Contact details
Dr Vincent Luboz
Department of Surgery & Cancer
Paterson Wing
St Mary's Campus
Tel: +44 (0)20 7886 6379
Email:
Dr Vincent Luboz
Current work
Vincent belonged to the Department of Biosurgery and Surgical Technology research staff as a Research Associate from March 2007 to February 2011. He was working on a vascular interventional radiology simulator with Fernando Bello. This project of the CRAIVE consortium was funded by EPSRC. The aim of this project was to develop and validate a computer generated virtual environment (VE) with variable virtual anatomy, in which the appearance, 'feel' and human factors of invasive radiological procedures (interventional radiology, IR) in patients can be reproduced and assessed. The final simulator encompasses needle puncture as well as guidewire and catheter insertion and manipulation.
The needle puncture is most of the time performed in the upper femoral artery part, and is called the Seldinger technique. Figure 1 shows an example of how this technique is simulated: with two Omni phantoms (one for the virtual needle, one for the virtual ultra sound probe) placed under a mirror reflecting the simulated patient body and the interactions.
Figure 1 – Left: Seldinger technique. Right: CRaIVE simulator.
More specifically, Vincent is working on four tasks:
- Developing, integrating and validating a complete virtual environment for training in vascular interventional radiology, encompassing needle puncture as well as guide wire and catheter insertion and manipulation (see Figure 1).
- Semi-automatically processing imaging data. Generation of variable virtual anatomy from patient specific data sets will be achieved via a direct facility to load DICOM data and semi-automatically produce a range of 3D geometry of vascular and surrounding tissue structures. Figure 2 presents a typical data set and a fly over the anatomy concerned in the simulator. Figure 3 and 4 show three different anatomies of the vasculature treated in interventional radiology. The needle will be introduced in the lowar part of the segmented dataset (iliac or femoral arteries). It will then be followed by the guidewires and the catheters.
Figure 2 – Typical data set from a dicom CT angiography. After the contrast adjustment, one can see the ribs, the heart, the pelvis, and the vasculature treated in this study: the aorta, the renal, the iliac and femoral arteries.
Figure 3 – Three different anatomies of the studied vasculature from femoral arteries (at the bottom) to aorta (in the middle) to neck vessels (at the top): A. non pathological aorta, B. two aortic aneurysms, and C. aortic dissection.
Figure 4 – Three different anatomies of the studied vasculature from femoral arteries (at the bottom) to aorta (in the middle) to neck vessels (at the top). From left to right: non pathological aorta, two aortic aneurysms, and aortic dissection.
- Simulating needle puncture and introduction of guide wire (Figure 5) and catheter into a blood vessel by combining suitable deformation models incorporating tissue resistance, tissue and vessel deformation, pathology and physiological pulsation.
- Developing a haptic device interface to give the trainee a realistic feel and feedback of the virtual procedure. To this end, a tracking and force feedback device, from the company Mentice, is used, see Figure 6.
Figure 6 - VSP Haptic Device from Mentice.
- Simulating real time contrast agent injection to allow highlighting the vessels under fluoroscopy. In figures 7 and 8, the vasculature is overlaid in pink to show how accurate the contrast is flowing in the vessels. In a real clinical set up, they would not be seen before the injection. In figure 8, the heart beat is visible thanks to the modelling of the speed of the blood flow and its interaction on the contrast medium.
Figure 7 - Contrast agent injection in the iliac artery.
Figure 8 - Contrast agent injection in the aorta.
- Simulating stenosis treatment by inflating a balloon and deploying a stent, Figure 9.
Background:
Vincent was previously holding a post doctoral position of Harvard Medical School in the Simulation Group, in the CIMIT laboratory and in the Massachusetts General Hospital, in Boston, MA. From December 2003 to March 2007, Vincent has worked on two projects: (1) the Truth Cube project, in order to characterize the mechanical properties of different materials, and (2) the Eve project, an endovascular simulator for training surgeons and fellows in interventional radiology for strokes and aneurisms. See his previous webpage for more details. His advisors were Stephane Cotin and Steve Dawson.
Before hand, Vincent received his PhD in Computer Science from the University Joseph Fourrier and the Institut National Polytechnique in Grenoble, France, in December 2003. He has done his PhD in the TIMC laboratory, team GMCAO, on the planning and surgery of dysthyroidian exophthalmia, see here for his PhD web page. His advisors were Yohan Payan and Jocelyne Troccaz.
His research interests are computer assisted surgery, biomechanics, surgical simulation/planning, and medical imaging/segmentation.
Teaching:
Lectures:
- Title: 3D visualisation and manipulation
lecture notes visualisation and manipulation for MSc
Course: Module - Information technology
Website: http://www1.imperial.ac.uk/medicine/about/divisions/sora/teaching/postgraded/mscsurgicaltech/
Degree: MSc. surgical technology, Department: Faculty of Medecine, Faculty: Imperial College London, Years: 2009/2010 and 2010/2011.
- Title: Image segmentation and pre-processing
lecture notes preprocessing and segmentation for MSc
Course: Module - Information technology
Website: http://www1.imperial.ac.uk/medicine/about/divisions/sora/teaching/postgraded/mscsurgicaltech/
Degree: MSc. surgical technology, Department: Faculty of Medecine, Faculty: Imperial College London, Years: 2009/2010 and 2010/2011.
- Title: Development of Simulation Tools
Course: Module 3 - Technology and bioengineering advances and clinical safety
Website: https://education.med.imperial.ac.uk/Years/4-0708/surgery/
Degree: BSc. surgery and anaesthesia, Department: Faculty of Medecine, Faculty: Imperial College London, Year: 2009/2010.
- Title: Development of Simulation Tools - Software
Course: Module 3 - Technology and bioengineering advances and clinical safety
Website: https://education.med.imperial.ac.uk/Years/4-0708/surgery/
Degree: BSc. surgery and anaesthesia, Department: Faculty of Medecine, Faculty: Imperial College London, Year: 2008/2009.
- Title: Development of Simulation Tools - Hardware
Course: Module 3 - Technology and bioengineering advances and clinical safety
Website: https://education.med.imperial.ac.uk/Years/4-0708/surgery/
Degree: BSc. surgery and anaesthesia, Department: Faculty of Medecine, Faculty: Imperial College London, Year: 2008/2009.
- Title: Appearance-based modelling
lecture notes appearance based modelling for MSc
Course: Advanced Graphics and Visualisation
Website: http://www.doc.ic.ac.uk/~dr/teaching/2008/Visualization/
Degree: Msc. in Advanced Computing, Department: Departement: Departement of Computing, Faculty: Imperial College London, Years: 2007/2008, 2008/2009.
- Title: Physics-based modelling II
lecture notes physics based modelling for MSc
Course: Advanced Graphics and Visualisation
Website: http://www.doc.ic.ac.uk/~dr/teaching/2008/Visualization/
Degree: Msc. in Advanced Computing, Department: Departement: Departement of Computing, Faculty: Imperial College London, Years: 2007/2008, 2008/2009.
- Title: 3D visualisation and manipulation
lecture notes visualisation and manipulation for MSc
Course: Module - Information technology
Website: http://www1.imperial.ac.uk/medicine/about/divisions/sora/teaching/postgraded/mscsurgicaltech/
Degree: MSc. surgical technology, Department: Faculty of Medecine, Faculty: Imperial College London, Year: 2008/2009.
- Title: Image segmentation and pre-processing
lecture notes pre-processing and segmentation for MSc
Course: Module - Information technology
Website: http://www1.imperial.ac.uk/medicine/about/divisions/sora/teaching/postgraded/mscsurgicaltech/
Degree: MSc. surgical technology, Department: Faculty of Medecine, Faculty: Imperial College London, Year: 2008/2009.
- Title: Developing VR Surgical Simulation Software
Course: Module 3 - Technology and bioengineering advances and clinical safety
Website: https://education.med.imperial.ac.uk/Years/4-0809/surgery/
Degree: BSc. surgery and anaesthesia, Department: Faculty of Medecine, Faculty: Imperial College London, Year: 2007/2008.
Tutorials (theoretical and practical):
- Tutorials for the Msc. in Surgical Technology in 2009/2010. Faculty: Imperial College London.
Website: http://www1.imperial.ac.uk/medicine/about/divisions/sora/teaching/postgraded/mscsurgicaltech/
- Tutorials for the Msc. in Surgical Technology in 2008/2009. Faculty: Imperial College London.
Website: http://www1.imperial.ac.uk/medicine/about/divisions/sora/teaching/postgraded/mscsurgicaltech/
- Tutorials for the Msc. in Advanced Computing in 2008/2009. Faculty: Imperial College London.
Website: http://www.doc.ic.ac.uk/~dr/index.php?id=51
- Tutorials for the Msc. in Advanced Computing in 2007/2008. Faculty: Imperial College London.
Website: http://www.doc.ic.ac.uk/~dr/teaching/2008/Visualization/
- Tutorials in Computer Sciences for the DEUG (years 1 and 2 of BSc) MIAS (Mathematique et Informatique Appliquees aux Sciences) in 2000/2001, 2001/2002, and 2002/2003. Faculty: Universite Joseph Fourrier (Grenoble 1).
Website: http://www.ujf-grenoble.fr/36392593/0/fiche___pagelibre_accueil/
Student projects:
2010:
- Blood Vessel Skeletonisation and Skeleton Based Fluid Simulation for Virtual Angiography in a Vascular Interventional Radiology Simulator (Petros Dracos, Department of Computing, MSc student).
Abstract:
Virtual reality endovascular simulators, introduced to acquire interventional radiology skills, rely heavily on efficient blood vessel representations due to the plethora of simultaneous tasks involved during an operation that have to be simulated in real-time.
Using the concept of skeletonisation, this project aims to produce anatomically accurate and efficient representations of the vasculature of a number of real patients. This is a major challenge in medical imaging and simulation as we are dealing with noisy datasets; also different patients, depending on their age and possible underlying conditions, have different anatomical geometry and characteristics.
Angiography uses injection of a contrast medium into blood vessels to visualise vascular anatomy and pathology for diagnosis and instrument guidance during endovascular procedures. In this project, we introduce a skeleton based methodology with constant O(1) time complexity for simulating contrast medium propagation in 3D virtual vasculatures. The contrast medium is modelled as a line of particles represented by spheres that propagate in the vessels purely based on a precomputed skeleton in the form of a centreline and are constrained by the forces produced by the initial injection and the blood flow.
Skeletonisation has been successfully applied to a number of datasets from which a suitable centreline skeleton has been obtained for skeleton based angiography. The injection of contrast medium in a number of virtual patients achieves a realistic rendering at an average of 34 frames per second while at the same time a catheter is being manipulated.
- Virtual Natural Orifice Transluminal Endoscopic Surgery Simulator – Instrument Modelling (Przemyslaw Korzeniowski, Department of Computing, MSc student).
Abstract:
Natural Orifice Transluminal Endoscopic Surgery (NOTES) is an emerging experimental technique in surgery that eliminates abdominal external incisions. It is considered as a next stage in Minimal Access Surgery development and it is expected to further reduce operation trauma, recovery time, and overall clinical costs.
The foundations of NOTES were laid by recent improvements in flexible endoscopy. Because this technique is at its early stage, very few efficient training programs are available for clinicians. Computer simulation of a flexible endoscope would contribute to surgical training without putting patients at risk, popularizing NOTES and keeping practitioners up to date with new methods.
This project aims to create such virtual simulator to help training NOTES surgeons. It is inspired by an existing catheterisation simulator created at Imperial College. The endoscope shaft and actuators models are based on a mass-spring model. A number of forces affect the model to realistically simulate the endoscope behaviour via a newly design collision response algorithm. The virtual instrument was also partially integrated with the SOFA framework to interact with deformable organ models.
According to tests conducted in a silicon phantom model, the overall performance of the virtual flexible endoscope yields the real instrument in a convincing way, in real time, and establishes promising foundations for further developments. This was confirmed by Mr. Sudip Sarker, a Consultant Surgeon in General, Colorectal, Endoscopic and Minimally Invasive Surgery, who tested the simulator.
- Virtual Natural Orifice Transluminal Endoscopic Surgery (NOTES) simulator – Tissue Modelling (Danial Sheikh, Department of Computing, MSc student).
Abstract:
The field of surgery has dramatically improved a lot over the past years. Open surgeries is not always needed and can be replaced with keyhole surgeries in many cases. Open surgery typically requires an incision large enough for the surgeon's hands to enter the patient, leaving behind big scars in particular. Keyhole surgeries, such as laparoscopic, arthroscopy and thoracoscopy (which belong to the broader field of endoscopy), leaves less traces of surgery. It is performed through small incisions (usually 0.5 – 1.5 cm). These types of surgery require high skill levels to ensure a safe and successful operation.
The current training method is limited to traditional apprenticeships, which comes with a high risk for the patient. An alternative way, and a better way, to train surgeons are to use computer-based simulation. A virtual environment lets the trainee to explore and even to trial-and-error with the simulated patient. At current time, there is no such approved method for Natural Orifice Transluminal Endoscopic Surgery (NOTES).
This project aims to create a virtual environment for clinicians to use. There has not been any commercial or validated simulator for this yet, as it still is in a research stage. This project focuses on the tissue deformation and collision detection with an instrument. As there is not previous work on this, there will be a high amount of research, testing and experimentation, until a last conclusion is drawn and the right models are selected.
In this project, a number of tools are used. The tissues are modelled by a particular spring model, and different deformation models, force fields, collision models and collision detection algorithms are combined to give the most realistic and efficient approximation of a real tissue or organ.
In the end, the results were convincing and promising according to clinician. We have managed to merge this project with the Instrument modelling-project, to get a proof of concept of a complete simulator for NOTES. This will be used for future development within this area.
2009:
- Simulation of Blood Flow and Contrast Medium Propagation for a Vascular Interventional Radiology Simulator (Yingtao Wang, Department of Computing, MSc student).
Abstract:
Medical education and training offers a virtual clinical environment for doctors to have a high qualified professional skills capability, and provide high-quality care to patients as well. Minimally invasive technique provides a revolutionized clinical therapy, which significantly reduces operation trauma, recovery time, and overall clinical costs. Interventional radiologists use this technique to open surgery through vasculature systems, and use catheterization to arrive the region of interest with the help of the medical imaging technique.
This project is mainly based on a framework of virtual catheterisation simulator (VCSim) developed in St Mary’s hospital, Imperial College London. This simulator models instruments using hybrid mass-spring model, and provides the interfaces with interventional radiology specific haptic device. Clinicians are able to use this simulator to practice the technique of catheterization through different realistic vascular models. However, during the virtual reality of fluoroscopy, the surfaces of vasculature cannot be seen clearly without the enhanced substance called medical contrast medium. With the effect of beating blood flow produced by heart rate, contrast medium mixes and propagates through vasculature for angiography. Furthermore, using injector of haptic device, contrast medium injected through catheter is an important training program of the virtual catheterisation simulator.
This project aims to simulate blood flow and contrast medium propagation in the vasculature segmented and three dimensional constructed from the real patient CT scans. The useful information such as bifurcation and cross-section of vasculature are obtained from the corresponding centreline generated from patient datasets and processed through a centreline reconstruction tool. The blood flow is controlled by beating heart model, performing on the reaction of contrast medium propagation. The contrast medium is modeled using smoothed-particle hydrodynamics, and is constrained by three forces produced by initial injection, collision with vessel walls and beating blood flow. Moreover, infinite times of injection are achieved, and an initial review system is developed. The simulation is tested in three different vasculatures, and theoretically supports complex arteriovenous with a large number of branches and sub-branches. Further, the simulation is evaluated both by clinicians and through comparison with real injection videos. The result is convincing and can be used as the foundation for model more realistic contrast medium performance under specific and complex blood flow cases.
- Virtual Catheterisation Simulator – Vessel deformations (Xiaoyuan Zou, Department of Computing, MSc student).
Abstract:
Recently, virtual simulation has been widely developed in many fields, such as flight operations, nuclear power, medical training and other industries as a training tool and method to evaluate performance. Realistic deformation is an important feature for virtual simulators, as most objects tend to deform when constraints are applied on them. It is the case in medical simulation and especially in interventional radiology.
Interventional radiology uses medical images to guide minimal invasive techniques using needles, wires and catheters to access vascular and organ systems. It aims at diagnosing and treating pathologies such as aneurysm or stenos. Because of vessels’ high elasticity, when instruments are navigated in the vasculature, it tends to deform. The deformation can be slight when the instruments are flexible but it can reach large deformations in the case of rigid instruments. In the past few years, a trend aiming at training interventional radiologists through virtual simulators emerged. A global, fast and realistic deformation is therefore required in a virtual simulator in reply to collision between instrument and vessel wall.
This thesis introduces a methodology for simulating vessel deformation in a virtual simulator. The novel model introduced in this paper is based on vessel centrelines. Once collision between a medical instrument and the vasculature is detected, all the displacements of surface triangles are controlled by their dominant skeleton points.
In our model, a mass-spring model is incorporated into a set of skeleton lines representing the patient vasculature. All the skeleton points are connected with their neighbours by springs and are subject to several forces: external force, dumping force, bending force and spring force. During a deformation, first, a set of displacements of the skeleton points is computed in response to the force which caused the deformation. Then, all the displacements are applied to the surface triangles proportionally. The skeleton model allows a fast application of the deformation since a skeleton point is responsible for the motion of a large set of vascular triangles.
The model has been successfully incorporated into an existing framework, which was developed in C++, VTK and FLTK. Several experiments have been performed on synthetic models and real patient datasets. They show that a fast and realistic deformation can be achieved in some patient models in the virtual simulator.
- Haptic Device for Virtual Endovascular Procedural Training (Adeel Sarwar, Department of Mechanical Engineering, BSc student).
Abstract:
Objectives: Creating a new haptic device for coaxial tracking of a guidewire and a catheter.
2008:
- Building of three-dimensional vascular models for interactive simulation in Interventional Radiology (Daniel King, Department of Computing MSc student).
Abstract:
Objectives: To provide a set of tools for generating 3D vascular models from patient datasets for interventional radiology simulations.
Background: As interventional radiology procedures become more common place, improved training methods must be developed to better prepare surgeons. Virtual reality simulators provide a safe and effective way for surgeons to gain the required experience. These simulations use models of the involved anatomy to provide visualisation and interaction. A set of tools to generate these 3D anatomical models from patient datasets will improve the variety of situations available to trainees, as well as allow experienced surgeons to practise specific procedures.
Methods: An application was developed in C++ to provide access to a set of interactive modelling tools using supporting functionality from FLTK, VTK, ITK, CGAL, and TetGen. Several segmentation algorithms were integrated and extended to allow the extraction of structures from datasets. A 2.5D extension of Lewis Griffin's hierarchical segmentation was developed. Another segmentation method using the Hessian matrix was also adapted to provide a fast method for finding vessel-like structures in an image. A distance transform skeletonisation algorithm was included to find a compact set of skeletal points from a segmented vessel. These skeletal points can then be connected together to form an approximate centreline through a minimum spanning tree algorithm. Mesh generation algorithms allow the creation of triangular and tetrahedral meshes from the segmented datasets.
Results: Experiments were conducted to test the effectiveness of the included algorithms. Hierarchical segmentation was found to be a good alternative to level set methods for vessels and other high contrast structures. By combining a vesselness measure, calculated from the Hessian matrix, with a fast marching segmentation, the vessels of a dataset can be quickly and automatically extracted. Vessels were automatically skeletonised in a short amount of time, and the reconstructions validated the skeletons as good representations of the original vessels. The meshing algorithms were found to generate good quality representations of anatomical structures that are suitable for simulations.
Conclusions: The structure of the program allows the user to easily select and combine tools in a sequence to generate anatomical models from specific patient datasets. The tools also allow the user to interactively improve the results from segmentation and mesh generation so that the models meet the simulation's needs and requirements. Future algorithms can be easily added to provide improved and extended functionality.
- Virtual Catheterisation Simulator - VCSim (Rafal Blazewski, Department of Computing MSc student).
Abstract:
Interventional radiologists use medical imaging to perform minimal invasive techniques involving needles, wires and catheters to access and treat vascular structures and internal organs such as the liver. These manipulations require high skill levels to ensure a safe and positive outcome. They are founded on core skills which must be learnt and automated before moving to more complex tasks.
The current training method relies on traditional apprenticeships, which are widely acknowledged to be risky for the patient and expensive for the hospital. Furthermore, because of the reduced number of hours in hospital, the cases accessible for trainees and therefore the key basis of core skills training have decreased. Several alternative training methods making use of computer-based simulation have been proposed to improve the training of the clinicians. Such virtual environments offer a safe framework to learn and practice the specific skills of interventional radiology without putting patients at risk. Furthermore, trainees are free to practice at their own pace and are given useful and objective feedback on their performance, thus quantifying their procedural competency. At the current time no validated method of training, and objectively assessing, core skills using computer-based simulation exists.
This project aimed to extend the previously prototyped simulation framework called the Virtual Catheterisation Simulator (VCSim), developed at Imperial College in 2006. Unfortunately, this previous framework was not able to adapt to the goals set for this project: introduction of a new interventional radiology specific haptic device, and catheterisation within realistic vascular models from real patient datasets. In consequence, the whole framework needed to be revised. In this project, the instruments are modelled as a set of particles linked by rigid springs, forming a hybrid mass-spring model. A unique approach to collision detection coupled with a realistic collision response resulted in an efficient and accurate approximation of the catheterisation. Four forces are interacting during the collision response, external force, spring force, 2D and 3D bendings, therefore increasing the realism of the instruments’ behaviour. The vascular model is obtained from the segmented CT scans of the vascular phantom and patient specific data. A basic approximation of small deformations of the vasculature was implemented and a number of possible improvements was proposed. The simulator has been evaluated on a silicon vascular phantom and tested on real patient datasets. Overall, the simulation results were very convincing and promising according to the interventional radiologists who tested the framework. It will now be used as the foundation for further developments of the simulator in order to allow teaching core skills in interventional radiology.
- Generation of 3D Models of Anatomy for the Simulation of Endoluminal Minimal Access Therapy (Nizar Din, Department of Surgery BSc student).
Abstract:
Objectives: To generate a library of 3D arterial models for interventional radiology simulation and to evaluate the most accurate and efficient segmentation algorithm for vessel generation.
Background: The trend towards minimally access therapy has necessitated a realistic endovascular simulator. Existing endovascular simulators fail to accurately streamline the process of generating three-dimensional geometric models directly and rapidly from patient-specific CTA scans. For this to occur, appropriate segmentation algorithms, optimised scanning protocols and explicit criteria for aortic delineation are required.
Methods: A clearly defined segmentation criteria was established and twenty CTA-scans were obtained. Four different segmentation methods were applied to each patient. ITK-SNAP manual segmentation was used as the reference point for comparison. Three semi-automatic segmentations were also applied: ITK-SNAP Level-set, LEEDS level-set and hierarchical segmentation, and compared with the reference segmentation using data-analysis software, MESH. Segmentation time and visual validation were also analysed.
Results: Of the 20 patient datasets, ITK-SNAP level-set, LEEDS level-set and hierarchical demonstrated, compared to manual segmentation, a mean Hausdorff distance of 2.36mm(RMS 7.45mm), 4.93mm(RMS 14.96mm) and 2.37mm(RMS 0.371mm) respectively, with ITK-SNAP and hierarchical segmentation statistically different to Leeds’ level-set (P<0.05). On average, manual segmentation took 527.5mins, ITK-SNAP 119.5mins, hierarchical segmentation 104.2mins and LEEDS level-set 20mins.
Conclusions: Of the compared segmentation techniques, the hierarchical algorithm strikes the best balance between Hausdorff-distance and time-efficiency. Suitably realistic models obtained from this project will be incorporated into a new-generation of operative simulation, with potential applications in training, assessment and mission-rehearsal in minimally-access therapies.
