Section: New Results

Medical Image Computing in Brain Pathologies

Detection of dysplasia and heterotopia

Participants : Elise Bannier, Camille Maumet, Jean-Christophe Ferré, Christian Barillot.

Focal cortical dysplasia and heterotopias are a recognized cause of epilepsy. Indication for surgery relies on precise localization and delineation. However, visual depiction of focal cortical dysplasia and heterotopias is difficult, time-consuming and reader dependant. Several 3D T1 voxel based morphometry methods have been proposed to automatically identify and suggest potential abnormalities to the reader. Several studies have shown the ability of Double Inversion Recovery imaging to detect intracortical lesions in MS and Epilepsy. In this study we propose to evaluate the ability of Double Inversion Recovery voxel based analysis to detect cortical and juxtacortical lesions in pharmaco resistant epileptic patients. This work was performed in collaboration with Arnaud Biraben, Anca Pasnicu and Eduardo Pasqualini, Béatrice Carsin-Nicol [24] .

MRI Estimation of $T_1$ Relaxation Time Using a Constrained Optimization Algorithm

Participants : Fang Cao, Olivier Commowick, Elise Bannier, Jean-Christophe Ferré, Gilles Edan, Christian Barillot.

We propose a new method to improve $T_1$ mapping with respect to the popular DESPOT1 algorithm. A distance function is defined to model the distance between the pure signal and the measurements in presence of noise. We use a constrained gradient descent optimization algorithm to iteratively find the optimal values of $T_1$ and $M_0$. The method is applied to MR images acquired with 2 gradient echo sequences and different flip angles. The performance of $T_1$ mapping is evaluated both on phantom and on in vivo experiments [25] .

Characterization and Modeling of Multidimensional MRI signatures in Multiple Sclerosis in clinically isolated syndromes.

Participants : Yogesh Karpate, Olivier Commowick, Gilles Edan, Christian Barillot.

Clinically Isolated Syndrome data contribute to critical factors in obtaining meaningful precursor and predictors of Multiple Sclerosis. Current methodologies don't go beyond segmentation and which generalize poorly over multi-modal MRI data. The project objective is to research and develop a framework for characterization and modeling of multidimensional MRI signatures in clinically isolated syndrome(disease's onset),based on earlier and concurrent research and developments in the lab. In on going work an algorithmic framework is being developed to address the MS lesions' classification ,identification and retrieval in USPIO-6 database.

As part of a battery of pre- processing techniques ,the module for intensity normalization of MRI volumes based on Spatio-Temporal Robust Expectation Maximization (STREM) is developed. This work is primarily based on 3 MRI modalities viz T1-w,T2-w and FLAIR. Complementary to this work , an another intensity normalization algorithm is devised based on parametric robust as well as efficient estimation by minimizing a density power divergence (beta divergence).The proposed method is indexed by a single parameter alpha which controls the trade-off between robustness and efficiency. The methodology affords a robust extension of maximum likelihood estimation for which alpha tends to be zero. Choices of alpha near zero afford considerable robustness while retaining efficiency close to that of maximum likelihood.

Moving forward, to facilitate accurate lesion tracking,features must be selected which are robust to photo-metric and geometric distortions. Energy measures are used to capture lesion's multiscale orientation structure in space.To illustrate utility with respect to a lesion detection, we have developed descriptor like local energy based on 3D steerable wavelets. This will be followed by the rigorous empirical evaluations of the resulting algorithm yielding better lesion identification and retrieval.

Multiple Sclerosis Lesions Evolution in Patients with Clinically Isolated Syndrome.

Participants : Alessandro Crimi, Olivier Commowick, Gilles Edan, Christian Barillot.

Multiple sclerosis (MS) is a disease with heterogeneous evolution among the patients. Some classifications have been carried out according to either the clinical course or the immunopathological profiles. Epidemiological data and imaging are showing that MS is a two-phase neurodegenerative in inflammatory disease. At the early stage it is dominated by focal in inflammation of the white matter (WM), and at a latter stage it is dominated by diffuse lesions of the grey matter and spinal cord. A Clinically Isolated Syndrome (CIS) is a first neurologic episode caused by in inflammation/demyelination in the central nervous system which may lead to MS. Few studies have been carried out so far about this initial stage. Better understanding of the disease at its onset will lead to a better discovery of pathogenic mechanisms, allowing suitable therapies at an early stage. We propose a new data processing framework able to provide an early characterization of CIS patients according to lesion patterns, and more specifically according to the nature of the inflammatory patterns of these lesions. Our method is based on a two layers unsupervised clustering. Initially, the spatio-temporal lesion patterns are classified using a tensor-like representation. The discovered lesion patterns are then used to identify group of patients and their correlation to one year follow-up total lesion loads, which is so far the only image-based figure that can potentially correlate to future evolution of the pathology. We expect that the proposed framework can infer new prospective figures from the earliest imaging sign of MS since it can provide a classification of different types of lesion across patients [30] .