Subdivision of the occipital lobes: An anatomical and functional MRI connectivity study

Abstract

Exploring brain connectivity is fundamental to understanding the functional architecture of the cortex. In our study we employed tractography-based parcellation, combined with the principal component analysis statistical framework, to divide the occipital lobes into seven areas in a group of eighteen healthy participants. Tractography-based parcellation is a method based on diffusion imaging tractography, which segregates the living human brain into distinctive areas showing sharp differences in their anatomical connectivity. The results were compared to covarying functional networks involving distinct areas within the occipital lobes, that we obtained using resting state functional magnetic resonance imaging (fMRI), as well as to other existing subdivisions of the occipital lobes. Our results showed similarities with functional imaging data in healthy controls and cognitive profiles in brain-damaged patients, although several differences with cytoarchitectonic, myelogenetic, myeloarchitectonic and functional maps were reported. While the similarities are encouraging, the potential validity and limitations of the differences observed are discussed. Taken together these results suggest that tractography-based parcellation may provide a new promising anatomical subdivision of the living human brain based on its anatomical connectivity, which may benefit the understanding of clinical-neuroanatomical dissociations and functional neuroimaging results.

Introduction

During the course of the last 50 years, many developments in our knowledge of the human's occipital lobe stem from the monkey's occipital lobe having been segregated into several sharply defined subdivisions of visually responsive cortex using direct electrical recording (Felleman and Van Essen, 1991; Hubel and Wiesel, 1962; Van Essen, 2003; Van Essen and Maunsell, 1983; Zilles and Clarke, 1997). Most of these subdivisions show clear differences in their anatomical features, and given that electrical recording is very difficult in the living human brain, anatomists used these anatomical features as landmarks to parcellate the occipital lobes in humans. These anatomical landmarks included, for instance, the visible variation in the neuronal pattern of the six cortical layers (i.e., cytoarchitectonic) (Amunts et al., 2000; Brodmann, 1909; Campbell, 1905; Von Economo and Koskinas, 1925; Zilles et al., 1986), the density of myelinated fibres (i.e., myeloarchitectonic) (Smith, 1907; Vogt and Vogt, 1919; Zilles and Schleicher, 1993), the progressive myelination during maturation of the human brain (Flechsig, 1920) or how sub-regions of the occipital lobe are connected with the rest of the brain (Burkhalter and Bernardo, 1989; Clarke, 1993; Clarke and Miklossy, 1990). Two common factors underlying these subdivisions of the living human brain are that they employ post-mortem techniques and that they are difficult to translate into in-vivo specimens due to inter-individual variability. Alternative methods to subdivide the human living brain into distinct areas sharing close anatomical features are therefore required to allow us to associate variability in these subdivisions with healthy and pathological conditions.

Exploring brain connectivity is a fundamental approach to study the functional architecture of the cortex (Mesulam, 2005). For example, although neurons from visual and auditory cortices share similar anatomy and organisation, they support different functions, and show clear differences in their anatomical connectivity with the rest of the brain. These differences in connectivity, may explain their functional specificity (in humans see Catani et al., 2003, 2005; ffytche et al., 2010; in monkeys see Petrides and Pandya, 2009; Yeterian and Pandya, 2010). This suggests that one of the best methods to studying the functional specialisation of specific brain regions is to examine the nature of the input and output of that region (Van Essen and Maunsell, 1983).

The study of both anatomical and functional connectivity in the living human brain has shown a considerable expansion in the last decade thanks to the development of diffusion-weighted imaging (DWI) tractography techniques (Basser et al., 2000; Dell'acqua and Catani, 2012; Dell'acqua et al., 2010; Jones, 2008) as well as functional magnetic resonance imaging (fMRI) connectivity methods (Deco et al., 2011; Fox and Raichle, 2007; Greicius et al., 2003; Mantini et al., 2007). Tractography studies reconstruct white matter anatomical features in the living human brain, which show similarities with those reported in post-mortem animal tracing studies (Dauguet et al., 2007; Rilling et al., 2008; Thiebaut de Schotten et al., 2011a, 2012) and in human brain dissections (Catani et al., 2012; Lawes et al., 2008; Thiebaut de Schotten et al., 2011b). Resting state fMRI (rsfMRI) connectivity studies identify brain areas with similar dynamics of blood/oxygen level (BOLD) signal changes, i.e., covarying functional networks. Hence, rsfMRI connectivity has been employed to decompose data into a set of distinct spatial maps, each with its own time course (Kiviniemi et al., 2003). Recently, independent component analysis has been optimised to extract the major covarying networks in the resting brain, as imaged with rsfMRI (Beckmann et al., 2005). These networks are very consistent across subjects (Damoiseaux et al., 2006) and closely correspond to major task-related networks as identified using standard fMRI paradigms (Smith et al., 2009). However, it is unknown whether all these functional connectivity networks reflect specific anatomical features. Some rsfMRI connectivity studies showed patterns of connectivity comparable to those revealed with anatomical tractography (Greicius et al., 2009; Honey et al., 2009; Skudlarski et al., 2008), but functional and anatomical connectivities are not necessarily similar. For example, rsfMRI connectivity may vary according to the functional state of the measured brain (Fox et al., 2006; Ginestet and Simmons, 2010; Hampson et al., 2006; He et al., 2007; McAvoy et al., 2008; Seeley et al., 2007; Zhou et al., 2011). Hence, anatomical connectivity and functional connectivity are not strictly equivalent but rather complementary (Deco et al., 2011; Honey et al., 2009).

Tractography can also be used to divide a brain area into sub-regions defined by a similar anatomical connectivity pattern. Tractography-based parcellation has been employed to segregate functionally different but anatomically adjacent areas, such as the supplementary motor area (SMA) and the pre-SMA (Johansen-Berg et al., 2004) or Brodmann areas (BA) 44 and 45 (Anwander et al., 2007), because they exhibit sharp changes in their connectivity. These results suggest a close alignment between tractography-based parcellation of the living human brain and functional and cytoarchitectonic regions (Jbabdi and Behrens, 2012). More recently other authors employed tractography-based parcellation to describe a new in-vivo parcellation of the inferior parietal cortex into five regions (Mars et al., 2011) that correspond to recent post-mortem cytoarchitectonic atlases (Caspers et al., 2008, 2006). However, the statistical framework used for tractography-based parcellation is limited by the need to determine a priori the number of sub-regions to expect from the parcellation, which relies on the experimenter's subjectivity. This is a crucial issue in cases of blind parcellation approaches that aim to detect functional sub-units without strong prior hypotheses (Jbabdi et al., 2009).

In the present study, we employed a principal component analysis, which divided the occipital lobe into several regions that showed a different pattern of anatomical connectivity as revealed by probabilistic tractography, without determining a priori the number of sub-regions. We report seven distinct sub-regions in each occipital lobe, as well as their hemispheric asymmetries. Results are compared to rsfMRI covarying networks obtained in the same subjects, and also to standard myeloarchitectonic (Smith, 1907), cytoarchitectonic (Caspers et al., 2012, 2008, 2006; Kujovic et al., 2012; Rottschy et al., 2007), myelogenetic (Flechsig, 1920) maps obtained from post-mortem dissections, functional maps transposed from monkey to humans (Van Essen, 2003), fMRI results and clinical-anatomical conclusions built from neuropsychological studies.