Elsevier

NeuroImage

Volume 86, 1 February 2014, Pages 404-416
NeuroImage

Automated longitudinal intra-subject analysis (ALISA) for diffusion MRI tractography

https://doi.org/10.1016/j.neuroimage.2013.10.026Get rights and content

Highlights

  • A framework for longitudinal fiber tractography (FT) analyses (ALISA) is presented.

  • ALISA performs time-efficient automated longitudinal intra-subject analyses.

  • The ALISA approach performs equally well as manually segmented FT results.

  • Precision and accuracy of ALISA are comparable to the reference standard.

  • Inter-subject automated analyses do not perform as well as ALISA or manual analyses.

Abstract

Fiber tractography (FT), which aims to reconstruct the three-dimensional trajectories of white matter (WM) fibers non-invasively, is one of the most popular approaches for analyzing diffusion tensor imaging (DTI) data given its high inter- and intra-rater reliability and scan-rescan reproducibility. The major disadvantage of manual FT segmentations, unfortunately, is that placing regions-of-interest for tract selection can be very labor-intensive and time-consuming. Although there are several methods that can identify specific WM fiber bundles in an automated way, manual FT segmentations across multiple subjects performed by a trained rater with neuroanatomical expertise are generally assumed to be more accurate. However, for longitudinal DTI analyses it may still be beneficial to automate the FT segmentation across multiple time points, but then for each individual subject separately. Both the inter-subject and intra-subject automation in this situation are intended for subjects without gross pathology. In this work, we propose such an automated longitudinal intra-subject analysis (dubbed ALISA) approach, and assessed whether ALISA could preserve the same level of reliability as obtained with manual FT segmentations. In addition, we compared ALISA with an automated inter-subject analysis. Based on DTI data sets from (i) ten healthy subjects that were scanned five times (six-month intervals, aged 7.6–8.6 years at the first scan) and (ii) one control subject that was scanned ten times (weekly intervals, 12.2 years at the first scan), we demonstrate that the increased efficiency provided by ALISA does not compromise the high degrees of precision and accuracy that can be achieved with manual FT segmentations. Further automation for inter-subject analyses, however, did not provide similarly accurate FT segmentations.

Introduction

To date, diffusion magnetic resonance imaging (MRI) is the only non-invasive method for probing soft tissue microstructure and its 3D architectural organization in vivo, offering the possibility of exploring the microstructural organization and architectural configuration of distinct anatomical fiber networks within the brain white matter (WM) (Jones, 2008, Tournier et al., 2011). Diffusion tensor imaging (DTI) (Basser et al., 1994), in particular, has been widely used to investigate e.g., WM abnormalities in pathological conditions (Caeyenberghs et al., 2010, Ciccarelli et al., 2006, Concha et al., 2005a, Concha et al., 2005b, Concha et al., 2009, Deprez et al., 2011, Jones et al., 2006, Price et al., 2008, Sage et al., 2009, Van Hecke et al., 2010b, Yogarajah et al., 2008), WM changes in normal development (Eluvathingal et al., 2007, Lebel et al., 2008, Lebel et al., 2010, Verhoeven et al., 2010, Zhang et al., 2007), and aging (Hsu et al., 2008, Hsu et al., 2010, Sullivan and Pfefferbaum, 2007, Van Hecke et al., 2008a). In many of these studies, fiber tractography (FT) (Basser et al., 2000, Conturo et al., 1999, Jones et al., 1999a, Koch et al., 2001, Mori et al., 1999, Parker et al., 2002, Parker et al., 2003, Poupon et al., 2000) has been used to identify specific WM fiber bundles, from which diffusion characteristics, such as fractional anisotropy (FA) and mean diffusivity (MD), can be derived (Jones et al., 2005a). Other approaches for investigating DTI data include region-of-interest (ROI) (e.g., Madsen et al., 2011, Snook et al., 2005), histogram (e.g., Cercignani et al., 2001), voxel-based (e.g., Giorgio et al., 2010) including TBSS (Smith et al., 2006), network-based (e.g., Hagmann et al., 2008, Reijmer et al., 2013) and atlas-based (e.g., Faria et al., 2011) analyses. A detailed description of these techniques is considered beyond the scope of this article — the interested reader is referred to surveys by Cercignani (2010) and Hasan et al. (2011).

With the ability to observe diffusion changes over time in the same population of subjects, longitudinally designed DTI studies can provide more specific insights into the microstructural dynamics of brain WM tissue compared to studies with a cross-sectional population setup (Beaulieu, 2002, Johansen-Berg, 2010, Lebel and Beaulieu, 2011). Boosted by the advent of more stable and high performance MR equipment, the interest to perform longitudinal DTI studies for capturing such subject-specific changes in microstructural organization is increasing rapidly (e.g., Concha et al., 2007, Deprez et al., 2012, Gong et al., 2008, Keller and Just, 2009, Kumar et al., 2009, Ljungqvist et al., 2011, Schlaug et al., 2009, Scholz et al., 2009, Sullivan et al., 2010, Yogarajah et al., 2010). Although there are several well-established methods for analyzing cross-sectional DTI data sets, they may not be ‘optimal’ for longitudinal studies. In voxel-based DTI analyses, for instance, which are notorious for their high sensitivity with respect to the amount/type of filtering (Jones et al., 2005b, Van Hecke et al., 2009, Van Hecke et al., 2010a) and choice of template/atlas (Sage et al., 2009, Van Hecke et al., 2008b, Van Hecke et al., 2011), subtle intra-subject changes may not be detected due to the much larger residual inter-subject misalignments. In this case, and if there is also a clear hypothesis regarding a specific WM fiber bundle (or a segment thereof — Colby et al., 2012), FT may be preferred over the voxel-based approach.

By combining objective protocols for extracting WM fiber pathways of interest on the one hand (Catani and Thiebaut de Schotten, 2008, Wakana et al., 2004, Wakana et al., 2007) and incorporating subjective prior knowledge from the neuroanatomical expert on the other hand, many studies have already demonstrated the high inter- and intra-rater reliability and scan–rescan reproducibility of manual FT segmentations (Ciccarelli et al., 2003, Danielian et al., 2010, Heiervang et al., 2006, Kristo et al., 2013a, Kristo et al., 2013b, Malykhin et al., 2008, Pfefferbaum et al., 2003, Wakana et al., 2007). The major drawback of manual FT segmentations, however, is that placing ROIs for tract selection can be very labor-intensive and time-consuming, which – for obvious reasons – can become problematic for large-cohort studies. Notwithstanding the existence of methods that can identify WM fiber bundles in an automated way in the absence of gross pathology (Clayden et al., 2007, Hagler et al., 2009, Lebel et al., 2008, Leemans et al., 2006, O'Donnell et al., 2009, Reich et al., 2010, Suarez et al., 2012, Verhoeven et al., 2010, Yendiki et al., 2011, Zhang et al., 2010), manual FT segmentations across multiple subjects performed by a trained rater with neuroanatomical expertise are generally more specific and, therefore, more accurate. For longitudinal DTI analyses, however, it may still be beneficial to automate the FT segmentation across multiple time points, but then for each individual subject separately. In doing so, the adverse effect of inter-subject variability on the reliability of the FT results may be circumvented, while maintaining the main advantages of the automated approach, i.e., higher efficiency and objectivity.

In this work, we developed an automated longitudinal intra-subject analysis (ALISA) for investigating FT segmentations and compared its performance in terms of precision and accuracy to the “bronze standard”, i.e. the manual FT segmentations. In addition, we compared these results to those obtained with FT segmentations obtained in an automated way over all time points and all subjects (Lebel et al., 2008). Sixty DTI data sets, which are part of the HUBU cohort database (“Hjernens Udvikling hos Børn og Unge”: Brain maturation in children and adolescents; see Madsen et al., 2010, Madsen et al., 2011 for more details), were included in this study: five acquisitions at six-month intervals for ten healthy subjects and a set of ten repeats of one control subject scanned at weekly intervals. Important to note here is that these automated methods are not intended for case-based clinical use in cases with gross pathology, but rather for the use of longitudinal group studies of healthy development and aging, and pathologies without large displacing lesions.

Without loss of generality, the analyses were evaluated with tractography results from four WM fiber bundles: (i) the superior segment of the cingulum (SSCing) bundle, part of the collection of WM fibers that interconnect limbic structures, relevant for the regulation of emotional processes (e.g., Karaus et al., 2009); (ii) the cortico-spinal tracts (CST), running from the spinal cord to the motor cortex; (iii) the uncinate fasciculus (UF), connecting the frontal and temporal lobes, which has been shown to be important in the interaction between cognition and emotion (e.g., Price et al., 2008); and (iv) the forceps major (FM), or splenium of the corpus callosum, providing interhemispheric occipital connections that are affected in, for instance, schizophrenia (Catani and Thiebaut de Schotten, 2008, Clark et al., 2011).

Section snippets

Subjects and data acquisition

Sixty DTI data sets were acquired for this study, five acquisitions at six-month intervals for ten healthy subjects (8 F/2 M) aged 7.6 to 8.6 years (mean age of 8.1 ± 0.4 years) at the first acquisition date and a set of ten repeats of one control subject (female, age 12.2 years) scanned on four occasions, with respectively two, three, two, and three separate scan sessions. These four occasions were separated by two, two, and seven weeks, respectively. The data sets are part of the HUBU cohort

Subject-specific template construction

The reliability of automating fiber tractography segmentations for serial DTI data depends heavily on the coregistration quality and, consequently, the construction of the subject-specific template that was used in the ALISA approach. Indicating the quality of the applied normalization procedure (Jones et al., 2002), Fig. 3 shows the resulting template derived from the five serial DTI scans of a representative subject. Overlaid on the FA skeleton of this template, the dominant diffusion

Discussion

With the advent of large-cohort DTI studies, there is an increased demand for data analyses that require minimal user input. While several methods have been developed to analyze WM fiber bundles in an automated way (Clayden et al., 2007, Hagler et al., 2009, Lebel et al., 2008, Leemans et al., 2006, O'Donnell et al., 2009, Reich et al., 2010, Suarez et al., 2012, Verhoeven et al., 2010, Yendiki et al., 2011, Zhang et al., 2010), there is typically a trade-off that needs to be made between prior

Acknowledgments

The HUBU project is supported by the Danish Medical Research Council, and the Lundbeck Foundation.

References (110)

  • E. Heiervang et al.

    Between session reproducibility and between subject variability of diffusion MR and tractography measures

    NeuroImage

    (2006)
  • J.L. Hsu et al.

    Gender differences and age-related white matter changes of the human brain: a diffusion tensor imaging study

    NeuroImage

    (2008)
  • J.L. Hsu et al.

    Microstructural white matter changes in normal aging: a diffusion tensor imaging study with higher-order polynomial regression models

    NeuroImage

    (2010)
  • T.L. Jernigan et al.

    Postnatal brain development: structural imaging of dynamic neurodevelopment processes

    Prog. Brain Res.

    (2011)
  • D.K. Jones

    Studying connections in the living human brain with diffusion MRI

    Cortex

    (2008)
  • D.K. Jones et al.

    Spatial normalization and averaging of diffusion tensor MRI data sets

    NeuroImage

    (2002)
  • D.K. Jones et al.

    The effect of filter size on VBM analyses of DT-MRI data

    NeuroImage

    (2005)
  • D.K. Jones et al.

    White matter integrity, fiber count, and other fallacies: the do's and don'ts of diffusion MRI

    NeuroImage

    (2013)
  • T.A. Keller et al.

    Altering cortical connectivity: remediation-induced changes in the white matter of poor readers

    Neuron

    (2009)
  • C.M.R. Kitchen

    Nonparameteric vs parameteric tests of location in biomedical research

    Am J. Ophthalmol.

    (2009)
  • C. Lebel et al.

    Microstructural maturation of the human brain from childhood to adulthood

    NeuroImage

    (2008)
  • C. Lebel et al.

    Age-related regional variations of the corpus callosum identified by diffusion tensor tractography

    NeuroImage

    (2010)
  • K.S. Madsen et al.

    Response inhibition is associated with white matter microstructure in children

    Neuropsychologia

    (2010)
  • K.S. Madsen et al.

    Brain microstructural correlates of visuospatial choice reaction time in children

    NeuroImage

    (2011)
  • N. Malykhin et al.

    Diffusion tensor imaging tractography and reliability analysis for limbic and paralimbic white matter tracts

    Psychiatry Res.

    (2008)
  • S. Mori et al.

    Stereotaxic white matter atlas based on diffusion tensor imaging in an ICBM template

    NeuroImage

    (2008)
  • C. Nimsky et al.

    Intraoperative visualization of the pyramidal tract by diffusion-tensor-imaging-based fiber tracking

    NeuroImage

    (2006)
  • L.J. O'Donnell et al.

    Tract-based morphometry for white matter group analysis

    NeuroImage

    (2009)
  • G.J.M. Parker et al.

    Initial demonstration of in vivo tracing of axonal projections in the macaque brain and comparison with the human brain using diffusion tensor imaging and fast marching tractography

    NeuroImage

    (2002)
  • C. Poupon et al.

    Regularization of diffusion-based direction maps for the tracking of brain white matter fasciculi

    NeuroImage

    (2000)
  • G. Price et al.

    White matter tracts in first-episode psychosis: a DTI tractography study of the uncinate fasciculus

    NeuroImage

    (2008)
  • D.S. Reich et al.

    Automated vs. conventional tractography in multiple sclerosis: variability and correlation with disability

    NeuroImage

    (2010)
  • S.M. Smith et al.

    Tract-based spatial statistics: voxelwise analysis of multi-subject diffusion data

    NeuroImage

    (2006)
  • L. Snook et al.

    Diffusion tensor imaging of neurodevelopment in children and young adults

    NeuroImage

    (2005)
  • R.O. Suarez et al.

    Automated delineation of white matter fiber tracts with a multiple region-of-interest approach

    NeuroImage

    (2012)
  • W. Van Hecke et al.

    On the construction of an inter-subject diffusion tensor magnetic resonance atlas of the healthy human brain

    NeuroImage

    (2008)
  • W. Van Hecke et al.

    On the construction of a ground truth framework for evaluating voxel-based diffusion tensor MRI analysis methods

    NeuroImage

    (2009)
  • W. Van Hecke et al.

    The effect of template selection on diffusion tensor voxel based analysis results

    NeuroImage

    (2011)
  • P.J. Basser et al.

    In vivo fiber tractography using DT-MRI data

    Magn. Reson. Med.

    (2000)
  • C. Beaulieu

    The basis of anisotropic water diffusion in the nervous system — a technical review

    NMR Biomed.

    (2002)
  • M. Cercignani

    Strategies for Patient–Control Comparison of Diffusion MR Data

  • M. Cercignani et al.

    Mean diffusivity and fractional anisotropy histograms of patients with multiple sclerosis

    AJNR Am. J. Neuroradiol.

    (2001)
  • L.C. Chang et al.

    RESTORE: robust estimation of tensors by outlier rejection

    Magn. Reson. Med.

    (2005)
  • O. Ciccarelli et al.

    Probabilistic diffusion tractography: a potential tool to assess the rate of disease progression in amyotrophic lateral sclerosis

    Brain

    (2006)
  • J.D. Clayden et al.

    A probabilistic model-based approach to consistent white matter segmentation

    IEEE Trans. Med. Imaging

    (2007)
  • J.B. Colby et al.

    Along-tract statistics allow for enhanced tractography analysis

    NeuroImage

    (2012)
  • L. Concha et al.

    Bilateral limbic diffusion abnormalities in uni-lateral temporal lobe epilepsy

    Ann. Neurol.

    (2005)
  • L. Concha et al.

    Diffusion tensor tractography of the limbic system

    AJNR Am. J. Neuroradiol.

    (2005)
  • L. Concha et al.

    Bilateral white matter diffusion changes persist after epilepsy surgery

    Epilepsia

    (2007)
  • L. Concha et al.

    White-matter diffusion abnormalities in temporal-lobe epilepsy with and without mesial temporal sclerosis

    J. Neurol. Neurosurg. Psychiatry

    (2009)
  • Cited by (0)

    1

    These authors contributed equally.

    View full text