Miguel Guevara

Reproducibility of superficial white matter tracts using diffusion-weighted imaging tractography

ABSTRACT Human brain connection map is far from being complete. In particular the study of the superficial white matter (SWM) is an unachieved task. Its description is essential for the understanding of human brain function and the study of pathogenesis triggered by abnormal connectivity. In this work we automatically created a multi‐subject atlas of SWM diffusion‐based bundles of the whole brain. For each subject, the complete cortico‐cortical tractogram is first split into sub‐tractograms connecting pairs of gyri. Then intra‐subject shape‐based fiber clustering performs compression of each sub‐tractogram into a set of bundles. Proceeding further with shape‐based clustering provides a match of the bundles across subjects. Bundles found in most of the subjects are instantiated in the atlas. To increase robustness, this procedure was performed with two independent groups of subjects, in order to discard bundles without match across the two independent atlases. Finally, the resulting intersection atlas was projected on a third independent group of subjects in order to filter out bundles without reproducible and reliable projection. The final multi‐subject diffusion‐based U‐fiber atlas is composed of 100 bundles in total, 50 per hemisphere, from which 35 are common to both hemispheres. HIGHLIGHTSWe propose an hybrid method for the study of the reproducibility of superficial white matter bundles of the whole brain, using diffusion‐weighted imaging.The method combines cortical parcellation and fiber clustering in order to determine reproducible well‐defined bundles across subjects.A multi‐subject atlas of 100 reproducible bundles is finally created, from which 35 are common to both hemispheres.

Spatial normalization of brain images and beyond

The deformable atlas paradigm has been at the core of computational anatomy during the last two decades. Spatial normalization is the variant endowing the atlas with a coordinate system used for voxel-based aggregation of images across subjects and studies. This framework has largely contributed to the success of brain mapping. Brain spatial normalization, however, is still ill-posed because of the complexity of the human brain architecture and the lack of architectural landmarks in standard morphological MRI. Multi-atlas strategies have been developed during the last decade to overcome some difficulties in the context of segmentation. A new generation of registration algorithms embedding architectural features inferred for instance from diffusion or functional MRI is on the verge to improve the architectural value of spatial normalization. A better understanding of the architectural meaning of the cortical folding pattern will lead to use some sulci as complementary constraints. Improving the architectural compliance of spatial normalization may impose to relax the diffeomorphic constraint usually underlying atlas warping. A two-level strategy could be designed: in each region, a dictionary of templates of incompatible folding patterns would be collected and matched in a way or another using rare architectural information, while individual subjects would be aligned using diffeomorphisms to the closest template. Manifold learning could help to aggregate subjects according to their morphology. Connectivity-based strategies could emerge as an alternative to deformation-based alignment leading to match the connectomes of the subjects rather than images.

Clustering of Whole-Brain White Matter Short Association Bundles Using HARDI Data

Human brain connectivity is extremely complex and variable across subjects. While long association and projection bundles are stable and have been deeply studied, short association bundles present higher intersubject variability, and few studies have been carried out to adequately describe the structure, shape, and reproducibility of these bundles. However, their analysis is crucial to understand brain function and better characterize the human connectome. In this study, we propose an automatic method to identify reproducible short association bundles of the superficial white matter, based on intersubject hierarchical clustering. The method is applied to the whole brain and finds representative clusters of similar fibers belonging to a group of subjects, according to a distance metric between fibers. We experimented with both affine and non-linear registrations and, due to better reproducibility, chose the results obtained from non-linear registration. Once the clusters are calculated, our method performs automatic labeling of the most stable connections based on individual cortical parcellations. We compare results between two independent groups of subjects from a HARDI database to generate reproducible connections for the creation of an atlas. To perform a better validation of the results, we used a bagging strategy that uses pairs of groups of 27 subjects from a database of 74 subjects. The result is an atlas with 44 bundles in the left hemisphere and 49 in the right hemisphere, of which 33 bundles are found in both hemispheres. Finally, we use the atlas to automatically segment 78 new subjects from a different HARDI database and to analyze stability and lateralization results.

Automatic segmentation of short association bundles using a new multi-subject atlas of the left hemisphere fronto-parietal brain connections.

Human brain connection map is far from being complete. In particular the study of the superficial white matter (SWM) is an unachieved task. Its description is essential for the understanding of human brain function and the study of the pathogenesis associated to it. In this work we developed a method for the automatic creation of a SWM bundle multi-subject atlas. The atlas generation method is based on a cortical parcellation for the extraction of fibers connecting two different gyri. Then, an intra-subject fiber clustering is applied, in order to divide each bundle into sub-bundles with similar shape. After that, a two-step inter-subject fiber clustering is used in order to find the correspondence between the sub-bundles across the subjects, fuse similar clusters and discard the outliers. The method was applied to 40 subjects of a high quality HARDI database, focused on the left hemisphere fronto-parietal and insula brain regions. We obtained an atlas composed of 44 bundles connecting 22 pair of ROIs. Then the atlas was used to automatically segment 39 new subjects from the database.

Automatic clustering of short association white matter fibers from HARDI tractography datasets

In this paper we studied the short brain association fibers based on an inter-subject clustering over 20 subjects of a high quality HARDI database. Fibers from all the subjects were clustered together in Talairach space. Generic fascicles present in most of the subjects were then labeled using a cortical parcelation and several criterion for the selection of the final bundles and labels. For the left hemisphere, we obtained 52 representative fascicles, present in at least 17 of the 20 subjects, twelve of which were found in all the subjects. For the right hemisphere, we found 28 representative fascicles, eleven of which were found in all the subjects.

T145. ALTERATIONS IN SUPERFICIAL WHITE MATTER IN THE FRONTAL CORTEX IN SCHIZOPHRENIA: A DWI STUDY USING A NOVEL ATLAS

Abstract Background Alterations in brain connectivity are strongly implicated in the pathophysiology of schizophrenia (SZ). Very recently, evidence is mounting to suggest that changes in superficial white matter (SWM) U-shaped short range fibers are integral components of disease neuropathology, a theory that is supported by findings from postmortem studies and less often in vivo in patients with SZ. This diffusion weighted imaging (DWI) study aimed to investigate SWM microstructure in the frontal cortex in people with SZ. Methods Whole brain tractography was performed in 31 people with SZ and 54 healthy controls using BrainVISA and Connectomist 2.0 software. Segmentation and labelling of superficial white matter tracts were performed using a novel atlas characterizing 100 bundles. Principal Components Analysis (PCA) using a varimax orthogonal rotation was performed on mean generalised fractional anisotropy (gFA) of bundles located in the frontal cortex. Composites scores were computed for each subject, reflecting a linear combination of mean gFA values. Results PCA revealed three components explaining 19.7 %, 5.8 %, and 5.4 % of the total variance. The mean score of the second component was significantly lower in the people with SZ compared with controls (p = 0.01) and included 13 bundles connecting regions in the pars orbitalis, insula, pars triangularis, pars opercularis, orbitofrontal cortex, anterior cingulate, superior frontal cortex and middle frontal cortex. Discussion Our results support findings of reduced white matter integrity in the frontal cortex in people with SZ. Moreover, PCA may be helpful in identifying specific networks as the deficits do not appear to be widespread. Identifying patterns of superficial white matter dysconnectivity may be helpful in understanding the prominent symptoms and cognitive deficits and observed in SZ.

Short association bundle atlas based on inter-subject clustering from HARDI data

This paper is focused on the study of short brain association fibers. We present an automatic method to identify short bundles of the superficial white matter based on inter-subject hierarchical clustering. Our method finds clusters of similar fibers, belonging to the different subjects, according to a distance measure between fibers. First, the algorithm obtains representative bundles and subsequently we perform an automatic labeling based on the anatomy, of the most stable connections. The analysis was applied to two independent groups of 37 subjects. Results between the two groups were compared, in order to keep reproducible connections for the atlas creation. The method was applied using linear and non-linear registration, where the non-linear registration showed significantly better results. A final atlas with 35 bundles in the left hemisphere and 27 in the right hemisphere from the whole brain was obtained. Finally results were validated using the atlas to segment 26 new subjects from another HARDI database.

Creation of a whole brain short association bundle atlas using a hybrid approach

The Human brain connection map is far from being complete. In particular the study of the superficial white matter (SWM) is an unachieved task. Its description is essential for the understanding of human brain function and the study of pathogenesis triggered by abnormal connectivity. In this work we expanded a previously developed method for the automatic creation of a whole brain SWM bundle atlas. The method is based on a hybrid approach. First a cortical parcellation is used to extract fibers connecting two regions. Then an intra-and inter-subject hierarchical clustering are applied to find well-defined SWM bundles reproducible across subjects. In addition to the fronto-parietal and insula regions of the left hemisphere, the analysis was extended to the temporal and occipital lobes, including all their internal regions, for both hemispheres. Validation steps are performed in order to test the robustness of the method and the reproducibility of the obtained bundles. First the method was applied to two independent groups of subjects, in order to discard bundles without match across the two independent atlases. Then, the resulting intersection atlas was projected on a third independent group of subjects in order to filter out bundles without reproducible and reliable projection. The final multi-subject U-fiber atlas is composed of 100 bundles in total, 50 per hemisphere, from which 35 are common to both hemispheres. The atlas can be used in clinical studies for segmentation of the SWM bundles in new subjects, and measure DW values or complement functional data.