Elsevier

NeuroImage

Volume 27, Issue 4, 1 October 2005, Pages 842-851
NeuroImage

Spatial registration of multichannel multi-subject fNIRS data to MNI space without MRI

https://doi.org/10.1016/j.neuroimage.2005.05.019Get rights and content

Abstract

The registration of functional brain data to the common brain space offers great advantages for inter-modal data integration and sharing. However, this is difficult to achieve in functional near-infrared spectroscopy (fNIRS) because fNIRS data are primary obtained from the head surface and lack structural information of the measured brain. Therefore, in our previous articles, we presented a method for probabilistic registration of fNIRS data to the standard Montreal Neurological Institute (MNI) template through international 10–20 system without using the subject's magnetic resonance image (MRI). In the current study, we demonstrate our method with a new statistical model to facilitate group studies and provide information on different components of variability. We adopt an analysis similar to the single-factor one-way classification analysis of variance based on random effects model to examine the variability involved in our improvised method of probabilistic registration of fNIRS data. We tested this method by registering head surface data of twelve subjects to seventeen reference MRI data sets and found that the standard deviation in probabilistic registration thus performed for given head surface points is approximately within the range of 4.7 to 7.0 mm. This means that, if the spatial registration error is within an acceptable tolerance limit, it is possible to perform multi-subject fNIRS analysis to make inference at the population level and to provide information on positional variability in the population, even when subjects' MRIs are not available. In essence, the current method enables the multi-subject fNIRS data to be presented in the MNI space with clear description of associated positional variability. Such data presentation on a common platform, will not only strengthen the validity of the population analysis of fNIRS studies, but will also facilitate both intra- and inter-modal data sharing among the neuroimaging community.

Introduction

With the advent and development of various brain mapping techniques, there is a growing demand to access and compare data acquired from different modalities. Ideally, all the functional neuroimaging data should be presented on, or at least can be translated to, a common platform across all modalities.

Currently, the Montreal Neurological Institute (MNI) standard coordinate system serves as the most common stereotaxic platform (reviewed in Brett et al., 2002). The MNI standard template was created by registering magnetic resonance images (MRIs) of many subjects to the Talairach brain and their subsequent averaging (Collins et al., 1994). In conventional mapping techniques, first, the structural brain imaging data are registered to the standard MNI brain template by linear and nonlinear transformation processes to create the normalized brain image. Then, the functional data from neuroimaging experiments are registered to the normalized brain image, thereby to MNI coordinate system. A common stereotaxic platform, such as MNI system, facilitates a relatively consistent spatial expression of functional localization data. Thus, presenting functional neuroimaging data in the MNI standard coordinate system has virtually grown to be the golden standard for tomographic functional brain mapping methods, such as functional magnetic resonance imaging (fMRI) and positron emission tomography (PET). Furthermore, MNI coordinate system is compatible to another common stereotaxic system, Talairach coordinate system that is based on a single patient's brain with detailed expression of anatomical features including Brodmann estimates (Brett et al., 2002, Talairach and Tournoux, 1988).

Although the MNI coordinate system has a potential to become the golden standard for inter-modal data presentation, it poses a major obstacle in assimilating transcranial mapping data. The transcranial mapping data are primarily obtained from the head surface without any structural information of the underlying brain. In order to register such data in the MNI coordinate system, it is necessary to map it from the head surface to the cortical region that corresponds to the signal source. For conventional transcranial methods such as magnetoencephalography (MEG) and electroencephalography (EEG), researchers have developed three-dimensional signal source estimation algorithms, including LORETA for EEG (Pascual-Marqui et al., 2002), SAM, linear beamforming, and MUSIC for MEG (reviewed in Vrba and Robinson, 2001). These methods enabled researchers to present transcranial functional data in the standard MNI and Talairach brains.

Unlike conventional transcranial neuroimaging techniques, functional data from functional near-infrared spectroscopy (fNIRS) and transcranial magnetic stimulation (TMS) describe cortical activity only on the lateral cortical surface. fNIRS utilizes the tight coupling between neural activity and regional cerebral blood flow and monitors regional cerebral blood volume changes as relative changes of hemoglobin concentration (Colacino et al., 1981, Hoshi and Tamura, 1993, Jobsis, 1977, Kato et al., 1993, Villringer et al., 1993). fNIRS requires only compact experimental systems, is less restrictive, and is relatively more robust to body movement. Owing to these features, fNIRS provides researchers with a wide variety of flexible experimental setups for long-term and repetitive monitoring. The near-infrared light emitted from a light source is detected by a detector located several centimeters away from the source on the head surface after traveling laterally through cortical tissue. Therefore, the signal source is estimated only on the lateral cortical surface (Firbank et al., 1998, Strangman et al., 2002, Villringer and Chance, 1997). In TMS, the applied magnetic field is effective only within 4 cm of the head surface (reviewed in Jalinous, 1991), and the estimation of the target point is limited to the lateral cortical surface. Therefore, fNIRS and TMS require specialized methods to estimate the signal source on the lateral cortical surface and to present their data onto the standard MNI and Talairach systems. Such methods have evolved but are only at the beginning stage (Okamoto and Dan, 2005, Okamoto et al., 2004a).

Our recent publications presented such specialized methods for fNIRS and TMS in order to provide a probabilistic and topological link between the head surface where measurements are obtained and the brain surface where signal or target exists. In fNIRS and TMS studies, head surface locations are often described via the international 10–20 positions which are determined in reference to the head surface landmarks (Jasper, 1958). Therefore, we established correspondence between the international 10–20 and the MNI coordinate systems for nineteen standard 10–20 positions and clarified the error factors associated with such cross-platform correspondence (Okamoto et al., 2004a). We developed transcranial projection algorithms and computer programs to project a given head surface point onto the cortical surface in MNI space for the purpose of functional brain mapping (Okamoto and Dan, 2005). Like other transcranial techniques, fNIRS and TMS have required MRIs to obtain structural information of the brain that is examined. However, since acquisition of MRIs is expensive, there are often situations when structural images of subjects are not available. The exclusion of fNIRS and TMS data because of non-availability of structural images will result in a tremendous loss to neuroimaging studies. Therefore, we extended our projection method to the situation when structural images of the subject are not available. This method probabilistically estimates the cortical projection point for a given head surface point in reference to the established transcranial correspondence of 10–20 standard points (Okamoto and Dan, 2005).

Although our previous article mainly emphasized the theoretical aspects of the probabilistic link between the MNI coordinate and the international 10–20 systems, it did not include details on the practical aspects and application of the method. Furthermore, the former method only dealt with a single-subject study. However, in functional studies, the use of multi-subject analysis is inevitable for making the inference at the population level.

In this study, first we will introduce a practical way of determining a given head surface point in the real-world (RW) coordinate system using a 3D digitizer. Then, we will demonstrate the probabilistic registration of multichannel fNIRS data to the MNI template for multiple subjects via the 10–20 standard positions. We regard multichannel fNIRS as the most generalized case. Furthermore, we will examine the error factor arising from different sources of variability in the data and method using a random effects statistical model.

In essence, the current method enables the presentation of transcranial mapping data in MNI coordinate system without any structural image, thereby establishing a close topological link between transcranial and tomographic brain-mapping modalities.

Section snippets

Subjects

Twelve healthy volunteers participated in the head surface measurement (two males and ten females, aged 24 to 39 years). All were right-handed. Written informed consents and approval from the institutional ethics committee were obtained.

Head surface measurement

We used a 3D magnetic space digitizer, FASTRAK (Polhemus, Colchester, VT), to obtain relative locations of 10–20 standard positions and fNIRS optodes in real-world (RW) coordinate system. For this measurement, we devised a relatively comfortable and reproducible

Results and discussion

We used the proposed probabilistic registration method to estimate the CP points corresponding to all channel/optode locations based on 10–20 system in prefrontal region. We estimated the standard deviations using one-way ANOVA for all channel/optode locations. Table 3 presents a list of the estimated coordinates and standard deviations of CP points for all the channels and optodes. The cortical-surface centroid is just a slight variation of mean, and therefore the similar approximated

Acknowledgments

We thank Professor Tadashi Takemori at Tsukuba University for helpful discussions. This work is supported by the Industrial Technology Research Grant Program in 03A47022 from the New Energy and Industrial Technology Development Organization (NEDO) of Japan and the Program for Promotion of Basic Research Activities for Innovative Bioscience (PROBRAIN).

References (40)

  • A. Villringer et al.

    Non-invasive optical spectroscopy and imaging of human brain function

    Trends. Neurosci.

    (1997)
  • A. Villringer et al.

    Near infrared spectroscopy (NIRS): a new tool to study hemodynamic changes during activation of brain function in human adults

    Neurosci. Lett.

    (1993)
  • J. Vrba et al.

    Signal processing in magnetoencephalography

    Methods

    (2001)
  • E. Watanabe et al.

    Non-invasive functional mapping with multi-channel near infra-red spectroscopic topography in humans

    Neurosci. Lett.

    (1996)
  • A. Abbott

    Neuroscience: a new atlas of the brain

    Nature

    (2003)
  • J. Ashburner et al.

    Nonlinear spatial normalization using basis functions

    Hum. Brain Mapp.

    (1999)
  • J. Ashburner et al.

    Image registration using a symmetric prior—In three dimensions

    Hum. Brain Mapp.

    (2000)
  • M. Brett et al.

    The problem of functional localization in the human brain

    Nat. Rev. Neurosci.

    (2002)
  • J.M. Colacino et al.

    Infra-red technique for cerebral blood flow: comparison with 133Xenon clearance

    Neurol. Res.

    (1981)
  • D.L. Collins et al.

    Automatic 3D intersubject registration of MR volumetric data in standardized Talairach space

    J. Comput. Assist. Tomogr.

    (1994)
  • Cited by (526)

    View all citing articles on Scopus
    1

    The two authors contributed equally to this work.

    View full text