11:30
Surgical Intervention
Chair: Gabrielle Tuijthof
11:30
15 mins
|
Toward Data-driven Control of Origami-Inspired Actuators for Colonoscopy
Milad Hayati, Mostafa Atalla, Giulio Dagnino, Momen Abayazid
Abstract: Origami-inspired actuators are foldable, lightweight mechanisms engineered to replicate the flexible and precise movements characteristic of origami structures. Their capacity to fold into complex configurations, combined with these features, offers great promise in the field of soft robotics, making them particularly suitable for precision-demanding applications such as minimally invasive endoscopic interventions. In this application, the adaptability of origami actuators can potentially enable safer navigation through complex pathways, allowing endoscopes to maneuver through the body with greater control and minimal tissue impact. While origami actuators often demonstrate more predictable, linear dynamics than silicone-based actuators, achieving precise control over their movements remains challenging due to their complex shapes, material variability, and unique motion characteristics. Traditional control methods often result in inaccuracies when faced with uncertain disturbances, as in colonoscopy, where the actuator must adapt to various anatomical bends and flexures to ensure safe navigation.
This study explores data-driven approaches to develop a precise and adaptable control system for origami actuators. Using supervised learning and predictive modeling techniques, this research aims to design a closed-loop control system that utilizes real-time sensor data to optimize the actuator’s movements. Through data-driven learning, the control system enables the actuator to achieve target configurations with enhanced precision and adaptability. In colonoscopy, this data-driven control strategy can potentially enhance safety by autonomously adjusting the actuator’s response to unexpected behaviors, such as looping or path deviations, thereby maintaining safe interaction with surrounding tissues. State-of-the-art research in soft robotics suggests that data-driven control methods are superior in trajectory tracking, response speed, and adaptability over traditional approaches. Leveraging these insights, our study will focus on developing and testing a data-driven control strategy specifically for origami-inspired actuators designed for colonoscopy application. This control strategy aims to enhance the precision and stability of actuator tip bending and positioning, which are critical for autonomous navigation in endoscopic procedures. These potential improvements can reduce the risk of tissue injury and empower the interventionists with navigation autonomy that enhances the accuracy and control of the endoscope, making endoscopic procedures safer and more effective for both diagnostic and therapeutic applications.
|
11:45
15 mins
|
Towards reliable handheld optical microcirculatory blood flow Imaging
Ata Chizari, Mirjam J Schaap, Tom Knop, Marieke M.B. Seyger, Wiendelt Steenbergen
Abstract: Background: Handheld laser speckle contrast imaging (LSCI) plays a vital role in clinical environments; however, motion artifacts (MA) can undermine the reliability of perfusion images. Existing methods for preventing and suppressing MA are often impractical or overly complex. While machine vision techniques show promise in enhancing medical imaging quality, their application in mitigating MA remains largely unexplored.
Objective: This study introduces an innovative linear regression-based method for motion artifact correction (MAC) in LSCI, validated through measurements on psoriasis patients [1].
Methods: We conducted paired handheld and mounted LSCI measurements on 14 psoriasis lesions using the handheld perfusion imager (HAPI), previously validated for eye safe and user friendly clinical investigation [2]. By delineating lesion boundaries for clinical assessment, the HAPI utilized a monochromatic camera for both speckle imaging and motion detection, thus simplifying hardware requirements [3]. We accurately estimated the relative displacements between the test object and the LSCI probe, enabling the application of MAC to the perfusion images by correlating the calculated speed with local perfusion.
Results: Linear regression of spatial perfusion and on-surface speed extrapolated the zero-speed point as the predicted motion artifact-corrected value for each location. Based on this correction, the discrepancy in mean perfusion between handheld and mounted modes significantly decreased (median error of 14.2 perfusion units (p.u.) on lesions before correction (p<0.0005), compared to 0.5 p.u. after correction (p=0.2)).
Conclusions: Our findings support the efficacy of handheld LSCI and validate our MAC approach in a psoriasis context. We address one of the two primary causes of MA—on-surface speeds—and successfully correct mean perfusion, assuming constant temporal perfusion at each site.
Significance: We present a practical, non-contact, marker-free technique for reliable handheld perfusion imaging, paving the way for enhanced clinical applications in dermatology, burn treatment, and plastic surgery [4].
References:
[1] A. Chizari, et al., “Mitigation of Motion Artifacts in Handheld Laser Speckle Contrast Imaging Illustrated on Psoriasis Lesions,” IEEE Trans Biomed Eng, pp. 1–9, 2024, doi: 10.1109/TBME.2024.3438375.
[2] M. J. Schaap et al., “Perfusion measured by laser speckle contrast imaging as a predictor for expansion of psoriasis lesions,” Skin Research and Technology, 2021, doi: 10.1111/srt.13098.
[3] A. Chizari, et al., “Handheld versus mounted laser speckle contrast perfusion imaging demonstrated in psoriasis lesions,” Sci Rep, vol. 11, no. 1, p. 16646, Aug. 2021, doi: 10.1038/s41598-021-96218-6.
[4] A. Rook, et al., “Handheld wireless laser speckle contrast imaging (LSCI) during DIEP flap breast reconstruction: a pilot study,” in Optical Diagnostics and Sensing XXIV: Toward Point-of-Care Diagnostics, J. S. Baba and G. L. Coté, Eds., SPIE, Mar. 2024, p. 18. doi: 10.1117/12.3001925.
|
12:00
15 mins
|
Assessment of Multiple Budget Handheld 3D Cameras for Breast Imaging Purposes withing Reconstructive Surgery
Judith Waldner-Troost, Rob van Doremalen, Ruud Verdaasdonk
Abstract: Introduction: Three-dimensional (3D) surface imaging of the breasts gains relevance within reconstructive medicine. The medically certified Vectra XT system provides good quality, accuracy and has a proven reproducibility. However, this system is expensive and upright stationary. Low-cost non-medical handheld 3D scanners could provide a less expensive or mobile alternative. The aim of this study was to assess three low-cost, non-medical, handheld 3D scanners on clinical applicability for 3D breast imaging in comparison to the established Vectra XT. These scanners were assessed on 1; Accuracy, 2; Reproducibility, and 3; Usability.
Methods: The Vectra XT, the Structure Sensor Pro, the iPhone X, and Revopoint Pop 2 were used during this study. Our first experiment used a clay phantom of a breast within a research environment. This phantom was scanned five times by 6 observers. For the second experiment six healthy female volunteers were scanned three times by one operator in a clinical environment. Accuracy was determined by comparing the volume of the scanned phantom with the true volume of the phantom. Reproducibility was determined by deviation maps of every individual scanner.
Results: The accuracy of the Revopoint Pop 2 is similar to the accuracy of the Vectra. The Structure on the other hand shows the lowest accuracy. Both the Vectra and the Revopoint Pop 2 differed only 5 cm3 from the true volume of the phantom. The Structure Sensor Pro, together with the Vectra XT, shows the highest reproducibility in clinical setting with a mean deviation of 3.4 ± 0.8 mm. Lastly, the usability of the Structure Sensor Pro was highest.
Conclusion:The Revopoint Pop 2 seems to have the highest accuracy and reproducibility during the phantom experiment. However, during the experiment in the clinical environment the Revopoint Pop 2 had a below average reproducibility. A potential explanation could be the larger area the scanner had to scan during the clinical experiment. The Structure has the lowest reproducibility in the phantom study, but scores best in de clinical study. More experiments have to be conducted before recommendations for the optimal handheld scanner in clinical setting can be done.
|
12:15
15 mins
|
Active Lubrication of Endoluminal Instruments Using Ultrasonic Waves
Mostafa A. Atalla, Michaël Wiertlewski, Aimée Sakes
Abstract: Minimally invasive endoluminal procedures use catheters and endoscopes that are guided through body lumens to perform interventions in internal anatomical structures, resulting in an inevitable frictional interaction between the instrument and the luminal walls. While this friction enhances stability during the intervention, it poses a risk of damaging the luminal wall during navigation, leading to post-operative complications including bleeding, perforation and infectious diseases, to name a few. To mitigate the risk of adverse complications, we propose a new concept of actively-lubricated instruments capable of transitioning from low friction during navigation to high friction for increased stability while performing the intervention, allowing for safer, yet more stable endoluminal interventions. The proposed actively-lubricated instrument features discrete friction control modules along its shaft, enabling localized real-time friction control on demand. These modules use transverse ultrasonic micron-level surface waves to induce fluid film lubrication at the modules-tissue interfaces, enabling active control of friction during the procedure. In a catheter use case, our experiments demonstrated that the prototype effectively reduced friction by 42% on ex-vivo porcine aorta tissue and 82% on rigid substrates, representing its potential performance on healthy and calcified tissue, respectively. This result underscores the feasibility of the design and its potential to improve the safety and efficacy of minimally invasive endoluminal procedures.
|
12:30
15 mins
|
5DoF pose estimation of Intracardiac Echocardiography in 2D X-ray images during tricuspid valve repair
Annelies Severens, Richard Lopata, Midas Meijs, Vipul Pai Raikar
Abstract: Tricuspid valve regurgitation presents significant challenges due to high surgical risk and poor outcomes of conventional treatment. Transcatheter therapies offer a safer, minimally invasive alternative. Current image-guiding methods, like transesophageal echocardiography, face limitations such as poor visibility due to shadowing from left-sided prostheses and thoracic pathology. Intracardiac echocardiography (ICE) enables tricuspid valve therapy by providing improved imaging insights into the complex 3D shape of the valve. However, the adoption of 3D ICE is limited by the complexity of catheter manipulation, steep learning curves, and lack of available tracking algorithms.
This study aims to overcome these challenges by developing a solution for accurately locating the ICE transducer within X-ray fluoroscopy (XRF) images to facilitate real-time registration and improve procedural efficiency. A deep learning-based framework is proposed to estimate the five degrees of freedom (5DoF) pose of the ICE probe in 2D XRF images. The method consists of a two-step pipeline: first, coarse localization is performed using the YOLOv5 object detection network, followed by refined pose estimation on a smaller image patch using ResNet34. Due to limited in vivo data, a synthetic dataset was generated using digital reconstructed radiographs to simulate realistic probe positioning and orientations in X-ray projections. The models were trained on synthetic data and validated on both synthetic and manually annotated in vivo images.
Validation from four clinical cases showed that the framework provides accurate and reliable probe localization, showing strong detection performance and accurate 5DoF pose estimation in various clinical settings. The object detection model achieved perfect detection for the synthetic dataset, and high precision (0.97) and recall (0.89) for the in vivo cases. The pose estimation model showed mean absolute position errors of less than 1 mm on both datasets and rotation errors of less than 4° on the synthetic data, and under 7° for the in vivo dataset.
This robust, real-time solution for ICE pose estimation allows for seamless integration of ICE and XRF images. By facilitating automatic detection and pose estimation of the ICE probe, this framework aims to enhance the efficiency and accessibility of tricuspid procedures.
|
12:45
15 mins
|
Design and Evaluation of a Ball Spline Wasp-Inspired Needle
Jette Bloemberg, Zola Fung-A-Jou, Aimée Sakes, Paul Breedveld
Abstract: Introduction: In percutaneous interventions such as focal laser ablation to treat prostate cancer, needles are used to reach target locations inside the body. However, when the needle is pushed through the tissue, forces arise at the needle tip and along the needle body, making the needle prone to buckling. Recently, needles that prevent buckling inspired by the ovipositor of female parasitic wasps have been developed. Building on these needle designs, this study proposes a manual actuation unit that allows the operator to drive the wasp-inspired needle through stationary tissue. Materials and methods: The needle consists of six 0.3-mm rods, of which one is advanced while the others are retracted. The advancing needle segment has to overcome a cutting and friction force while the retracting ones experience a friction force in the opposite direction. The actuation unit moves the needle segments in the required sequence using a low-friction ball spline mechanism. The moving components of the needle have low inertia, and its connection to the actuation unit using a ball spline introduces a small friction force, generating a small push force of 0.18 N ± 0.12 N on the needle that facilitates the needle's propulsion into tissue while preventing needle buckling. Results: Experimental testing evaluated the needle's ability to move through 15-wt% gelatin tissue phantoms for different actuation velocities. It was found that the needle moved through the tissue phantoms with mean propulsion efficiencies of 65%, 69%, and 71% for actuation velocities of π, 2π, and 3π rad/s, respectively. Discussion and conclusion: The propulsion efficiency of the needle evaluated in 15-wt% gelatin was higher for the conditions where the central ball spline was able to move like intended (i.e., 68% ± 5.9%) than for the conditions where the central ball spline was fixed (i.e., 39% ± 7.0%). This indicates that the actuation system's low net insertion force helps propel the needle through the tissue with a high propulsion efficiency and without buckling. The needle actuation system design is a step forward in developing a wasp-inspired needle for percutaneous interventions that prevents buckling.
|
|