evaluate their control energy function building both on image features and algorithms on physical testbeds. mechanical plausibility [5–7]. Vandini et al. [8] search for image features that possibly represent guidewire segments and Keywords: endovascular, surgical robotics, testbed, combine them to find the guidewire. In recent years tracking, guidewire, convolutional neural networks have been successfully used to extract the position of guidewires from fluoroscopy images [9–11]. For application in control algorithms a sufficiently 1 Introduction high tracking frequency and a low tracking induced delay are necessary. Clinicians typically use an image frequency of 4- Vascular diseases, especially ischemic heart (16%) and 10Hz while the stated research is able to process images within cerebrovascular disease (11.2%), are the leading causes of 50-175ms. These values can be used as a baseline requirement death worldwide [1]. The endovascular intervention is a for tracking algorithms in real time control loops for minimally invasive surgical method to diagnose and treat these autonomous guidewire navigation. Implementation and diseases. During this type of intervention thin and flexible runtime optimization of such professional solutions can be a challenging task, however, it might not be necessary during early stages of development of endovascular robotic systems. ______ In these stages facile solutions and interfaces that allow easy *Corresponding author: Christoph Eyberg: Fraunhofer IPA, replacement of each part of the testbed are required. Theodor-Kutzer-Ufer 1-3, Mannheim, Germany, e-mail: Christoph.Eyberg@ipa.fraunhofer.de The contribution of this paper is twofold: We present an Lennart Karstensen, Tim Pusch, Johannes Horsch, Jens architecture for a testbed using the ROS2 [12] interface which Langejürgen Fraunhofer IPA, Mannheim, Germany Open Access. © 2022 The Author(s), published by De Gruyter. This work is licensed under the Creative Commons Attribution 4.0 International License. 89 allows decentralized and wireless communication between the processing software, a guidewire manipulator, and a joystick. different components of the testbed. This allows to solve each The communication between the different elements is task independently and substitute between different solutions, performed through the ROS2 interface. This also allows us to e.g. switching to a state-of-the-art tracking solution when the place all control devices inside the secure control room stage of animal or clinical trials is reached. Secondly, a omitting exposure to any radiation. The setup is shown in tracking algorithm is suggested that is solely based on open Figure 1. The ROS2 architecture is built using the following source packages and yet meets the stated performance nodes: requirements and therefore allows in-vitro development and  The feedback node, orange, receives the live webcam or testing of navigation algorithms and instrument manipulators. fluoroscopy images and obtains tracking points of the guidewire. Those tracking points are then published to the feedback topic.  The target node, green, publishes the target position for 2 Method the guidewire tip to the target topic. The target position can be chosen in the displayed image.  The control node, yellow, subscribes to both the target 2.1 Testbed Architecture and the feedback topic and determines the control signal which is the desired translational and rotational velocity The testbed consists of a transparent phantom of a vascular for the guidewire. The control message is published to the system, which is either placed inside in an x-ray imaging action topic for each incoming feedback message. system (Artis Zeego, Siemens Helthineers) or mounted with a  The manipulation node, blue, subscribes to the action camera, a laptop with a 12-core, 2.6GHz processor and image topic and moves the guidewire accordingly. This architecture allows to easily substitute solutions for the different nodes e.g. replacing the manually controlled joystick in the control node with an autonomous control algorithm. Analogously, feedback and manipulation node can be replaced by a simulation. Additionally, multiple control nodes with different priorities can be utilized, e.g. to allow manual override via the joystick while navigating autonomously. 2.2 Tracking Algorithm Our novel guidewire tracking algorithm is placed inside the feedback node to retrieve guidewire tracking points from the image. The tracking algorithm uses the standard image processing functionalities of the publicly available Open CV [13] package. Instead of relying on complex algorithms our approach builds on manual parameter tuning utilizing two facts: First, the endovascular instruments are inserted at a static position which is known during the intervention. Second, endovascular instruments are slender devices, i.e. 𝑙 ≫ 𝑙 , 𝑧 𝑟 where 𝑙 is the straight length and 𝑙 is the radius of the 𝑧 𝑟 guidewire. Before the tracking loop is started, brightness and contrast of the incoming video stream can be manually adjusted. This allows to obscure the edges of the vascular tree and to enhance the visibility of the guidewire, when using a camera image. Figure 1: Phantom of vascular system with guidewire manipulator Likewise the thresholds for the edge detection can be hand under X-ray imaging (top), ROS topics (center) and guidewire tuned. Furthermore, the insertion area is manually marked (red tracking with joystick control pad in control room (bottom). A rectangle in Figure 2) and the coordinate system for the filled circle represents messages being published to the topic position of the guidewire is specified. In our setup the phantom while a ring represents a subscription 90 7. Retrieving and publishing tracking points that represent the centreline of the two edges of the guidewire. Figure 2 displays the successful retrieval of the guidewire from the fluoroscopy image. In evenly illuminated images it was also found possible to substitute steps 2 and 3 by a threshold filter. Setting all binary image values on the lower edge of the insertion area to zero after step 3 can additionally assert that a contour approximation point is placed inside the insertion area. Substituting step 6 by using the angle between the edges of the polygon approximation or adding a correction for guidewire bending during distance calculation were found to be less efficient. 2.3 Experiment Setup We validate our setup and evaluate the tracking algorithm by manually navigating through all branches of the vascular phantom using the joystick controller, as displayed in Figure 2, resulting in images with a variety of guidewire positions and lengths. During this task the delay of the image processing from retrieving a new image until returning the tracking points and the frequency of incoming tracking signals is measured. Also the number of time steps, where the algorithm is not able to retrieve a guidewire position from the image is counted. The experiment is conducted using both camera and x-ray imaging. Additionally, the experiment is repeated placing the vessel phantom on top of an anatomic phantom in x-ray imaging to Figure 2: Fluoroscopy image with additionally displayed vessel assess the capability of the tracking algorithm in a setup closer tree (top) and processed image with tracked guidewire to the clinical application. (bottom). of the vascular system is placed inside the x-z-plane with the guidewire pointing into the direction of the z-axis. This corresponds to the coordinate system of the x-ray imaging 3 Evaluation system. To track the guidewire, the following steps are performed: The tracking algorithm achieves an average processing delay 1. Transformation to grayscale image and adjustment of of less than 50ms under both camera and x-ray imaging which brightness and contrast as specified. is below the baseline of current research. The achieved control 2. Extracting the edges of the guidewire into a binary image frequency matches or even exceeds the stated requirements of using the Canny-Edge-Detector [14]. 4-10 Hz. All values are gathered in Table 1. 3. Applying a closing operation (consecutive usage of dilation and erosion filters) to fill up the edge-contours of Table 1: Average values for control frequency and delay, and the slender guidewire to achieve an enclosed contour. percentage of time-steps where the guidewire was not found 4. Retrieving the shapes and a polygon approximation of all enclosed contours in the binary image. Control Processing Guidewire not 5. Iterating through the approximation points of all contours Frequency Delay found until the guidewire is found as the contour with points inside the insertion area. Camera 17Hz 41ms 0.0% Tracking 6. Finding the tip of the guidewire by measuring the distance of both edges starting from the edge of the insertion area x-ray Tracking 10Hz 25ms 0.7% in positive coordinate direction. The tip is found as the point where both edges meet under equal distance. 91 While the guidewire was tracked in the camera image in References all time steps, there were a few time steps where the guidewire [1] World Health Organization (2020) Global Health Estimates could not be tracked or the guidewire tip was not found 2019 Summary Tables: Deaths by cause, age and sex, by correctly using the fluoroscopy image input. This was due to world bank income group, 2000-2019. rapid guidewire movements under relatively low image https://www.who.int/docs/default-source/gho- documents/global-health- frequency which resulted in a smeared guidewire. In the estimates/ghe2019_cod_global_2000_20194e572f53-509f- experiment an imaging frequency of 30p/s was used. The 4578-b01e-6370c65d9fc5.xlsx?sfvrsn=eaf8ca5_7. Accessed failure rate increases with a lower image frequency. 16 May 2022 The tracking algorithm failed to distinguish the guidewire [2] Karstensen L, Ritter J, Hatzl J et al. (2022) Learning-based autonomous vascular guidewire navigation without human from the background when the vessel phantom was placed on demonstration in the venous system of a porcine liver. top of the anatomic phantom, due to equal illumination of International Journal of Computer Assisted Radiology and guidewire and skeletal structures under x-ray imaging. The Surgery tracking algorithm also fails to extract the correct guidewire [3] Schegg P, Dequidt J, Coevoet E et al. (2022 - 2022) tip if the guidewire overlaps itself or is kinked by Automated Planning for Robotic Guidewire Navigation in the Coronary Arteries. In: 2022 IEEE 5th International maloperation. Conference on Soft Robotics (RoboSoft). IEEE, pp 239–246 [4] Kweon J, Kim K, Lee C et al. (2021) Deep Reinforcement Learning for Guidewire Navigation in Coronary Artery Phantom. IEEE Access 9:166409–166422. 4 Discussion https://doi.org/10.1109/ACCESS.2021.3135277 [5] Baert SAM, Viergever MA, Niessen WJ (2003) Guide-wire A testbed for endovascular robotic systems was presented. The tracking during endovascular interventions. IEEE Trans Med Imaging 22:965–972. utilization of the ROS2 communication protocol divides the https://doi.org/10.1109/TMI.2003.815904 navigation task into the subtasks of receiving a feedback about [6] Slabaugh G, Kong K, Unal G et al. (2007) Variational the instruments position, finding an appropriate manoeuvre Guidewire Tracking Using Phase Congruency. In: Springer, and executing this manoeuvre. This allows researchers to Berlin, Heidelberg, pp 612–619 [7] Chang P-L, Rolls A, Praetere H de et al. (2016) Robust focus on each individual task during development while Catheter and Guidewire Tracking Using B-Spline Tube Model having the ability to easily replace different solutions for all and Pixel-Wise Posteriors. IEEE Robot Autom Lett 1:303– other components. 308. https://doi.org/10.1109/LRA.2016.2517821 Furthermore, a tracking algorithm for guidewires was [8] Vandini A, Glocker B, Hamady M et al. (2017) Robust presented that can be implemented straightforwardly using guidewire tracking under large deformations combining segment-like features (SEGlets). Medical Image Analysis open source software packages. It meets the requirements for 38:150–164. https://doi.org/10.1016/j.media.2017.02.001 application in navigation control loops and reliably tracks the [9] Zhou Y-J, Xie X-L, Bian G-B et al. (2019) Fully Automatic guidewire in two-dimensional vascular phantoms. Dual-Guidewire Segmentation for Coronary Bifurcation The testbed allows researchers to evaluate control Lesion. In: 2019 International Joint Conference on Neural Networks (IJCNN). IEEE, Piscataway, NJ, pp 1–6 algorithms or robots for endovascular instruments by replacing [10] Gherardini M, Mazomenos E, Menciassi A et al. (2020) the respective ROS2 node with their solution. Algorithms that Catheter segmentation in X-ray fluoroscopy using synthetic successfully navigate through the presented physical testbed data and transfer learning with light U-nets. Computer are promising candidates for testing in phantoms with higher Methods and Programs in Biomedicine 192:105420. https://doi.org/10.1016/j.cmpb.2020.105420 complexity or even animals. The tracking algorithm can then [11] Wagner MG, Laeseke P, Speidel MA (2019) Deep learning easily be replaced by a professional solution. based guidewire segmentation in x-ray images. In: Gilat- Schmidt T, Chen G-H, Bosmans H (eds) Medical Imaging 2019: Physics of Medical Imaging: 17-20 February 2019, San Diego, California, United States. SPIE, Bellingham, Washington, USA, p 150 [12] Macenski S, Foote T, Gerkey B et al. (2022) Robot Operating Author Statement System 2: Design, architecture, and uses in the wild. Sci Research funding: This project is funded by the Ministry of Robot 7:eabm6074. Economics, Labour and Tourism Baden-Württemberg within https://doi.org/10.1126/scirobotics.abm6074 [13] Bradski G (2000) The OpenCV Library. In: Dr. Dobb's the framework of the Forum Gesundheitsstandort Baden- Journal of Software Tools Württemberg. [14] Canny J (1986) A Computational Approach to Edge Conflict of interest: Authors state no conflict of interest. Detection. IEEE Trans Pattern Anal Mach Intell PAMI-8:679– 698. https://doi.org/10.1109/TPAMI.1986.4767851 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Current Directions in Biomedical Engineering de Gruyter

A ROS2-based Testbed Environment for Endovascular Robotic Systems

Download PDF
4 pages

Loading next page...
 
/lp/de-gruyter/a-ros2-based-testbed-environment-for-endovascular-robotic-systems-vQFIbUOAcq

References

References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.

Publisher
de Gruyter
Copyright
© 2022 by Walter de Gruyter Berlin/Boston
eISSN
2364-5504
DOI
10.1515/cdbme-2022-0023
Publisher site
See Article on Publisher Site

Abstract

DE GRUYTER Current Directions in Biomedical Engineering 2022;8(1): 8 9 - 92 Christoph Eyberg*, Lennart Karstensen, Tim Pusch, Johannes Horsch, and Jens Langejürgen A ROS2-based Testbed Environment for Endovascular Robotic Systems https://doi.org/10.1515/cdbme-2022-0023 instruments, i.e. guidewire and catheter, are navigated through the patient’s vascular system to the site of the lesion under Abstract: Developing autonomous endovascular robotic medical imaging, where the treatment is performed. systems requires physical testbeds to test control algorithms. Navigating through the vascular system is a complex task that Typically, such testbeds comprise of several hard- and exposes the surgeon to radiation and requires the usage of software components along with a way of having these contrast agent to guide the surgeon through the vascular components communicate with each other. Building such a system. testbed is a multidisciplinary task which can be beyond the Current research aims to automate the navigation task in scope of expertise for research groups. The goal of this work order to enhance patient safety and allow the physicians to is to facilitate setting up such testbeds in two ways: First, we focus on the actual treatment while reducing the required propose a testbed architecture that allows to develop tracking, dosage of contrast agent and radiation [2–4]. Yet, the control and instrument manipulation systems separately by development and testing of automated guidewire and catheter utilizing the ROS2 communication protocol. Secondly, we navigation in physical testbeds is challenging as it requires present a reliable yet straightforward to implement tracking solving multiple tasks at once: Receiving feedback about the algorithm for endovascular instruments that is built using only position of the instruments, e.g. by using the unfiltered open-source software packages. The tracking algorithm is medical image or tracking the instrument position, evaluated using both video camera and x-ray imaging and is determining an appropriate navigation manoeuvre to move the found to meet the requirements for real time control instruments closer to the target position and executing this algorithms. Furthermore, we show an example of the proposed manoeuvre. modular testbed architecture as it is used in our lab. Both the A frequently used approach to tracking guidewires is to modular testbed architecture and the open-source tracking represent the guidewire with a B-spline and then update the algorithm may serve as helpful building blocks for other position of its control points at every step by optimizing an researchers in the field seeking to evaluate their control energy function building both on image features and algorithms on physical testbeds. mechanical plausibility [5–7]. Vandini et al. [8] search for image features that possibly represent guidewire segments and Keywords: endovascular, surgical robotics, testbed, combine them to find the guidewire. In recent years tracking, guidewire, convolutional neural networks have been successfully used to extract the position of guidewires from fluoroscopy images [9–11]. For application in control algorithms a sufficiently 1 Introduction high tracking frequency and a low tracking induced delay are necessary. Clinicians typically use an image frequency of 4- Vascular diseases, especially ischemic heart (16%) and 10Hz while the stated research is able to process images within cerebrovascular disease (11.2%), are the leading causes of 50-175ms. These values can be used as a baseline requirement death worldwide [1]. The endovascular intervention is a for tracking algorithms in real time control loops for minimally invasive surgical method to diagnose and treat these autonomous guidewire navigation. Implementation and diseases. During this type of intervention thin and flexible runtime optimization of such professional solutions can be a challenging task, however, it might not be necessary during early stages of development of endovascular robotic systems. ______ In these stages facile solutions and interfaces that allow easy *Corresponding author: Christoph Eyberg: Fraunhofer IPA, replacement of each part of the testbed are required. Theodor-Kutzer-Ufer 1-3, Mannheim, Germany, e-mail: Christoph.Eyberg@ipa.fraunhofer.de The contribution of this paper is twofold: We present an Lennart Karstensen, Tim Pusch, Johannes Horsch, Jens architecture for a testbed using the ROS2 [12] interface which Langejürgen Fraunhofer IPA, Mannheim, Germany Open Access. © 2022 The Author(s), published by De Gruyter. This work is licensed under the Creative Commons Attribution 4.0 International License. 89 allows decentralized and wireless communication between the processing software, a guidewire manipulator, and a joystick. different components of the testbed. This allows to solve each The communication between the different elements is task independently and substitute between different solutions, performed through the ROS2 interface. This also allows us to e.g. switching to a state-of-the-art tracking solution when the place all control devices inside the secure control room stage of animal or clinical trials is reached. Secondly, a omitting exposure to any radiation. The setup is shown in tracking algorithm is suggested that is solely based on open Figure 1. The ROS2 architecture is built using the following source packages and yet meets the stated performance nodes: requirements and therefore allows in-vitro development and  The feedback node, orange, receives the live webcam or testing of navigation algorithms and instrument manipulators. fluoroscopy images and obtains tracking points of the guidewire. Those tracking points are then published to the feedback topic.  The target node, green, publishes the target position for 2 Method the guidewire tip to the target topic. The target position can be chosen in the displayed image.  The control node, yellow, subscribes to both the target 2.1 Testbed Architecture and the feedback topic and determines the control signal which is the desired translational and rotational velocity The testbed consists of a transparent phantom of a vascular for the guidewire. The control message is published to the system, which is either placed inside in an x-ray imaging action topic for each incoming feedback message. system (Artis Zeego, Siemens Helthineers) or mounted with a  The manipulation node, blue, subscribes to the action camera, a laptop with a 12-core, 2.6GHz processor and image topic and moves the guidewire accordingly. This architecture allows to easily substitute solutions for the different nodes e.g. replacing the manually controlled joystick in the control node with an autonomous control algorithm. Analogously, feedback and manipulation node can be replaced by a simulation. Additionally, multiple control nodes with different priorities can be utilized, e.g. to allow manual override via the joystick while navigating autonomously. 2.2 Tracking Algorithm Our novel guidewire tracking algorithm is placed inside the feedback node to retrieve guidewire tracking points from the image. The tracking algorithm uses the standard image processing functionalities of the publicly available Open CV [13] package. Instead of relying on complex algorithms our approach builds on manual parameter tuning utilizing two facts: First, the endovascular instruments are inserted at a static position which is known during the intervention. Second, endovascular instruments are slender devices, i.e. 𝑙 ≫ 𝑙 , 𝑧 𝑟 where 𝑙 is the straight length and 𝑙 is the radius of the 𝑧 𝑟 guidewire. Before the tracking loop is started, brightness and contrast of the incoming video stream can be manually adjusted. This allows to obscure the edges of the vascular tree and to enhance the visibility of the guidewire, when using a camera image. Figure 1: Phantom of vascular system with guidewire manipulator Likewise the thresholds for the edge detection can be hand under X-ray imaging (top), ROS topics (center) and guidewire tuned. Furthermore, the insertion area is manually marked (red tracking with joystick control pad in control room (bottom). A rectangle in Figure 2) and the coordinate system for the filled circle represents messages being published to the topic position of the guidewire is specified. In our setup the phantom while a ring represents a subscription 90 7. Retrieving and publishing tracking points that represent the centreline of the two edges of the guidewire. Figure 2 displays the successful retrieval of the guidewire from the fluoroscopy image. In evenly illuminated images it was also found possible to substitute steps 2 and 3 by a threshold filter. Setting all binary image values on the lower edge of the insertion area to zero after step 3 can additionally assert that a contour approximation point is placed inside the insertion area. Substituting step 6 by using the angle between the edges of the polygon approximation or adding a correction for guidewire bending during distance calculation were found to be less efficient. 2.3 Experiment Setup We validate our setup and evaluate the tracking algorithm by manually navigating through all branches of the vascular phantom using the joystick controller, as displayed in Figure 2, resulting in images with a variety of guidewire positions and lengths. During this task the delay of the image processing from retrieving a new image until returning the tracking points and the frequency of incoming tracking signals is measured. Also the number of time steps, where the algorithm is not able to retrieve a guidewire position from the image is counted. The experiment is conducted using both camera and x-ray imaging. Additionally, the experiment is repeated placing the vessel phantom on top of an anatomic phantom in x-ray imaging to Figure 2: Fluoroscopy image with additionally displayed vessel assess the capability of the tracking algorithm in a setup closer tree (top) and processed image with tracked guidewire to the clinical application. (bottom). of the vascular system is placed inside the x-z-plane with the guidewire pointing into the direction of the z-axis. This corresponds to the coordinate system of the x-ray imaging 3 Evaluation system. To track the guidewire, the following steps are performed: The tracking algorithm achieves an average processing delay 1. Transformation to grayscale image and adjustment of of less than 50ms under both camera and x-ray imaging which brightness and contrast as specified. is below the baseline of current research. The achieved control 2. Extracting the edges of the guidewire into a binary image frequency matches or even exceeds the stated requirements of using the Canny-Edge-Detector [14]. 4-10 Hz. All values are gathered in Table 1. 3. Applying a closing operation (consecutive usage of dilation and erosion filters) to fill up the edge-contours of Table 1: Average values for control frequency and delay, and the slender guidewire to achieve an enclosed contour. percentage of time-steps where the guidewire was not found 4. Retrieving the shapes and a polygon approximation of all enclosed contours in the binary image. Control Processing Guidewire not 5. Iterating through the approximation points of all contours Frequency Delay found until the guidewire is found as the contour with points inside the insertion area. Camera 17Hz 41ms 0.0% Tracking 6. Finding the tip of the guidewire by measuring the distance of both edges starting from the edge of the insertion area x-ray Tracking 10Hz 25ms 0.7% in positive coordinate direction. The tip is found as the point where both edges meet under equal distance. 91 While the guidewire was tracked in the camera image in References all time steps, there were a few time steps where the guidewire [1] World Health Organization (2020) Global Health Estimates could not be tracked or the guidewire tip was not found 2019 Summary Tables: Deaths by cause, age and sex, by correctly using the fluoroscopy image input. This was due to world bank income group, 2000-2019. rapid guidewire movements under relatively low image https://www.who.int/docs/default-source/gho- documents/global-health- frequency which resulted in a smeared guidewire. In the estimates/ghe2019_cod_global_2000_20194e572f53-509f- experiment an imaging frequency of 30p/s was used. The 4578-b01e-6370c65d9fc5.xlsx?sfvrsn=eaf8ca5_7. Accessed failure rate increases with a lower image frequency. 16 May 2022 The tracking algorithm failed to distinguish the guidewire [2] Karstensen L, Ritter J, Hatzl J et al. (2022) Learning-based autonomous vascular guidewire navigation without human from the background when the vessel phantom was placed on demonstration in the venous system of a porcine liver. top of the anatomic phantom, due to equal illumination of International Journal of Computer Assisted Radiology and guidewire and skeletal structures under x-ray imaging. The Surgery tracking algorithm also fails to extract the correct guidewire [3] Schegg P, Dequidt J, Coevoet E et al. (2022 - 2022) tip if the guidewire overlaps itself or is kinked by Automated Planning for Robotic Guidewire Navigation in the Coronary Arteries. In: 2022 IEEE 5th International maloperation. Conference on Soft Robotics (RoboSoft). IEEE, pp 239–246 [4] Kweon J, Kim K, Lee C et al. (2021) Deep Reinforcement Learning for Guidewire Navigation in Coronary Artery Phantom. IEEE Access 9:166409–166422. 4 Discussion https://doi.org/10.1109/ACCESS.2021.3135277 [5] Baert SAM, Viergever MA, Niessen WJ (2003) Guide-wire A testbed for endovascular robotic systems was presented. The tracking during endovascular interventions. IEEE Trans Med Imaging 22:965–972. utilization of the ROS2 communication protocol divides the https://doi.org/10.1109/TMI.2003.815904 navigation task into the subtasks of receiving a feedback about [6] Slabaugh G, Kong K, Unal G et al. (2007) Variational the instruments position, finding an appropriate manoeuvre Guidewire Tracking Using Phase Congruency. In: Springer, and executing this manoeuvre. This allows researchers to Berlin, Heidelberg, pp 612–619 [7] Chang P-L, Rolls A, Praetere H de et al. (2016) Robust focus on each individual task during development while Catheter and Guidewire Tracking Using B-Spline Tube Model having the ability to easily replace different solutions for all and Pixel-Wise Posteriors. IEEE Robot Autom Lett 1:303– other components. 308. https://doi.org/10.1109/LRA.2016.2517821 Furthermore, a tracking algorithm for guidewires was [8] Vandini A, Glocker B, Hamady M et al. (2017) Robust presented that can be implemented straightforwardly using guidewire tracking under large deformations combining segment-like features (SEGlets). Medical Image Analysis open source software packages. It meets the requirements for 38:150–164. https://doi.org/10.1016/j.media.2017.02.001 application in navigation control loops and reliably tracks the [9] Zhou Y-J, Xie X-L, Bian G-B et al. (2019) Fully Automatic guidewire in two-dimensional vascular phantoms. Dual-Guidewire Segmentation for Coronary Bifurcation The testbed allows researchers to evaluate control Lesion. In: 2019 International Joint Conference on Neural Networks (IJCNN). IEEE, Piscataway, NJ, pp 1–6 algorithms or robots for endovascular instruments by replacing [10] Gherardini M, Mazomenos E, Menciassi A et al. (2020) the respective ROS2 node with their solution. Algorithms that Catheter segmentation in X-ray fluoroscopy using synthetic successfully navigate through the presented physical testbed data and transfer learning with light U-nets. Computer are promising candidates for testing in phantoms with higher Methods and Programs in Biomedicine 192:105420. https://doi.org/10.1016/j.cmpb.2020.105420 complexity or even animals. The tracking algorithm can then [11] Wagner MG, Laeseke P, Speidel MA (2019) Deep learning easily be replaced by a professional solution. based guidewire segmentation in x-ray images. In: Gilat- Schmidt T, Chen G-H, Bosmans H (eds) Medical Imaging 2019: Physics of Medical Imaging: 17-20 February 2019, San Diego, California, United States. SPIE, Bellingham, Washington, USA, p 150 [12] Macenski S, Foote T, Gerkey B et al. (2022) Robot Operating Author Statement System 2: Design, architecture, and uses in the wild. Sci Research funding: This project is funded by the Ministry of Robot 7:eabm6074. Economics, Labour and Tourism Baden-Württemberg within https://doi.org/10.1126/scirobotics.abm6074 [13] Bradski G (2000) The OpenCV Library. In: Dr. Dobb's the framework of the Forum Gesundheitsstandort Baden- Journal of Software Tools Württemberg. [14] Canny J (1986) A Computational Approach to Edge Conflict of interest: Authors state no conflict of interest. Detection. IEEE Trans Pattern Anal Mach Intell PAMI-8:679– 698. https://doi.org/10.1109/TPAMI.1986.4767851

Journal

Current Directions in Biomedical Engineering – de Gruyter

Published: Jul 1, 2022

Keywords: endovascular; surgical robotics; testbed; tracking; guidewire

There are no references for this article.