Multimodal fMRI tractography in normal subjects and in clinically recovered traumatic brain injury patients Andrea Cherubini, ⁎ Giacomo Luccichenti, Patrice Péran, Gisela E. Hagberg, Carmen Barba, Rita Formisano, and Umberto Sabatini Neuroimaging Laboratory, IRCCS Fondazione Santa Lucia, Via Ardeatina 306, 00179 Rome, Italy Received 5 May 2006; revised 13 September 2006; accepted 1 November 2006 Available online 2 January 2007 In this study, we defined an operator-independent protocol for reconstructing the anatomical connections originating from fMRI activations in order to demonstrate that results obtained with this protocol are affected by alterations of functional activations. Seven healthy volunteers and two patients who sustained traumatic brain injury underwent an fMRI with a finger tapping task and a DTI scan. Cortical fMRI activations were used directly as seed mask for tractography for the reconstruction of individual motor pathways. On patients we observed a different motor network if compared to healthy subjects. However, when the activations of healthy subjects were used as seed masks for the tractography in patients, we observed for the patients a pattern of connectivity more similar to what was observed for healthy subjects. At the same time, when the activations of patients were used for the tractography on healthy subjects, we obtained patterns of connectivity similar to those obtained for patients. These results show the potential of the integration of fMRI and tractography for clarifying the mechanisms of cortical plasticity in the recovery of motor functions. © 2006 Elsevier Inc. All rights reserved. Keywords: Probabilistic tractography; Connectivity; Functional MRI; Motor pathways; Motor-deficit; Cortical plasticity Introduction The past few years have seen an exponential growth of fiber tracking (FT) studies, aimed at reconstructing in vivo the anatomical connections within the human brain using diffusion- weighted magnetic resonance imaging (DWI) (Basser et al., 2000; Behrens et al., 2003b; Conturo et al., 1999; Jones et al., 1999; Mori and van Zijl, 2002; Parker et al., 2002). In these studies, the pathways of nervous fibers are inferred from DWI data, under the assumption that water diffusion is influenced by the spatial anisotropy of local microstructure in myelinated fiber bundles (Beaulieu, 2002). Although the application of FT to the investigation of connectivity is very promising, it is currently difficult to interpret the results of these studies in functional terms. This difficulty originates from the fact that conventional tractography is an anatomy-based technique, whose results depend on the prior knowledge of functional neuroanatomy. In fact, in tractography studies, a seed point is usually arbitrarily defined as a region corresponding to known functional areas from where to start the reconstruction from (Mori and van Zijl, 2002). However, the location of functional areas is known with an inter-individual variability, which may be consistently prominent in patients with psychiatric, neurological primary or secondary disease. In other words, when the seed area for tractography is manually defined by an operator, FT algorithms could possibly generate trajectories that are uncorrelated to the actual functional area of the individual. Functional magnetic resonance imaging (fMRI), in virtue of its ability to locate functional areas as cortical activations that are due to signal changes occurring during specific tasks or stimuli, could be integrated with FT studies in order to overcome these limitations. Nevertheless, conventional “streamline” FT algorithms (sFT) make this integration difficult because activated areas, which could represent the seed mask where to start the fiber reconstruc- tion from, are mainly located within gray matter, which present a degree of diffusion anisotropy inadequate for sFT. Moreover, since sFT provide an information of “binary” nature (fibers are either traced or not), results obtained from these algorithms are particularly sensitive to noise in the diffusion dataset (Behrens et al., 2003b; Ciccarelli et al., 2003a; Lori et al., 2002). Recently, it has been proposed to reconstruct white matter tracts using probabilistic fiber tracking (pFT) methods, which take into account the local uncertainty in fiber orientation and are able to provide information of “parametric” nature about the level of anatomical connectivity (Behrens et al., 2003a,b; Guye et al., 2003; Koch et al., 2002; Lazar et al., 2003; Parker and Alexander, 2003; Parker et al., 2002). Thanks to their peculiar characteristics, these probabilistic algorithms are able to track fibers robustly through areas of low diffusion anisotropy, where conventional sFT methods www.elsevier.com/locate/ynimg NeuroImage 34 (2007) 1331 – 1341 ⁎ Corresponding author. Fax: +39 06 5150 1213. E-mail address: a.cherubini@hsantalucia.it (A. Cherubini). Available online on ScienceDirect (www.sciencedirect.com). 1053-8119/$ - see front matter © 2006 Elsevier Inc. All rights reserved. doi:10.1016/j.neuroimage.2006.11.024