Jean-François Mangin

Robust multimodality registration for brain mapping

We present a robust intrasubject registration method for the synergistic use of multiple neuroimaging modalities, with applications to magnetic resonance imaging (MRI), functional MRI, perfusion MRI, MR spectroscopy, and single‐photon emission computed tomography (SPECT). This method allows user‐friendly processing of difficult examinations (low spatial resolution, advanced pathology, motion during acquisition, and large areas of focal activation). Registration of three‐dimensional (3D) brain scans is initially estimated by first‐order moment matching, followed by iterative anisotrophic chamfer matching of brain surfaces. Automatic brain surface extraction is performed in all imaging modalities. A new generalized distance definition and new specific methodologies allow registration of scans that cover only a limited range of brain surface. A new semiautomated supervision scheme allows fast and intuitive corrections of possible false automatic registration results. The accuracy of the MRI/SPECT anatomical‐functional correspondence obtained was evaluated using simulations and two difficult clinical populations (tumors and degenerative brain disorders). The average discrimination capability of SPECT (12.4 mm in‐plane resolution, 20 mm slice thickness) was found to be better than 5 mm after registration with MRI (5 mm slice thickness). Registration accuracy was always better than imaging resolution. Complete 3D MRI and SPECT registration time ranged between 6–11 min, in which surface matching represented 2–3 min. No registration failure occurred. In conclusion, the application of several new image processing techniques allowed efficient and robust registration. Hum. Brain Mapping 5:3–17, 1997.

Segmenting internal structures in 3D MR images of the brain by Markovian relaxation on a watershed based adjacency graph

The authors present a fast stochastic method aiming at segmenting cerebral internal structures in 3D magnetic resonance images. An original method introducing context permits the authors to obtain reliable radiometric characteristics even for hardly discriminable brain structures. Segmentation is formulated as the labeling of a region adjacency graph. The graph is constructed by an extension to 3D of the watershed algorithm and the labeling is performed using a Markovian relaxation process. This leads to consistent results with a very low computational burden.

Differential features of cortical folds

Analysis of functional images of the brain increasingly relies on individual anatomy of subjects. The aim is the construction of accurate anatomy-indexed functional mappings which would be of crucial importance for neurosurgical planning. The major difficulty relies in the important inter-subject structural variability of the cortical anatomy. In this paper, we assume that a large part of this variability can be overcome if more elementary and stable units than sulci and gyri are chosen to analyse cortex anatomy. Hence, we try to highlight such features using differential geometry. A method extracting cortical fold crest lines is described first. Then a morphological decomposition of sulci along crest lines is proposed. Finally, the relevance of this approach is demonstrated by a segmentation of central sulcus into two stable parts for ten subjects.

How interindividual differences in brain anatomy shape reading accuracy

The capacity to read develops throughout intensive academic learning and training. Several studies have investigated the impact of reading on the brain, and particularly how the anatomy of the brain changes with reading acquisition. In the present study, we investigated the converse issue, namely whether and how reading acquisition is constrained by the anatomy of the brain. Using multimodal MRI, we found that (a) the pattern (continuous or interrupted sulcus) of the posterior part of the left lateral occipito-temporal sulcus (OTS) hosting the visual word form area (VWFA) predicts reading skills in adults; that (b) this effect is modulated by the age of reading acquisition; and that (c) the length of the OTS sulcal interruption is associated with reading skills. Because the sulcal pattern is determined in utero, our findings suggest that individual difference in reading skills can be traced back to early stages of brain development in addition to the well-established socioeconomic and educational factors.

Sulci as Landmarks

When spatially normalizing images of the human cerebral cortex, the folding pattern is often used as a proxy for architecture. However, the variability of the folding pattern across individuals creates a lot of difficulties. Furthermore, the mechanisms underlying gyrification are still under exploration, and the links between folding and architecture are unclear outside primary areas. New computational methods focused on cortical sulci have been designed to support the research programs aiming at improving the role of sulci as architectural landmarks. They provide automatic recognition of the sulci, sulcal-based spatial normalization, and models of the variability of the shape of the sulci.

Sulcus Identification and Labeling

The complexity and the variability of the cortical folding pattern are overwhelming for human experts. Computational anatomy helps the field to harness the folding variability considered as a proxy for architectural variability. First, bottom-up processing pipelines convert the implicit encoding of the cortical folding pattern embedded in the geometry of the cortical surface into a synthetic graphic representation. Then, learning-based pattern recognition methods assemble the building blocks of the folding making up this representation in order to reconstruct the sulci of the standard nomenclature. Some attempts at improving current folding models using the same bottom-up strategy could have some impact in the near future.

A Patch-Based Segmentation Approach with High Level Representation of the Data for Cortical Sulci Recognition.

Because of the strong variability of the cortical sulci, their automatic recognition is still a challenging problem. The last algorithm developed in our laboratory for 125 sulci reaches an average recognition rate around 86%. It has been applied to thousands of brains for morphometric studies (www.brainvisa.info). A weak point of this approach is the modeling of the training dataset as a single template of sulcus-wise probability maps, losing information about the alternative patterns of each sulcus. To overcome this limit, we propose a different strategy inspired by Multi-Atlas Segmentation (MAS) and more particularly the patch-based approaches. As the standard way of extracting patches does not seem capable of exploiting the sulci geometry and the relations between them, which we believe to be the discriminative features for recognition, we propose a new patch generation strategy based on a high level representation of the sulci. We show that our new approach is slightly, but significantly, better than the reference one, while we still have an avenue of potential refinements that were beyond reach for a single template strategy.

Group Analysis of Individual Activation Maps Using 3D Scale-Space Primal Sketches and a Markovian Random Field

We present here a new method of cerebral activation detection. This method is applied on individual activation maps of any sort. It aims at processing a group analysis while preserving individual information and at overcoming as far as possible problems induced by spatial normalization used to compare different subject. The analysis is made through a multi-scale object-based description of the individual maps and these descriptions are compared, rather than comparing directly the images in a stereotactic space. The comparison is made using a graph, on which a labeling process is performed. The label field on the graph is modeled by a Markovian random field, which allows us to introduce high-level rules of interrogation of the data.