Adaptive Dictionary Learning for Competitive Classification of Multiple Sclerosis Lesions

Participants : Hrishikesh Deshpande, Pierre Maurel, Christian Barillot.

The manual delineation of Multiple Sclerosis (MS) lesions is a challenging task pertaining to the requirement of neurological experts and high intra- and inter-observer variability. It is also time consuming because large number of Magnetic Resonance (MR) image slices are needed to obtain 3-D information. Over the last years, various models combined with supervised or unsupervised classification methods have been proposed for segmentation of MS lesions using MR images. Recently, signal modeling using sparse representations (SR) has gained tremendous attention and is an area of active research. SR allows coding data as sparse linear combinations of the elements of over-complete dictionary and has led to interesting image recognition results. In this work, we have proposed to use a sparse representation and an adaptive dictionary learning paradigm to automatically classify Multiple Sclerosis (MS) lesions from MRI. In particular, we investigate the effects of learning dictionaries specific to the lesions and individual healthy brain tissues, which include White Matter (WM), Gray Matter (GM) and Cerebrospinal Fluid (CSF). The dictionary size plays a major role in data representation but it is an even more crucial element in the case of competitive classification. We present an approach that adapts the size of the dictionary for each class, depending on the complexity of the underlying data. The proposed algorithm is evaluated on 3-D multi-channel MR images demonstrating improved classification.

Predictive Value of Imaging Markers at Multiple Sclerosis Disease Onset Based on Gadolinium- and USPIO- Enhanced MRI and Machine Learning

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

Robust detection of multiple sclerosis lesion from intensity-normalized multi-channel MRI

Participants : Yogesh Karpate, Olivier Commowick, Christian Barillot.

Multiple sclerosis (MS) is a disease with heterogeneous evolution among the patients. Better understanding of the disease will lead to improved patient-adapted therapeutic strategies. We propose a novel paradigm to detect MS lesions based on a statistical framework which consists of detection based on differences between multi-channel MRI of patients and controls. This framework fused with intensity standardization was applied to the study of MS and highlighted the great interest of quantitative MRI measurements for a better characterization of MS. Experimental results demonstrate that our technique accurately detects significant differences in lesions consequently improving the results of MS lesion detection. This work has been accepted to SPIE Medical Imaging 2015.

Multiple Sclerosis Lesions Recognition: One Class Learning Approach

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

We have developed an automatic algorithm for the detection of multiple sclerosis lesions (MSL) from multi-sequence magnetic resonance imaging (MRI). We build a probabilistic classifier that can recognize MSL as a novel class, trained only on Normal Appearing Brain Tissues (NABT). Patch based intensity information of MRI images is used to train a classifier at the voxel level. The classifier is in turn used to compute a probability characterizing the likelihood of each voxel to be a lesion. This probability is then used to identify a lesion voxel based on simple Otsu thresholding. This work has been submitted to ISBI 2015.