Overview

The Medical Image Analysis group conducts theoretical and applied research in image processing, computer vision, and artificial intelligence for the analysis of medical image datasets.

A particular focus of research is given to orthopaedic research and applications. However, the spectrum of topics and research interests is broad and open to other fields where medical image analysis techniques can provide solutions to biomedical-related problems.

SOFT TISSUE SIMULATION FOR CRANIO-MAXILLOFACIAL (CMF) SURGERY

Development of methodologies to predict Cranio-Maxilo-Facial surgery outcomes involving soft tissues deformations by creating robust, yet computationally inexpensive mathematical models that can be integrated into clinical practice. We have succesfully developed software to simulate patient-specific soft tissue deformation that are appropriate for integration into the clinical workflow. The developed methods highlight patient-specific muscle structure prediction, inhomogeneity, sliding contact, as well as tissue anisotropy. This project is part of the Co-Me network (http://co-me.ch/).

 In the software section (right hand menu), a tool for surface to surface distance computation is available for research purposes.

BRAIN IMAGE AND BRAIN IMAGE TUMOUR ANALYSIS AND SIMULATIONS USING HIGHLY EFFICIENT CUDA-ENABLED TECHNOLOGIES

Magnetic Resonance Imaging (MRI) is a powerful image modality that encompasses rich anatomical and physiological information at a high resolution. During the last years our group has developed several methodologies to analyze MRI images with focus on fast multimodal non-rigid image registration and multimodal image segmentation for brain image tumor analysis studies. The developments are driven by clinical requirements such as computation speed, robustness, use of standard clinical imaging protocols, and task-specific accuracy.

During 2011, application of our developments for the treatment of epileptic patients using image fusion of fMRI/MRI/CT imaging has been conducted in collaboration with our clinical partners at the Inselspital, Univ. Bern.

The problem of tissue classification in brain tumor studies has been also focus of research and developments in our group during the last years. Previously we have presented an algorithm for multimodal brain tissue segmentation in grade III and IV glioblastoma cases. The algorithm considers the standard clinical imaging protocol as well as the classification of healthy brain tissues (i.e. gray matter, white matter, csf) and tumor layers (necrotic, active and edema regions). In collaboration with the computational engineering group of our institute, a multiscale model of brain tumor growth and brain tissue classification has been developed into a unified framework for atlas-based brain tumor bearing segmentation.

In the software section (right hand menu) different software tools for automatic brain extraction from MRI, tissue segmentation and registration are available for research purposes.

COMPUTATIONAL ANATOMY TECHNIQUES FOR ORTHOPAEDIC RESEARCH

Computational anatomy enables analysis of biological variability on a population. Using statistical mathematical techniques, models can be built to represent the typical shape of an anatomical structure and the predominant patterns of variability across a given population.

In orthopaedic research we have used these techniques for automated patient-specific and model-based bone segmentation.
During 2011, we have developed algorithms to augment the information included in these models in order to assist radiologists in the task of region identification and generation of scanning plan in bone MRI imaging.

Computational anatomy techniques have also been developed to study the variability of bone shapes in a population using shape descriptors directly related to the clinical application and the existing knowledge of the clinicians about the anatomy. In this sense our goal during 2011 was to drive the development of these models using clinically known shape descriptors. This has established a new methodology that is anatomy-aware and driven by clinical information.

Last but not least, we continue developing algorithms to perform population-based orthopaedic implant design. During 2011 our efforts concentrated on taking into account both the quality of bone-implant fitting as well as the minimization of intra-operative implant reshaping. As result, we have been able to propose algorithms for population-based implant design, and patient-specific implant reshaping considering current devices used in
clinical practice.

For further information please contact Mauricio Reyes.