Emmanuel Piuze

Phd Report - Neurovascular Segmentation

The identification of cerebral vessel lesions and malformations in medical images strongly correlates with risk factors for arterial embolisms (vascular blockage), ischemic strokes, and aneurysms, all of which are relatively common con- ditions that can be debilitating and fatal. Segmentation methods can support the clinician in extracting neurovasculature from medical images. These vascular representations are helpful in the development of safer treatments for cerebral conditions, for a more efficient preoperative planning, operating room and postoperative monitoring, and for intra- and inter-patient comparative studies. Common features of interest include vessel branch points, quantification of the spatial relationship between vessels, vessel segment extraction, vessel diameter measurements, and vessel complexity assessment. When a sequence of images is available, it can also be useful to quantify changes in these quantities over time or across patients, thereby opening up a lot of research potential. The main focus of this literature review is to present a selection of key articles covering different approaches employed in the vessel segmentation literature.

Report (PDF, 2.9MB)

M.Sc Thesis in Computer Science (Computer Graphics)

Hair-like patterns, comprised of dense collections of curve-like elements, are ubiquitous in our world. Consider examples arising in nature such as hair, fur, grass, feathers, and white matter fiber tracts in the brain, or those arising in man-made structures, such as textiles, optical fibers or threads. This thesis investigates an approach for generating, fitting, and interpolating such patterns. An emphasis is placed on hair modeling in computer graphics as the driving application for the methods proposed. Many intricacies of hair modeling are shared with those of fur, feathers, and grass modeling, and so the proposed methods translate to these and related cases as well.

Thesis (PDF, 24.6MB)

A Pseudo-Physical Framework for Mapping Outdoor Terrains

COMP 765 - Mobile Robotics project with Prof. Gregory Dudek. Winter 2011

This course was an introduction to the broad area of mobile robotics. Topics included environment mapping, localization and path planning. The course project consisted of a summary and/or suggested extensions to a selected research problem. I designed and implemented a pseudo-physical mapping framework from the ground up in Java, which included implicit fitting and visualization, energy minimization, curvature and stationary point estimation and numerical derivation and integration. Java OpenGL was used for rendering.

Report (PDF, 10MB)

A Physical Simulation of Tubulin Dynamics in Yeast Cell Division

COMP 761 - Computational Structural Biology project with Prof. Jérôme Waldispühl. Winter 2011

This course was an introduction to mathematical and computational methods in structural biology, and in algorithmic foundations of the approaches used in this field. The course project consisted of building a physical simulation of the dynamics of microtubules involved in yeast cell division. The framework was developed in Java/OpenGL.

Multi-Modality Image Registration by Maximization of Mutual Information

ECSE 626 - Statistical Vision project with Prof. Tal Arbel. Winter 2010

The course project involved finding an interesting algorithm or technique in the medical imaging literature, implementing it, and testing it on a dataset. I picked a 1997 paper by Maes et al. describing one of the first approaches using mutual information as a registration criterion in the brain registration literature: Multi-Modality Image Registration by Maximization of Mutual Information. The algorithm was implemented using Matlab.

Report (PDF, 1MB)

Particle systems, cloth simulation, motion capture reuse

COMP 599 - Computer Animation project with Prof. Paul Kry. Winter 2010

This course provided an introduction to computational techniques for generating animation. Topics included physics-based animation, implicit and explicit methods, collision detection and response, Lagrange multiplier and energy constraints, motion capture, rigid body motion, forward and inverse kinematics. My final project consisted of implementing a hard constraint physical simulation: a particle system cloth colliding with a sphere and sliding on a curtain-like rail.

Interactive Fine-tuning of Performance-driven Facial Animation

Undergraduate NSERC project with Prof. Paul Kry. Summer 2009

Facial animation is typically produced using a character rig, which consists of a geometric model and a set of parametric deformations. Motion capture can greatly simplify the animation of the rig parameters, and has the benefit of directly capturing all the subtle movements of a real face. With motion capture, the problem is reduced to finding a mapping from the captured performance to an existing rig. We present a method for interactively tuning motion capture driven facial animation.