Registration of Brain Atlas to MR Images Using Topology Preserving Front Propagation

Abstract

Registration of brain atlases to MR images is important in both anatomic and functional studies of human brains. Existing intensity-based methods are confronted with the translation of image-similarity functions to desired anatomic correspondences; while feature-based methods are challenged with the automated extraction of required features. In this paper, we propose a non-rigid registration method, in which, a block matching method is first used to calculate boundary displacement of all structures in a brain atlas, and a topology preserving front propagation method is then used to deform the atlas by warping the structures according to their boundary displacements. The novelty of our method is that the registration procedure is automated and anatomically driven while there is no need to extract particular structures. Experiments on the registration of the Talairach-Tournoux brain atlas to phantom brain MR images and real data show that our method is robust to noise and intensity inhomogeneity, more accurate than the commonly used Talairach stereotaxic spatial normalization, and thus promising to open new applications for brain atlases.

Human Brain Atlases in Stroke Management

“…We have developed several methods for mapping of atlases to scans, including (Xu and Nowinski 2001 ; Liu et al 2009 ; Nowinski et al 2006b ; Volkau et al 2012 ). The initial development was guided by applications in stereotactic and functional neurosurgery (Nowinski 2009 ).…”

“…The initial development was guided by applications in stereotactic and functional neurosurgery (Nowinski 2009 ). Although we have devised non-rigid methods for mapping of the employed here atlas of anatomy to scans, such as (Xu and Nowinski 2001 ; Liu et al 2009 including Patent 8), we decided to use in stroke applications very fast and conceptually simple methods that are also potentially understandable to clinicians, including the FTT (Nowinski et al 2006b ) and the ellipse-based fitting method (Volkau et al 2012 ). The FTT requires 5 s only but it was devised for MR scans.…”

A Tutorial on Well-Composedness

“…For example, it can be made of one component at a given time and of two components at another time [118] (see Figure 7). Furthermore, in this way we consider only each label relatively to its associated background; the topological relationship between the different labels is then lost [142,11,111]. Another approach is to consider that the multi-label image can be represented with a connected components tree [114] where the root is the infinite background, and the rest of the tree represent the nested relationship between the connected components into the image.…”

Topology on Digital Label Images

“…The most common approach is to consider only one label at a time, the other labels being momentarily considered as a part of the background. However, except in the most simple cases where the label configuration leads to a binary modelling (see, e.g., [8,9]), one cannot directly deal with the relations between the labels but only with the topology of each label and of its associated background [10,11,12] (if necessary, one uses in addition an adjacency tree between labels in order to control their topological relations). These methods are often used with a cost function, which depends on the applicative context, whose *Manuscript Click here to download Manuscript: gestionLabels_en_href.tex Click here to view linked References 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 Fig.…”