This page shows a brief overview of my research interests. Students I am working with and ongoing/past projects are listed on separate pages
My work focuses on surgical robotics and image-guided control. My research goes from design of minimally invasive devices for targeted procedures up to control of those devices, mainly using visual servo control. My main contributions are in the following areas :
Mechatronic design : integration of distal actuators in minimally invasive surgical devices while respecting the sterilizability constraint ; sensors and imaging probe actuator design ; testing in realistic benchtop/in vivo conditions
Modelling : modelling of continuum robots ; modelling of interaction with the environment : soft tissue deformation ; modelling of gestures inside the operating room (teleoperation)
Control of surgical robots : visual servo control and associated image processing (exploitation of projective images, robust segmentation methods, estimation of positions and movements, 3D reconstruction and mosaicking) ; intuitive teleoperation for flexible robots.
Below are examples of different projects I've been involved in. A list of the related publications is available on this page, as well as on my google scholar profile
Image-based tracking of continuum robots - ongoing
Continuum robots are increasingly used in surgery thanks to their unique ability to provide tip dexterity while offering minimally invasive access. They are, however, difficult to model accurately. Recovering the shape of the instrument is therefore a topic of interest in the community. I'm currently working on developing tracking approaches requiring no other sensor than the endoscopic camera already present for navigating in the anatomy. Specifically, novel machine learning methods in which image information is fused with kinematic/dynamic information coming from models of the instruments are being developed, with the goal of online initialization of the tracker without needing any pre-training at the beginning of the surgery.
Robot-assisted endomicroscopy - 2010-2013
Endomicroscopy is an image modality which allows to take optical biopsies, i.e. without taking a tissue sample. The goal of the PERSEE project, in which I worked during my PhD, was to make a minimally invasive (5mm) robotized device to sweep the endomicroscopy probe in contact of the tissues in order to make large field of view images using a mosaicing algorithm. Challenges are microactuation of the probe at the micron-scale, compensation of physiological motion and control of the probe movements in order to make wide FOV mosaics with a controlled shape.
I developed methods to estimate the soft tissue deformations induced by the probe movements, visual servo control algorithms for compensating those deformations, and prototype minimally invasive instruments for in vivo validation. The end result is that the surgeon is able, with a 5 mm handheld instrument, to stabilize on an area of interest and then to produce large field of view mosaics in less than one minute by using the robotic control combined with the microactuation system embedded in the tip of the instrument.
Beating heart intracardiac cardioscopy-guided interventions - 2015-2019
Beating-heart interventions are typically less invasive than their open-heart counterparts, due to the absence of extracoporeal cardiopulmonary bypass and the minimally invasive approaches. Those approaches require, however, high dexterity as well as some visual guidance. The dexterity problem was solved by designing a concentric tube robot able to reach the anatomical target , while limiting undesired interactions with the anatomy.
In order to provide visual guidance, cardioscopic imaging was integrated in the prototype. This imaging modality allows visualizing tissues upon contact despite the presence of blood in the beating heart, thanks to an optically clear soft silicone window placed in front of the camera. It was validated first on handheld devices (see video below showing validation on a transapical paravalvular leak closure on a porcine model), and then placed at the tip of the concentric tube robot. The project is ongoing, and control methods based on cardioscopic images are being developed for stabilizing the interaction with the beating heart tissues.
Intuitive control of active catheters - 2014-2015
TAVI is a promising procedure in which catheters are used to navigate from the femoral artery to the aortic valve in order to replace it in the case of severe stenosis. Current catheters are difficult to maneuver and a direct consequence is an overuse of X-ray imaging and lenghty procedures. Within the CASCADE project, KU Leuven was (among other things) developing catheters with distal actuation for improving the procedure. In order to evaluate the effect of the control strategy and interface on the user efficiency, a simulation setup was developed, including a realistic catheter insertion simulator, a 4 DOF haptic master arm, a simulated fluoroscopic view, and a 3D rendering environment with navigation cues for guidance. Several mappings from the haptic master arm to the slave catheter were developed and tested, with a cohort comprising novices and expert surgeons.
The study showed that experts outperform novices in all experiments, which is expected. More importantly, it showed that visual guidance cues are highly beneficial for both cohorts. Unexpectedly, sometimes simpler control strategies are being misused by expert surgeons due to their previous experience with similar -- but slightly different still - gestures in clinical practice. This last result outlines the importance of user experience and training for selecting proper interfaces and control strategies.
Other projects I've been involved in :
Ureteroscopy (Engineering degree, 2009-2010). A ureteroscope is a flexible endoscope that can be inserted in the kidney via natural ways in order to treat some kidney diseases. By using a laser, one can treat kidney stones in a minimally invasive way. The problem is that the ureteroscope only has one distal degree of freedom while a second one would be nedded for efficient treatment. As a result, the procedure is long and tedious for the surgeon. In this project, I integrated smart actuators (Shape Memory Alloy wires) at the distal tip of a ureteroscope in order to be able to destroy the kidney stone efficiently. An algorithm to detect the stone was also developed and tested on pathological images, in order to control the laser movements with visual servoing.
Laparoscope holder robot (post-doc,2013). I work on a robotized laparoscope holder for assisting the surgeon during the operation. Particular attention has been paid on the alignment of the robot with the incision of the patient in order to minimize forces upon the patient's body wall. A dedicated algorithm was developed to find the best point possible for such alignment, the Optimal Pivot Point (OPP). It makes use of an external stereo camera placed in the operating room and looking at the incision, and a calibration procedure performed by the surgeon that takes less than 20 seconds. The gesture of the surgeon is modeled together with the projective cameras, and this knowledge is included in a RANSAC algorithm in order to extract the OPP in 3D. Experiments have reported a very high success rate (94.9%, failures are detected and discarded) and precision (1.85 mm in average). Preliminary experiments showed that this can be used tgether with a tracking of a robot in order to align it with the optimal pivot point at the incision.