Functional imaging studies identified a motion-sensitive area (V5/MT+) in the vicinity of the posterior branch of the inferior temporal sulcus that has no correlate in any classical cytoarchitectonic map. The aim of the present study was to identify a cytoarchitectonic correlate of this region in 10 human postmortem brains and to provide a probability map of this area. Observer-independent mapping revealed an area, hOc5 (h for human, Oc for occipital lobe), that has a broad layer III, a high cell density in layer II/III, and a low one in layer V. Most of area hOc5 is found in the depths of the anterior occipital sulcus and the anterior parts of either the inferior lateral occipital or the inferior occipital sulcus. After 3-dimensional reconstruction and registration to a standard reference space, a probability map of the area measured the individual variability of its size and location. The mean spatial locations of area hOc5 are −43, −73, 10 (left) and 49, −70, 11 (right). The locations and their relationships to sulci strongly suggest that hOc5 is the cytoarchitectonic correlate of human V5/MT+. This hypothesis was supported by comparing the cytoarchitectonic probabilistic map with results from a functional imaging study.
Recent functional imaging and neurophysiological studies have shown that the visual cortices contain numerous and complex visual areas (van Essen and Zeki 1978; Zeki 1978; Fox and others 1987; Tootell and others 1996). One functionally defined area is V5/MT+. The first evidence that area V5/MT+ exists in the human visual cortex came from a clinical study of a brain-damaged patient who had a deficit in perceiving moving stimuli. After a bilateral stroke, which affected the lateral temporo-occipital cortex and the underlying white matter, the patient suffered from motion blindness. Her abilities to detect visual stimuli and locate them in 3-dimensional (3D) space were not impaired (Zihl and others 1983). Although the lesion was quite large, its location both in her and in later patients and the nature of the neurological impairments suggested the existence of a cortical area homologous to the nonhuman primate motion-sensitive area V5/MT+, which is in the middle temporal region (Dubner and Zeki 1971; Zeki 1971, 1995; van Essen and others 1981; Maunsell and van Essen 1983).
The existence of area V5/MT+ has been demonstrated in healthy and dyslexic human subjects in electrophysiological and functional imaging studies using positron emission tomography, functional magnetic resonance imaging (fMRI), transcranial magnetic stimulation (TMS), and magnetoencephalography (MEG) (Corbetta and others 1991; Zeki and others 1991; Watson, Myers, and others 1993; Zeki 1993; Dupont and others 1994, 1997; Shipp and others 1994; Tootell and others 1995; Anderson and others 1996; Reppas and others 1997; Vanni and others 1997; Büchel and others 1998; Goebel and others 1998; Hasnain and others 1998; Smith and others 1998; Dumoulin and others 2000; Morrone and others 2000; Dukelow and others 2001; Huk and others 2002; Wunderlich and others 2002; Claeys and others 2004; McGraw and others 2004; Miki and others 2004; Zafiris and others 2004; Noguchi and others 2005; Walters and others 2006). Other studies reported activity in V5/MT+ that correlated with attention or choice when the subject viewed ambiguous images, dealt with binocular rivalry, or detected shape and tactile motion (Hagen and others 2002; Krug 2004).
fMRI has the highest spatial resolution among the above methods, and its use has located V5/MT+ on the lateral surface of the anterior part of the occipital lobe, at the continuation of the inferior temporal sulcus (Dumoulin and others 2000; Huk and others 2002). This region, however, shows a considerable variability in the presence and shape of sulci (Ono and others 1990; Duvernoy 1991). In general, it has been assumed (Zilles and Clarke 1997) that an anatomic V5/MT+ exists and is correlated with the borders of Brodmann areas (BAs) 19 and 37 or with von Economo and Koskinas' areas OA (Area peristriata) and PH (Area parietalis occipito-temporalis).
The presently available cytoarchitectonic maps of Campbell (1905), Elliot Smith (1907), Brodmann (1909), von Economo and Koskinas (1925), and the Russian school (Sarkisov and others 1949) do not indicate a separate cortical area along the lateral surface of the anterior part of the occipital lobe. Brodmann's influential map proposed a tripartite division of the visual cortex: a striate core and 2 extrastriate areas surrounding the former in beltlike forms. von Economo and Koskinas (1925) followed this line in general but further subdivided the extrastriate areas on the basis of cytoarchitectonic modifications. Area OA, which is the outermost of the 3 areas, was given 3 subregions: OA1, OA2, and OAm, but none of these subregions match the human V5/MT+ with respect to location and/or size. The 3-way partition of the occipital visual cortex presented in these classical cytoarchitectonic maps disagrees with the more complex parcellation of the macaque cortex (Gattass and others 1988; Colby and Duhamel 1991; Fellemann and van Essen 1991; van Essen and others 1992) and recent human data on retinotopic mapping (Sereno and others 1995; DeYoe and others 1996; Tootell and others 1996; Reppas and others 1997; Tootell and Hadjikhani 2001).
Several recent studies support a microstructural correlate of functionally defined V5/MT+ within the human extrastriate cortex (Clarke and Miklossy 1990; Zilles and Schleicher 1993; Clarke 1994; Tootell and Taylor 1995). These studies showed a heavy myelination of putative human V5/MT+ and widespread callosal connections. An increased density of myelin had also been observed by Flechsig (1927) in his area 16. The myelin density of area V5/MT+, measured in profiles spanning the cortex from the pial surface to the white matter, increased considerably from superficial to the deep layers (Zilles and Schleicher 1993). In particular, heavy myelination characterized layer IV, suggesting an anatomical correlate of the outer stria of Baillarger. These layers are also highly CAT-301 (a monoclonal antibody to a proteoglycan at the extracellular neuronal surface) positive (Tootell and Taylor 1995). The deep myelination most likely reflects the inner stripes of Baillarger. The callosal afferents are widespread throughout the entire V5/MT+ area but are densest in its lower (ventral) part, suggesting V5/MT+ may be further divided (Clarke and Miklossy 1990). Recently, an automated method for detection of borders in histological sections delineated a myeloarchitectonic area, which has been interpreted as being the human V5/MT+ (Annese and others 2005). Finally, cytochrome oxidase staining showed a very dark band in putative area V5/MT+, which was absent in adjacent areas (Clarke 1994). None of these anatomical studies, however, provided the putative area V5/MT+ with stereotaxic coordinates in a standard reference space, a necessary prerequisite for subsequent comparison with data from functional imaging.
Moreover, motion sensitivity has been demonstrated in regions beyond V5/MT+, especially in the temporal and more posterior occipital regions (Tootell and others 1995; Orban and others 1999; Sunaert and others 1999; Bristow and others 2005; Noguchi and others 2005). Several studies showed that subtle differences in motion paradigms produce topographic differences in the elicited activation within a relatively small brain area (Morrone and others 2000; Huk and others 2002; Pelphrey and others 2005). These data suggest that multiple cortical areas exist within the temporo-occipital transition region, but the degree of heterogeneity of this region has not yet been addressed in the neuroanatomical literature.
The methodology of associating structural with functional regions compares cortical locations of 2 sets of maps: a structural map and a functional one. Each set has a known probability of occupying a particular location in a reference space. The overlap of the 2 map sets can then be measured. In this way, a location with a known probability of containing a defined cellular architecture can be tested for its association with a location defined as containing particular brain activations from imaging studies. Such probabilistic anatomical maps of cortical areas have been successfully applied in studies of the visual system (Larsson and others 1999, 2002; Barnikol and others 2006; Wilms and others 2005; Wohlschläger and others 2005).
To begin analyzing the anatomical regions, we examined the presumed V5/MT+ region and within it defined a distinct cytoarchitectural area, its borders, its stereotaxic location within a reference space, and its relationship to the appearance of sulci and gyri. We used an objective and reproducible methodology to define the area without bias. To avoid a term suggesting an unproven relationship between the cytoarchitectonic parcellation and brain function, we use a neutral nomenclature for the cytoarchitectonically delineated area, that is, hOc5 (h for human, Oc for occipital lobe, 5 for the area number 5 when moving laterally from the primary visual cortex [BA 17/V1/hOc1]). Areas hOc2 to hOc4 are located between hOc5 and the primary visual cortex: hOc3v and hOc4v border BA 18 (V2/hOc2) in the ventral extrastriate cortex (Rottschy and others 2005). In 2 accompanying papers (Barnikol and others 2006; Wilms and others 2005), we tested the hypothesis that the cytoarchitectonic area hOc5 and the functionally defined area V5/MT+ share a common spatial position.
Materials and Methods
Preparation of the Postmortem Brains
Cytoarchitectonic analysis was performed in 10 human postmortem brains (5 male and 5 female; mean age 66.3 years, 37–85 years; postmortem delay less than 36 h; Table 1) obtained from the body donor program of the Institute of Anatomy, University of Düsseldorf, Germany, in accordance with the guidelines of the Ethics Committee of the University of Düsseldorf. All brains came from subjects with no history of neurological or psychiatric diseases in their clinical records, with the exception of brain number 3, which came from a patient with transitory motor disabilities. Handedness is unknown. The sample is the same as that of our earlier study of BAs 17 and 18 (Amunts and others 2000).
|Brain||Age (years)||Gender||Cause of death||Postmortem delay (h)||Weight prior to fixation (g)||Shrinkage factor|
|1||79||F||Carcinoma of the bladder||24||1350||1.93|
|10||85||F||Mesenteric artery infarction||14||1046||1.67|
|Brain||Age (years)||Gender||Cause of death||Postmortem delay (h)||Weight prior to fixation (g)||Shrinkage factor|
|1||79||F||Carcinoma of the bladder||24||1350||1.93|
|10||85||F||Mesenteric artery infarction||14||1046||1.67|
The brains were removed from the skull and fixed in either 4% formalin or Bodian's fixative (a mixture of ethanol, formaldehyde, and glacial acetic acid) for several months (Table 1).
The brains were further processed for 3D reconstruction, and detailed descriptions are available (Amunts and others 2000, 2004). In short, each postmortem brain was magnetic resonance (MR) imaged prior to embedding and cutting. This step gives a reference data set to correct the inevitable distortions of brain shape and size that occur during histological processing. T1-weighted images (3D-FLASH scans) of the entire brain were acquired using a Siemens 1.5-T magnetron scanner (Erlangen, Germany) (flip angle 40°, repetition time = 40 ms, echo time = 5 ms). The spatial resolution was 1 × 1 × 1.17 mm. There were approximately 128 sagittal sections per brain.
After MR imaging, the brains were embedded in paraffin and sectioned (coronal sections, 20 μm). During sectioning, images of the block faces of the embedded brain were digitized using a charge-coupled device camera (=block-face images). Sections were stained for cell bodies using a modified silver method (Merker 1983). This method resembles Nissl staining but provides a higher contrast between cell bodies and neuropil. Every 60th section of the entire histological series was digitized and subjected to cytoarchitectonic analysis. The MR sequences of the fixed brain, the block-face images, and the digitized histological sections were used to create 3D-reconstructed histological volumes corrected for distortions caused by histological processing (Schormann and Zilles 1998; Amunts and others 2000).
We used an algorithm-based approach to define borders between cytoarchitectonic areas in a reproducible and statistically testable manner (Schleicher and others 1999). The cellular architecture was analyzed in cortical regions of interest (ROIs), which were circumscribed by stereotaxic coordinates published in functional imaging studies of area V5/MT+, as well as by gross anatomical landmarks such as the temporo-occipital junction and the rostral third of the occipital lobe, that is, the presumed location of human V5/MT+ (Clarke and Miklossy 1990; Watson, Frackowiak, and Zeki 1993; Tootell and Taylor 1995; Dumoulin and others 2000; Huk and others 2002; Annese and others 2005).
The gray level index (GLI), a measure of cell-packing density, was obtained in these ROIs (Wree and others 1982; Schleicher and Zilles 1990) (Figs 1 and 2). The GLIs were collected using a KS400® image-analyzing system (Zeiss, Jena, Germany) connected to a microscope with a motorized scanning stage (lens: Planapo® 6.3 × 1.25, Zeiss). The ROIs were scanned with a TV camera (XC-75, Sony, Tokyo, Japan) by a continuous, mosaiclike pattern using TV frames (524 × 524 μm) of 512 × 512 pixels (Fig. 2). The size of each measuring field was 20 × 20 μm. A binary image was generated from each TV frame by adaptive thresholding and the GLI measured (Fig. 2). The procedure resulted in a GLI image for each ROI and was used to extract GLI profiles (see below). Because GLI images are calculated from binary images, variations in staining intensities do not influence the GLI image (Schleicher and Zilles 1990).
The GLI profiles, measures of cell volume densities, were extracted along trajectories oriented perpendicular to the cortical layers and extended from the border of layer I/II to that of layer VI/subcortical white matter (Fig. 2B,E,H).The GLI profiles were standardized to a cortical depth of 100% to compensate for variations in cortical thickness. The positions of the GLI profiles paralleled the cortical surface and were consecutively numbered from 1 to n (n = total number of profiles per ROI). The cytoarchitecture was measured, that is, its shape defined, by quantifying a set of 10 features (=feature vector), which were based on the central moments of the profiles. Features included mean GLI value, the cortical depth of the center of gravity of the profile, the standard deviation (SD), the skewness, the kurtosis, and the analogous parameters for the first derivative of the profile (Amunts and others 1999; Schleicher and others 1999). Profile values vary with changes in the laminae. We used in-house software based on MATLAB® (Version 5.3, MathWorks, Natick Massachusetts, USA) to define the cortical borders (Schleicher and others 1999).
Each cytoarchitectonic area has a GLI pattern, which is described by a set of feature vectors. Differing patterns of GLI profiles demarcate a border between 2 areas, and the maxima and minima are associated with significant alterations (Schleicher and others 1998, 1999, 2000, 2005). A multivariate distance (Mahalanobis distance, D2) measures the difference between the sets of profiles of 2 cytoarchitectonic areas (Fig. 2C). The statistical significance between the 2 populations was tested with Hotellings T2-test, and a Bonferroni correction controlled for multiple comparisons (Bartels 1979).
To detect and quantify differences in laminar patterns, the measurements were tested between pairs of adjacent blocks, each comprising n neighboring profiles (Fig. 2D,G,J). Adjacent blocks of profiles were moved as sliding windows along the cortex in steps of one profile position. For each position, a new D2 was calculated. This procedure was repeated with block sizes ranging from 8 (approximately 1024 μm) to 24 profiles (approximately 3070 μm). The distance function, D2, showed maxima at locations where the shape of the GLI profiles changed. The higher the D2 value between 2 adjacent blocks of profiles, the greater the difference in cytoarchitecture. Borders were defined using 3 or more sections per hemisphere and brain. Positions of D2 values were accepted as borders if the values were significant and if borders at comparable positions were confirmed in adjacent sections (Fig. 2B–J).
The neighboring cortical areas of hOc5 were named “V” (ventral neighbor of hOc5) and “D” (dorsal neighbor of hOc5).
Definition of Volumes
Depending on the extent of each area, 9–16 sections were analyzed per hemisphere and brain (Table 2). The areas were measured in images with a size of 14 000 × 12 000 pixels, that is, the spatial resolution was approximately 95 pixel/mm.
|Average||0.381 ± 0.127||0.462 ± 0.277||0.844 ± 0.387|
|Average||0.381 ± 0.127||0.462 ± 0.277||0.844 ± 0.387|
Note: Total refers to both hemispheres.
The shrinkage factors were calculated as ratios between the fresh volume of a brain and the volume reconstructed from its histological sections. The fresh volumes are the fresh weight of the brain × its mean specific density of 1.033 (Kretschmann and others 1982).
We compared the volumes between hemispheres using a paired t-test. The mean volume of a cytoarchitectonically defined area is the average volume between the left and the right hemispheres.
Probabilistic Mapping in Stereotaxic Space
Following the cellular analysis, we made a 3D reconstruction of the histological sections and its cytoarchitectonic areas. The extent of area hOc5 was determined in histological sections (Fig. 3) and then interactively mapped to the corresponding digitized images.
The association of structure and function requires maps with known probabilities of stereotaxic localities (Roland and Zilles 1994). We registered our reconstructed volumes to the T1-weighted single-subject reference space of the Montreal Neurological Institute, the MNI space (Evans and others 1993; Collins and others 1994; Holmes and others 1998), using a combination of an affine linear, nonlinear elastic, and gray-level transformations (Henn and others 1997; Schormann and Zilles 1998; Mohlberg and others 2003; Amunts and others 2004). Because the origin of the MNI space does not coincide with the orientation using the superior posterior edge of the anterior commissure in the interhemispheric fissure (Talairach and Tournoux 1988), we linearly transformed our anatomical data to the anatomical MNI space (Amunts and others 2005). The transformation was a simple linear translation in rostrocaudal and ventrodorsal directions. The origin and orientation of the anatomical MNI space accords with the convention of Talairach and Tournoux (1988).
The amounts of individual variation in reconstructed space and location were also registered to the standard format of a reference brain (Amunts and Zilles 2001; Zilles and others 2002). We superimposed individual hOc5 volumes onto the 3D reference space. An anatomical probabilistic map showed the relative frequency with which hOc5 was present in each voxel of the reference space. The degrees of overlap among the 10 individual areas were color coded. Centers of gravity of these maps were positioned in the anatomical MNI space for each brain and hemisphere.
Cytoarchitecture and Cortical Borders of hOc5
We delineated a new cytoarchitectonic area, hOc5, in the region of the anterior occipital sulcus, the superior and the inferior lateral occipital sulci, the inferior occipital sulcus, and the temporo-occipital incisure (preoccipital notch) (Fig. 1). Its cytoarchitecture differs from that of the adjacent areas by having a higher cell-packing density in layers II and III, a prominent layer III with a clear radial arrangement of neurons in columns, large pyramidal cells in sublayers IIIb and IIIc, a lower cell-packing density in layer V than in layer III, and, on average, smaller pyramidal cells in layer V than in layer III (Figs 4 and 5). The largest pyramidal cells of layer V, however, are approximately the same size as those in layer III. The thickness of layer III is greater than that of layers IV, V, and VI taken together. Layer III was subdivided into sublayers IIIa, IIIb, and IIIc. Pyramidal cells in these sublayers increase in size from sublayer IIIa to IIIc. Sublayer IIIc has a sharp border with granular layer IV. Layer IV is located approximately between the middle and the lower third of the cortical cross section. Layer V has a relatively low cell-packing density with the pyramidal cells arranged in radial columns. It is composed of 2 sublayers (Va and Vb). On average, layer V pyramidal cells are smaller than those of sublayer IIIc (Figs 4 and 5).
Dorsal to area hOc5, we found an extrastriate area (area D), which has the following cytoarchitectonic characteristics: layers II, IIIa, and IIIb have a low cell-packing density; the pyramidal cells in sublayer IIIc are small; and layer IV is thick. Layer IV is also located higher in cortical cross section than it is in area hOC5. Layers V and VI are less densely packed and are thicker in area D than in hOc5. Neurons in layers V and VI are smaller than those of area hOc5 (Figs 4 and 5).
Compared with cells in hOc5, the extrastriate area ventral to hOc5 (area V) has larger pyramidal cells in sublayer IIIc, a more pronounced radial arrangement of cells in layers V and VI, a larger mean size of pyramidal cells in layer V, and denser cell packing in layers V and VI (Figs 4 and 5).
The analysis of the cytoarchitectonic borders revealed highly significant changes in laminar patterns at the borders between area hOc5 and its dorsal and ventral neighbors, areas D and V. The external borders of hOc5 match the significant maxima of the distance function.
In some sections, we found a significant subdivision within hOc5, a dorsal and ventral part (hOc5d, hOc5v) (Figs 2 and 4). The dorsal subdivision has a higher cell-packing density in layer II and a lower one in layer V than the ventral one. The 2 subdivisions are quite similar to each other and clearly differ from their non-hOc5 dorsal and ventral neighbors. At this time, we analyzed them as one area, hOc5.
Location with respect to Sulci and Sulcal Variability
Area hOc5 is located in the region of the temporo-occipital junction, where both constant and variably present sulci are observed. The anterior occipital sulcus was identified in all 20 hemispheres as a principal sulcus. It has a ventrodorsal orientation along the border between the temporal and occipital lobes and is the continuation of the ascending branch of the inferior temporal sulcus in most of the hemispheres (13 out of 20, 65%). In the remaining 7 hemispheres, the anterior occipital sulcus is the superior continuation of the temporo-occipital incisure (preoccipital notch) directed upward from the inferolateral margin of the hemisphere (Fig. 7). The lateral surface of the occipital lobe sometimes has 2 lateral occipital sulci, a superior and an inferior one (Fig. 1B). The superior lateral occipital sulcus is an inconstant sulcus. It is always located immediately above the inferior lateral occipital sulcus. In 2 hemispheres, it connects to the anterior occipital sulcus. The inferior lateral occipital sulcus is a constant and clearly defined sulcus, identified in all 20 hemispheres. In 9 hemispheres, the inferior lateral occipital sulcus connects to the anterior occipital sulcus. The junction between these 2 sulci is the most constant point of the sulcal intersections on the lateral surface of the occipital lobe (Fig. 7).
In addition to the lateral occipital sulci (superior and inferior), another variable occipital sulcus was identified—the inferior occipital sulcus. This sulcus, present in 75% of cases (15 hemispheres), was located immediately below the inferior lateral occipital sulcus, along or near the inferolateral margin of the occipital lobe. In 10 hemispheres, the inferior occipital sulcus is a posterior continuation of the inferior temporal sulcus, whereas in 5 hemispheres, it was identified as the sulcus that runs caudally from the anterior occipital sulcus (Figs 1B and 7).
Area hOc5 is located close to the intersection (actual or interpolated) of the anterior occipital and the inferior lateral occipital sulci in the region of the temporo-occipital junction. Its major part is found in the depths of sulci, with only a minor part covering the free surface of the occipital gyri. Area hOc5 was located along the axis formed by the anterior occipital sulcus and 2 caudally located sulci, the inferior lateral occipital and inferior occipital sulci. It extends along the posterior bank of the anterior occipital sulcus (65% of the cases, 13 hemispheres) and farther posteriorly to the inferior lateral occipital sulcus, where it occupies its bottom and inferior bank (20% of the cases, 4 hemispheres). In the remaining 15% of the cases (3 hemispheres), it was located within the inferior occipital sulcus (Figs 3, 6, and 7). Area hOc5 was not found in either the superior lateral occipital sulcus or the inferior temporal sulcus.
Volume of the Area hOc5
The mean total volume (volume of both sides together) of area hOc5 is 0.844 cm3 (SD = 0.39). The mean right volume is slightly, but not significantly, larger (0.463 cm3, SD = 0.28) than that of the left side (0.381 cm3, SD = 0.13, P > 0.05). Six brains had a larger mean volume in the right than in the left hemisphere. Four brains showed a larger volume on the left than on the right. The individual volumes varied by a factor of 2.8 and 5.5 for the left and right hemispheres, respectively.
Stereotaxic Location and Individual Variation
The anatomical probability map of hOc5 quantified its spatial variability in extent and location. The individual variability of the area is considerable (Figs 7 and 8). Of the 10 brains, at most 6 overlapped on the left and 7 on the right hemisphere (Table 3). The volumes based on the total probability maps were considerably larger than the mean volumes obtained from measuring the histological sections (see above). The volumes derived from the probability maps with a threshold of 4 brains are closest to the mean volume estimated in the histological sections, that is, the volumes are derived using voxels that overlap in 4 or more of the 10 brains (=40% probability map). The 40% probability volumes are 622 and 719 mm3 for the left and right hemisphere, respectively.
Note: The 40% map contains all those voxels that overlapped in 4 or more out of 10 brains. The 40% volumes were 622 mm3 (527 + 87 + 8, left) and 719 mm3 (399 + 223 + 89 + 8, right).
The mean centers of gravity of area hOc5 based on the 40% probability maps are −43, −73, 10 (left) and 49, −70, 11 (right). The area is located more laterally (by 6 mm), more rostrally (by 3 mm), and more dorsally (by 1 mm) in the right than in the left hemisphere.
Quantitative Cytoarchitecture of hOc5
Applying an algorithm-based approach to detect cytoarchitectonic borders enabled us to define a new extrastriate cytoarchitectonic area objectively and precisely. When we compared our results with those of classical cytoarchitectonic criteria, we found that the sizes of pyramidal cells in layers III and V most reliably differentiate area hOc5 from its neighbors. The sizes of pyramidal cells in layers III and V gradually increase from area hOc5 to the ventrally located area V and the inferolateral margin of the occipital lobe.
von Economo and Koskinas (1925) noted that their subregion OAm has large pyramidal cells in deep layer III and a radial arrangement of cells in layers V and VI. According to these authors, subregion OAm lacks large pyramidal cells in layer V. These features agree with our finding, namely, that large pyramidal cells rarely appear in layer V of area hOc5 and their sizes are smaller than those in sublayer IIIc.
Our cytoarchitectonic description of hOc5 also agrees with much of the description of the eulaminate preoccipital cortex of Bailey and von Bonin (1951). This cortical region starts directly caudal to the anterior occipital sulcus and is characterized by quite large pyramidal cells in sublayer IIIc, whereas layer V contains medium-sized pyramidal cells (Bailey and von Bonin 1951, p. 133, Fig. 57, Block X, Section 100). The cortex below this region (along the inferolateral margin of the occipital lobe) changes clearly: It contains large pyramidal cells in IIIc and numerous pyramidal cells in layer V that are approximately the same size as those in IIIc. They tend to accumulate in the middle part of layer V. Bailey and von Bonin (1951, p. 136, Fig. 57, Block X, Section 500) described this region as the eulaminate, temporo-occipital cortex dominated by layer V.
The pigmentarchitectonic map of Braak (1977) contains a different set of extrastriate areas, but the study of neither Braak (1977) nor von Economo and Koskinas (1925) details the relationship of the architectonic areas with gross anatomical landmarks.
Our dorsal and ventral subdivision of hOc5 may correspond to the divisions observed by Clarke and Miklossy (1990) and Tootell and Taylor (1995). The results cannot be compared directly with each other, however, as the studies used different staining techniques (cell bodies, myelin, cytochrome oxidase) and the earlier ones were not able to use a common reference space. Other researchers did not subdivide the myeloarchitectonically defined MT (Annese and others 2005).
Another alternative is that the 2 subdivisions of hOc5 might correspond to MT and MST (medial superior temporal area), (Huk and others 2002). Area MST, however, was located along the anterior (temporal) bank of the anterior occipital sulcus (the ascending ramus of the inferior temporal sulcus), a region where we never observed the cytoarchitectonic area hOc5. In addition, the dorsal and ventral subdivisions of hOc5 do not have a caudal-to-rostral arrangement as observed in MT and MST (Huk and others 2002). Therefore, it is less probable that the 2 cytoarchitectonic subdivisions of hOc5 correspond to functionally defined MT and MST. The above authors did, however, report that only part of the MT was activated. This is an indirect evidence of a further subdivision of MT.
Morrone and others (2000) identified 2 subregions of V5/MT+ that differed in their responses to circular and radial flow motion as compared with translation motion. The 2 subregions showed a dorsal-to-ventral organization, but their centers were separated by more than 1 cm (Morrone and others 2000), whereas the cytoarchitectonic subdivisions we observed in hOc5 are direct neighbors. On the other hand, the topographic location of their translation-sensitive and flow areas on the lateral brain surface (their Fig. 8) accords with those of our cytoarchitectonic subareas (Fig. 7) suggesting some correspondence.
An alternative model suggests that the subparcellation may reflect a coarse somatotopy. A recent fMRI study analyzed the activity in the posterior temporal–occipital cortex region including the superior temporal sulcus (STS), which was elicited by moving the eyes, mouth, or hand (Pelphrey and others 2005). The researchers determined that hemodynamic response amplitudes of the different movements follow an anterior-to-posterior distribution and may reflect a differential involvement of the STS region in directing spatial attention and in social communication. They concluded that the topography of the right posterior lateral temporal–occipital region supports the observed differential responses to motion (Pelphrey and others 2005).
The differential activation patterns in the above motion studies suggest that our observed structural segregation resembles its functional complexity. In order to solve questions about the structural–functional relationships, it will be necessary to 1) complete the cytoarchitectonic mapping of the occipital cortex including its transitions to the parietal and temporal cortices, where responses to motion tasks have also been reported (Tootell and others 1995; Orban and others 1999; Sunaert and others 1999; Bristow and others 2005; Noguchi and others 2005), and 2) compare the anatomical probability maps of cytoarchitectonically defined areas with functional maps of this region within the same reference space, thereby excluding methodical confounds. As a first step, 2 recent studies (employing MEG and fMRI, respectively) combined the cytoarchitectonic map of hOc5 with functional activations during visual motion tasks (Wilms and others 2005; Barnikol others 2006).
Spatial Location of Area hOc5 and Intersubject Variability
Like V5/MT+, area hOc5 was found along the temporo-occipital junction, within or near the anterior occipital sulcus (Watson, Myers, and others 1993; Ffytche and others 1995; Tootell and Taylor 1995; Dumoulin and others 2000) or within the inferior lateral occipital sulcus (Clarke and Miklossy 1990; Clarke 1994; Dumoulin and others 2000; Annese and others 2005). In the anterior occipital sulcus, area hOc5 always occupied the posterior (occipital) bank of the sulcus. This accords with the data of Huk and others (2002).
It has to be mentioned, however, that the sulcal pattern of the junction region of the occipital and temporal cortices is complex and highly variable (von Kuhlenbeck 1928; Jouandet and others 1989; Ono and others 1990; Kennedy and others 1998; Thompson and others 1998). Moreover, there is no commonly accepted nomenclature of the sulci. For example, the anterior occipital sulcus was first described by Wernicke (1876) and later analyzed in more detail by others (Kohlbrugge 1909; Shellshear 1927; Connolly 1950; Bailey and von Bonin 1951; Papez 1961; Kodama and Oawa 1974; Nieuwenhuys and others 1988; Ono and others 1990; Duvernoy 1991; Tzourio-Mazoyer and others 2002). It was called “sulcus praeoccipitalis” by Meynert (1877) and Genna (1924). In other human studies (Watson, Myers, and others 1993; Ffytche and others 1995; Dumoulin and others 2000; Annese and others 2005), the anterior occipital sulcus has been described as the ascending limb of the inferior temporal sulcus, although Eberstaller (1890) first described this sulcus as the “middle temporal sulcus” rather than the inferior temporal sulcus. Our study frequently (two-third of our cases) agrees with those that observed the anterior occipital sulcus originating from the inferior temporal sulcus. In one-third of the cases, however, this sulcus originated from the temporo-occipital incisure and was clearly separated from the inferior temporal sulcus. Considering these different origins, the term “anterior occipital sulcus” seems to be more appropriate. This sulcus was always present in our sample.
The lateral surface of the occipital lobe may contain 2 lateral occipital sulci, the superior and the inferior. The inferior lateral occipital sulcus is the more constant of the two; the superior lateral occipital sulcus appears in only 50% of the cases. The inferior lateral occipital sulcus is more frequently connected to the anterior occipital sulcus than is the superior lateral. Therefore, the junction between the anterior occipital and the inferior lateral occipital serves as a relatively reliable anatomical landmark for the position of hOc5.
The inferior occipital sulcus, present in 75% of the cases in our sample, is located below the inferior lateral occipital sulcus and along the inferolateral margin of the occipital lobe. A similar sulcal pattern was found in the studies of von Kuhlenbeck (1928) and von Economo and Horn (1930). The inferior occipital sulcus was also mentioned by Broca (1878), Connolly (1950), and Papez (1961). The coexistence of all 3 occipital sulci in a hemisphere, that is, the 2 lateral sulci and the inferior one, is rare (20% of our sample).
Thus, area hOc5 is found in the anterior part of the occipital lobe, in a sulcus that may be described as the continuation of the inferior temporal sulcus. This oversimplification, however, is only a rough macroanatomical estimate, given the complex and variable patterns of occipital sulci, and the exact relationship of hOc5 to the surrounding sulci must be defined by a case-by-case manner.
The hOc5 volumes, as estimated from histological sections, vary by as much as a factor of 5. The variability is reduced after normalization to the standard reference space, but the combination of an absolutely small hOc5 volume with its large volumetric variability produces little overlap when the brains are registered in the common reference space.
After normalization to the standard reference space, the right area hOc5 occupied more lateral positions and reached the superficially exposed cortical surface more frequently than the left. A similar left–right difference has been reported for areas 17 (V1, hOc1) and 18 (V2, hOc2); their centers of gravity are more lateral and more rostral in the right than in the left hemisphere (Amunts and others 2000).
Individual variation in the location and extent of area hOc5 in the standard space is considerable. The relatively small overlap between subjects may be caused by its high volumetric variability (see above), but the elastic spatial normalization procedure may also contribute. A higher variability (smaller maximal overlap) is found in the right than in the left hemisphere. The 40% maps are rather symmetric but show slightly smaller volumes on the left than on the right. In general, the smaller the cytoarchitectonic area and the more the area deviates from an ellipsoid or sphere, the greater the influence of methodological factors (registration, smoothing, and interpolation) on the overlap within the probability map. Consequently, the higher variability in hOc5 on the left than on the right side corresponds with the smaller left than right volumes.
Using observer-independent techniques, we identified a cytoarchitectonically defined region, hOc5, in 10 human postmortem brains and produced probability maps of this area. Area hOc5, found in the depths of the anterior occipital and the anterior parts of either the inferior lateral occipital or inferior occipital sulcus, has a broad layer III, a high cell density in layer II/III, and a low one in layer V.
The interesting question whether or not hOc5 corresponds to V5/MT+ appears to be affirmative. Proving that the cytoarchitectonically defined area hOc5 is equivalent to the functionally defined area V5/MT+ requires that results from 2 types of studies be compared within a common reference space. A solid match of area hOc5 with V5/MT+ has been found (Barnikol and others 2006; Wilms and others 2005). The topography of hOc5 described in this paper also matches the region described in many studies as human V5/MT+, suggesting that hOc5 and V5/MT+ are the same region.
We also observed an inconsistent subdivision of hOc5 into dorsal and ventral component and the identification of neighboring dorsal and ventral regions, which do not neatly fit into the classical cytoarchitecture of Brodmann. The findings suggest that future investigations will refine our understanding of the anatomy of this complex visual region and help neuroscientists discover more about structure–functional relationships within the extrastriate visual cortex and in general.
The anatomical probability map of hOc5 is available at http://www.fz-juelich.de/ime/ime_start/ and http://www.bic.mni.mcgill.ca/. It is also part of the database of the International Consortium for Human Brain Mapping (http://www.loni.ucla.edu/ICBM/ and the surface-based atlas http://sumsdb.wustl.edu) and can easily be applied using a toolbox (Eickhoff and others 2005) developed for the application of cytoarchitectonic anatomical probability maps as an option integrated in the SPM software (http://www.fil.ion.ucl.ac.uk/spm). The toolbox can be downloaded from http://www.fz-juelich.de/ime/ime_start/.
This Human Brain Project/Neuroinformatics research is funded by the National Institute of Biomedical Imaging and Bioengineering, the National Institute of Neurological Disorders and Stroke, and the National Institute of Mental Health. Further support by the BMBF (Bundesministerium für Bildung und Forschung 01GO0104), Brain Imaging Center West (Bundesministerium für Bildung und Forschung 01GO0204), and the Helmholtz Gemeinschaft (VH-NG-012) is gratefully acknowledged. The authors thank Mrs Ursula Blohm for excellent histological assistance. Conflict of Interest: None declared.