Elsevier

NeuroImage

Volume 39, Issue 1, 1 January 2008, Pages 119-126
NeuroImage

Combining fMRI and DTI: A framework for exploring the limits of fMRI-guided DTI fiber tracking and for verifying DTI-based fiber tractography results

https://doi.org/10.1016/j.neuroimage.2007.08.025Get rights and content

Abstract

A powerful, non-invasive technique for estimating and visualizing white matter tracts in the human brain in vivo is white matter fiber tractography that uses magnetic resonance diffusion tensor imaging. The success of this method depends strongly on the capability of the applied tracking algorithm and the quality of the underlying data set. However, DTI-based fiber tractography still lacks standardized validation. In the present work, a combined fMRI/DTI study was performed, both to develop a setup for verifying fiber tracking results using fMRI-derived functional connections and to explore the limitations of fMRI based DTI fiber tracking. Therefore, a minor fiber bundle that features several fiber crossings and intersections was examined: The striatum and its connections to the primary motor cortex were examined by using two approaches to derive the somatotopic organization of the striatum. First, an fMRI-based somatotopic map of the striatum was reconstructed, based on fMRI activations that were provoked by unilateral motor tasks. Second, fMRI-guided DTI fiber tracking was performed to generate DTI-based somatotopic maps, using a standard line propagation and an advanced fast marching algorithm. The results show that the fiber connections reconstructed by the advanced fast marching algorithm are in good agreement with known anatomy, and that the DTI-revealed somatotopy is similar to the fMRI somatotopy. Furthermore, the study illustrates that the combination of fMRI with DTI can supply additional information in order to choose reasonable seed regions for generating functionally relevant networks and to validate reconstructed fibers.

Introduction

Diffusion tensor imaging (DTI) (Basser et al., 1994, Pierpaoli et al., 1996) is a Magnetic Resonance Imaging (MRI) technique which allows us to measure the three dimensional (3D) Brownian motion of water molecules in the living human brain. In this fashion, the orientation of nerve fibers can be probed indirectly because water molecules diffuse preferentially parallel to the axonal fibers rather than perpendicular to them. The diffusion properties within each voxel are characterized by a symmetric diffusion tensor, whose principle eigenvector indicates the 3D orientation of the underlying fiber structure.

Fiber tracking allows for the virtual reconstruction of axonal networks within the brain on the basis of DTI data (for a review, see Mori and van Zijl, 2002). A variety of different tracking algorithms have been proposed; however, no standard procedure to quantify or to validate tracking results has been established so far. A key obstacle of fiber tractography is the uncertainty of the main diffusion direction. Due to the fact that the principal eigenvector does not correspond to the main fiber orientation in some instances, false trajectories may occur (Basser et al., 2000, Jones, 2003, Watts et al., 2003). In brain voxels with fiber crossing, branching, kissing, or merging, the tensor’s principle eigenvector aligns along the averaged diffusion direction, and therefore, may not serve directly as a basis for fiber reconstruction. Furthermore, the tensors (and consequently the reconstructed fiber trajectories) can be compromised by partial volume effects or noise (Lazar and Alexander, 2003, Lazar and Alexander, 2005, Tournier et al., 2002). If the tracking algorithm incorporates solely the principal eigenvector for propagation, erroneous fiber pathways may be reconstructed. These obstacles have been addressed by recent studies. Many groups have proposed sophisticated techniques to improve fiber tracking (Lazar et al., 2003, Staempfli et al., 2006, Weinstein et al., 1999, Westin et al., 2002, Zhang et al., 2004).

A crucial input parameter for fiber tracking is the definition of an appropriate seed region. If seed areas are slightly changed by only a few voxels, completely different fiber trajectories may be reconstructed. Also, large intersubject anatomical variations often exist, making it difficult to reliably define tracking seed areas based on anatomical landmarks. The complexity of the situation is increased when considering not only healthy volunteers but also patients suffering from space-occupying pathological processes (e.g., tumors). Different studies have attempted to solve this obstacle by combining DTI with fMRI. Thereby, start regions are defined in anatomical areas which are revealed by functional MRI (fMRI) activation patterns (Guye et al., 2003, Hendler et al., 2003, Watts et al., 2003). This combined approach may become clinically relevant for providing improved treatment planning information and patient outcome (Mori et al., 2002, Pamar et al., 2004, Staempfli et al., 2004).

In the present work, the main focus was to investigate a part of the motor system with a combination of DTI and fMRI. Known cortico-subcortical connections between the striatum and the primary motor cortex (PriMC) were examined. These connections have already been studied in primates in the early forties by Dusser et al. (Dusser de Barenne et al., 1942). They investigated projections from PriMC and sensory motor cortex to the putamen and the caudate nucleus. Kunzle et al. (Kunzle, 1975) performed an autoradiographic study and showed that the putamen is a major target of bilateral projections from the PriMC. Alexander et al. (Alexander and DeLong, 1985) stimulated single neuron cells within the basal ganglia which induced discrete movements of different body parts. A superio-inferior somatotopic gradient was found with face lying inferiorly, foot superiorly, and hand representation in between. Both studies concluded that the primate putamen is more directly involved in motor functions, whereas the caudate nucleus is involved in more complex behavioral functions.

In humans, two studies (Lehericy et al., 2004a, Lehericy et al., 2004b) used fiber tractography to provide evidence that cortico-striatal connections are organized in discrete circuits, as in a monkey brain. However, the respective functions of these loops are still a matter of debate. The principal striatal circuit that is involved in the control of motor functions connects the cerebral cortex with the basal ganglia in a feedback loop. Fibers emerging from the motor cortex run along the internal and external capsule, respectively, and finally project in a somatotopic fashion upon the sensorimotor striatal territory (Nieuwenhuys et al., 1988). This is an area mainly within the superio-lateral sector of the postcommissural portion of the putamen. Several groups performed fMRI studies using different motor paradigms to asses this in vivo somatotopy of the basal ganglia (Gerardin et al., 2003, Lehericy et al., 1998, Maillard et al., 2000, Scholz et al., 2000). All studies showed a similar superio-inferior gradient of the foot, hand, and face representation within the putamen.

The two main goals of the present study were: (a) to derive a framework for verifying the results of DTI fiber tracking algorithms, and (b) to explore the limitations of fMRI-guided DTI fiber tracking. Therefore, known cortico-subcortical connections were reconstructed and investigated by deriving the somatotopic gradient in the putamen, twice. First, fMRI-based somatotopic maps were generated, using unilateral motor tasks. Second, DTI somatotopic maps were calculated, based on connection probabilities between striatal voxels and fMRI activations in the PriMC using two different DTI tracking algorithms. The resulting somatotopic maps were then compared and analyzed.

Section snippets

Subjects

Twelve healthy, right-handed volunteers were scanned. Each subject gave informed written consent. Six of them were excluded after fMRI data analysis (see sub-section Sites of activation in the Results section). Of the resulting six volunteers two were female and four male (mean age 26.5 ± 3.8 years). To test the handedness, the Edinburgh handedness inventory (Oldfield, 1971) was used. The subjects achieved values between 73.3 and 100, with a mean of 88.0 ± 8.6.

Data acquisition

For data acquisition, a 3 T whole body

Sites of activation

In all subjects, activation in the left PriMC was elicited by all tasks. Furthermore, the fMRI analysis revealed activated clusters for at least two of the three motor tasks in 6 out of the 12 subjects within the putamen (five subjects for the foot and hand task, six subjects for the face task). Only the data of these six subjects were considered in further postprocessing steps.

Segmentation and localization of the striatum

In Fig. 1, a schematical depiction of the exact location of the left striatum, consisting of caudate nucleus and

Discussion

DTI fiber tracking allows for the reconstruction of various trajectories. However, in most cases, there is no implicit quantification of the resulting fibers. Interpretation of the DTI data without a priori anatomical knowledge is hardly possible. Until today, no gold standard has been established for the verification of DTI-generated connection probabilities or DTI fiber tracks. In this study, the DTI data were complemented with information that was extracted from fMRI data for two reasons:

Acknowledgments

The authors are grateful for the continuing support of Philips Medical Systems and for the financial support given by the Strategic Excellence Project Program (SEP) of the ETH Zurich.

References (52)

  • V.H. Scholz et al.

    Laterality, somatotopy and reproducibility of the basal ganglia and motor cortex during motor tasks

    Brain Res.

    (2000)
  • P. Staempfli et al.

    Resolving fiber crossing using advanced fast marching tractography based on diffusion tensor imaging

    NeuroImage

    (2006)
  • D.S. Tuch et al.

    Diffusion MRI of complex neural architecture

    Neuron

    (2003)
  • C.F. Westin et al.

    Processing and visualization for diffusion tensor MRI

    Med. Image Anal.

    (2002)
  • G.E. Alexander et al.

    Microstimulation of the primate neostriatum. II. Somatotopic organization of striatal microexcitable zones and their relation to neuronal response properties

    J. Neurophysiol.

    (1985)
  • R. Bammer et al.

    Diffusion tensor imaging using single-shot SENSE-EPI

    Magn. Reson. Med.

    (2002)
  • P.J. Basser et al.

    In vivo fiber tractography using DT-MRI data

    Magn. Reson. Med.

    (2000)
  • P.G. Batchelor et al.

    Anisotropic noise propagation in diffusion tensor MRI sampling schemes

    Magn. Reson. Med.

    (2003)
  • J. Dusser de Barenne et al.

    Physiological neuronography of the cortico-striatal connections

    Res. Publ.-Assoc. Nerv. Ment. Dis.

    (1942)
  • A.W. Flaherty et al.

    Corticostriatal transformations in the primate somatosensory system. Projections from physiologically mapped body-part representations

    J. Neurophysiol.

    (1991)
  • A.W. Flaherty et al.

    Two input systems for body representations in the primate striatal matrix: experimental evidence in the squirrel monkey

    J. Neurosci.

    (1993)
  • L.R. Frank

    Anisotropy in high angular resolution diffusion-weighted MRI

    Magn. Reson. Med.

    (2001)
  • L.R. Frank

    Characterization of anisotropy in high angular resolution diffusion-weighted MRI

    Magn. Reson. Med.

    (2002)
  • E. Gerardin et al.

    Foot, hand, face and eye representation in the human striatum

    Cereb. Cortex

    (2003)
  • T. Hendler et al.

    “Two are Better than One”: Combining fMRI and DTI Based Fiber Tracking for Effective Pre-Surgical Mapping

  • T. Jaermann et al.

    SENSE-DTI at 3 T

    Magn. Reson. Med.

    (2004)
  • Cited by (82)

    • Pictorial essay: How co-registered BOLD fMRI and DTI data can improve diffusion tensor tractography

      2021, Interdisciplinary Neurosurgery: Advanced Techniques and Case Management
      Citation Excerpt :

      Blood oxygen level dependent (BOLD) fMRI can delineate the relationship of eloquent cortex to surgical targets, allowing a surgical plan that minimizes damage to areas involved in vision, language, or motor function [10,11]. Combining BOLD fMRI with DTI can reduce artificial truncation of white matter tracts [4,12–15]. This paper illustrates how coregistration of BOLD fMRI and DTI data can improve the pre-operative imaging assessment of brain tumor patients by providing more comprehensive tractograms of the pyramidal tract (PT) and arcuate fasciculus (AF).

    • Diffusion MRI-based connectivity

      2021, Connectomic Deep Brain Stimulation
    View all citing articles on Scopus
    1

    First and second author both contributed equally to this study.

    View full text