
Statistical Learning
Medical image analysis provides a challenge for conventional machine learning approaches due to the nature of the imaging and pathologies involved. The project aims at exploring how to better exploit statistical learning for medical image analysis.Selected Publications
-
Proceedings of the First ICML Workshop on Machine Learning Meets Medical Imaging. Springer, Lecture Notes on Computer Science, volume 9487, 2015.
-
Acceptance Rate: 30% [Highlight] [video] Spectral Forests: Learning of Surface Data, Application to Cortical Parcellation. In Medical Image Computing and Computer Assisted Intervention (MICCAI), 2015.
-
Acceptance Rate: 30% [Highlight] Laplacian Forests: Semantic Image Segmentation by Guided Bagging. In Medical Computing and Computer Assisted Intervention (MICCAI), 2014.
-
Acceptance Rate: 32% Joint Statistics on Cardiac Shape and Fiber Architecture. In Medical Computing and Computer Assisted Intervention (MICCAI), 2013.

Spectral Shape Analysis
The study of complex anatomical structures requires a precise alignment of images (i.e., with a good overlap, structures in images can be compared). This alignment is, however, challenging when very large deformations exist. Instead, structures are studied using shape representations, or so called spectral signatures, that are invariant to external deformations. For example, deformed shapes have curiously similar representations in the spectral domain. Spectral signatures and representations are my current focus of research.Selected Publications
-
Acceptance Rate: 25% Spectral Demons - Image Registration via Global Spectral Correspondence. In European Conference in Computer Vision (ECCV), 2012.
-
FOCUSR: Feature Oriented Correspondence using Spectral Regularization - A Method for Precise Surface Matching. In IEEE Transactions on Pattern Analysis and Machine Intelligence (PAMI), 35 (9), 2013.

Accurate Surface Matching
The analysis of surfaces is important in many fields, particularly in neuroscience where accuracy is critical. Spectral graph theory provides elegant solutions to the problem of surface matching.
[Fast Brain Matching with FOCUSR]
Selected Publications
-
Acceptance Rate: 28% Fast Brain Matching with Spectral Correspondence. In Information Processing and Medical Imaging (IPMI), 2011.
-
FOCUSR: Feature Oriented Correspondence using Spectral Regularization - A Method for Precise Surface Matching. In IEEE Transactions on Pattern Analysis and Machine Intelligence (PAMI), 35 (9), 2013.

Atlas Construction
Statistical atlases contain information on the average structure of organs as well as their variabilities. Their constructions require the development of shape averaging tools as well as accurate image registrations.
[Morphing between 2 Brains]
Selected Publications
-
Oral Presentation
Best Paper Award Groupwise Spectral Log-Demons Framework for Atlas Construction. In MICCAI Workshop on Medical Computer Vision (MICCAI MCV), 2012. -
Acceptance Rate: 25% Spectral Demons - Image Registration via Global Spectral Correspondence. In European Conference in Computer Vision (ECCV), 2012.

Cardiac Fiber Architecture
The complex organization of the cardiac fibers plays a key role in the mechanical function and electrophysiology of the heart. Its study covers various research areas, including cardiophysiology as well as shape analysis and statistics of complex structures.Selected Publications
-
FQRNT Etudiant Etoile
Best Paper Award Human Atlas of the Cardiac Fiber Architecture: Study on a Healthy Population. In IEEE Transactions on Medical Imaging (TMI), 31 (7), 2012. -
Oral Presentation
Best Paper Award Statistical Analysis of the Human Cardiac Fiber Architecture from DT-MRI. In Functional Imaging and Modeling of the Heart (FIMH), 2011.

Image Segmentation
Image segmentation is a building block in image processing and analysis. In this context, fast, accurate and automatic solutions allow the processing of large quantities of images.Selected Publications
-
Spatio-Temporal Segmentation of the Heartin 4D MRI Images using Graph Cuts with Motion Cues. In Intenational Symposium in Biomedical Imaging (ISBI), 2010.
-
Acceptance Rate: 19% A Multilevel Banded Graph Cuts Method for Fast Image Segmentation. In IEEE International Conference in Computer Vision (ICCV), 2005.

Medical Image Visualization
Visualizing and navigating through medical images are important steps for radiologists to establish diagnostics and convey their analyses to professionals.
