Experimental and imaging studies in monkeys have outlined various long association fiber bundles within the temporoparietal region. In the present study the trajectory of the middle longitudinal fascicle (MdLF) has been delineated in 4 human subjects using diffusion tensor magnetic resonance imaging segmentation and tractography. The MdLF seems to extend from the inferior parietal lobule (IPL), specifically the angular gyrus, to the temporal pole remaining within the white matter of the superior temporal gyrus (STG). Comparison of the superior longitudinal fascicle II-arcuate fascicle (SLF II-AF) with the MdLF in the same subjects revealed that MdLF is located in a medial and caudal position relative to SLF II-AF and that it extends more rostrally. Given the location of MdLF between the IPL (angular gyrus) and the STG, it is suggested that MdLF could have a role in language and attention functions.
Several long association fiber pathways in humans have been described since the time of Burdach (1822). Dejerine (1895) further systematized these association pathways in a classic monograph. In recent years a number of investigators using imaging techniques such as diffusion tensor magnetic resonance imaging (DT-MRI) have confirmed and further detailed these pathways (Makris et al. 1997; Catani et al. 2002; Mori 2002). One of the long association pathways, namely the middle longitudinal fascicle (MdLF), has been described in experimental and imaging studies in the macaque monkey (Seltzer and Pandya 1984; Schmahmann et al. 2007). To our knowledge, no definitive information regarding the MdLF is available in humans. In the monkey, the MdLF has been shown to originate from the caudal part of the inferior parietal lobule (IPL) and to extend into the white matter of the superior temporal gyrus (STG). It connects parietal area PG-Opt, with the multimodal areas in the superior temporal sulcus and the adjacent part of the STG and the Sylvian fissure (Seltzer and Pandya 1984). In monkeys, the MdLF is distinct from other parietofrontal and temporofrontal long association fiber pathways such as the superior longitudinal fascicle (SLF) and the arcuate fascicle (AF), respectively (Seltzer and Pandya 1984; Schmahmann et al. 2007).
Although in the experimental animal the anatomical connections between the IPL and the STG and the cortex of the superior temporal sulcus have been elucidated satisfactorily, less information is available in humans. Possible indications of such connections have been made by Dejerine (1901) and by Makris (1999) based on histological observations, with additional suggestions by Nieuwenhuys (2007). Dejerine utilized gross dissection and myelin stain techniques to diagram a connection between the rostral part of the angular gyrus and the posterior part of the STG (Dejerine 1901). However, gross dissection and fixation techniques as well as myelin staining can be limited in delineating a fiber pathway with certainty. On the other hand, DT-MRI has proven successful in confirming fiber pathways and distinguishing them from each other (Catani et al. 2002; Makris et al. 2005). As such, this technique could prove particularly useful in determining whether a temporoparietal pathway can be extracted and delineated that may correspond to the MdLF in monkeys. Furthermore, DT-MRI can assess whether the MdLF is a distinct fiber pathway from the AF that is also purported to connect to the STG.
If the MdLF exists in the human brain as a distinct fiber pathway, it may play a significant role in higher cortical functions such as language and attention, given its putative location. In the present study, therefore, we have attempted to outline this pathway in 4 healthy human subjects using DT-MRI tractography and segmentation. Our results reveal that there is a fiber pathway within the core of the white matter of the STG equivalent to that of MdLF in monkeys. It is further shown that MdLF is quite distinct from other adjoining pathways such as the AF and the SLF.
We used MRI to quantify the stem portions of the MdLF and delineate their trajectories in 4 normal adult right-handed human subjects (2 males and 2 females), age range 23–33 years old. We combined 2 different DT-MRI–based techniques, specifically fiber tract segmentation and tractography and a T1-based technique for cortical parcellation of the human brain (Caviness et al. 1996).
MRI was performed using a Siemens Trio 3 Tesla imaging system. Scans included a T1-weighted acquisition with the following parameters: time echo (TE) = 3.3 ms, time repetition (TR) = 2530 ms, time inversion = 1100 ms, flip angle = 7°, slice thickness = 1.33 mm, 128 contiguous sagittal slices, acquisition matrix = 256 × 256, in-plane resolution = 1 × 1 mm2 (i.e., field of view [FOV] = 256 mm × 256 mm), 2 averages and pixel bandwidth = 200 Hz/pixel. A DT-MRI echo-planar based protocol was also acquired, and it included automatic magnetic field shimming and axial diffusion tensor imaging covering the entire brain (60 axial sections). We sampled the diffusion tensor, D, using a 7-shot echo-planar imaging (EPI) technique that samples the magnitude and orientation of the diffusion tensor. The following parameters were used: TR = 8.9 s, TE = 79 ms, averages = 10, number of slices = 60 in the axial orientation, slice thickness = 2 mm, no slice spacing, FOV = 256 mm × 256 mm (i.e., data matrix = 128 × 128, in-plane resolution = 2 mm2), diffusion sensitivity b = 600 s/mm2. The total imaging time for DT-MRI was 10 min.
The stem portion of the MdLF was segmented based upon a priori anatomical knowledge from the human (Makris 1999) and experimental animal literature (Seltzer and Pandya 1984; Schmahmann and Pandya 2006; Schmahmann et al. 2007). This was done through the T2-EPI principal eigenvector color-coded coronal sections within the STG (T1 or STG) white matter in the rostrocaudal dimension by selecting and labeling voxels pertaining to the stem portion of the MdLF. To determine whether the selected voxels were part of the stem portion of the MdLF 2 criteria were used: 1) the topography of the voxels, and 2) the orientation of diffusion properties of the tissue. In the coronal plane, the MdLF consisted of voxels with tensor information of anterior–posterior orientation within the white matter of the STG (Fig. 1).
Characterization of Region of Interest
We calculated the size (as number of voxels) and the mean and standard deviation (SD) of anisotropy as fractional anisotropy (FA) index for the MdLF region of interest (Basser and Pierpaoli 1996; Pierpaoli and Basser 1996). We also calculated left–right volumetric and FA symmetry based upon a symmetry coefficient (L − R)/[0.5(L + R)] (Galaburda et al. 1987). The trajectory of MdLF was defined in each individual subject using the “color map approach” (Makris et al. 1997; Makris, Pandya, et al. 2002; Mori 2002). Three-dimensional reconstructions were done based on the voxels manually selected on each coronal section (Fig. 2). Moreover, the trajectory of each individual MdLF (in 8 hemispheres) was determined in the Talairach coordinate system (Talairach and Tournoux 1988). This was performed for the group as well by computing the Talairach coordinate of the center of mass for the MdLF in each coronal slice in which it was observed (Fig. 3). Tkmedit was used for manual segmentation of the fiber tracts (Makris et al. 2007). To ensure reproducibility of the segmentation procedure used for MdLF, intrarater and inter-rater reliability measures were calculated for the MdLF and compared with those obtained for the corticospinal tract (Ellis et al. 2001; Pierpaoli et al. 2001) and the cingulum bundle (Makris, Pandya, et al. 2002), which are well-known and well-described fiber pathways in the human using the DT-MRI technique. This was measured as follows. Given the intersection of 2 labelings (L1 and L2) for a structure, the percent common voxel agreement is given by the volume (V) of the intersection as a ratio to the average volume of both labelings. That is, the percent volume overlap is provided by: PCVA = [V(L1 ∩ L2)/(V(L1) + V(L2))] × 200.
To delineate more completely the trajectory of the MdLF we performed tractographic analysis using the DTI Task Card software (Massachusetts General Hospital) (Makris et al. 2005). The algorithm used by the software is a streamlined method that grows fiber tracts by following the direction of the principle eigenvector at each step starting from a seed point (Mori et al. 1999; Lori et al. 2002). Once segmentation was accomplished for the MdLF stem, the entire segmented stem ROI was used as a seed for the tractographic delineation of this fiber bundle (Fig. 4A). The tractographic parameters of the seeds were FA = 0.25 and angular threshold = 23°. To confirm the results observed by this tractographic technique, we used an additional approach to delineate the trajectory of the MdLF, namely we used a seed derived from the manual segmentation in a single coronal section (i.e., coronal section –13 of Fig. 3D) within the STG white matter. The tractographic parameters of these seeds were also FA = 0.25 and angular threshold = 23°. The capability to create the MdLF by seeding a single coronal section as opposed to using the entire segmented stem of the fiber tract (which involves multiple consecutive coronals) adds credence to the idea that MdLF is a long corticocortical white matter fiber pathway within the STG and the angular gyrus rather than a sequence of adjacent short or medium range intragyral fibers. The use of both approaches, lends additional confidence to the present results.
Given that segmentation does not allow the delineation of fiber pathways beyond their stem portions (Makris et al. 1997, 2005; Makris, Papadimitriou, et al. 2002), tractography is necessary for a more complete delineation of their trajectories. Furthermore, tractography offers better visualization of the relative topographic relations between fiber tracts in the cerebral cortex (Makris et al. 2005, 2007).
It should also be noted that the SLF II-AF were generated in these 4 subjects (8 hemispheres) using DT-MRI tractography (Catani et al. 2005; Makris et al. 2005). The seed points were placed following the specifications by Catani et al. and their tractographic parameters were FA = 0.25 and angular threshold = 33° (Catani et al. 2005).
In this study, 1) we segmented the stem portion of the MdLF in 4 normal adult subjects and made volumetric and FA measurements (Table 1, Figs 1 and 2); 2) we mapped the MdLF stems in the Talairach coordinate space (Fig. 3); 3) we traced the trajectory of the MdLF fascicle using the tractographic technique (Fig. 4); 4) we followed the tractographic renderings into the cortical origin and termination of MdLF fibers using cortical parcellation results (Fig. 5); and 5) we differentiated the MdLF from adjacent fiber tracts such as the SLF II and AF by tracing them tractographically (Figs 6 and 7).
|Measure||Side||Subject 1||Subject 2||Subject 3||Subject 4||Group||Symmetry Index|
|Number of voxels||L||341||245||301||371||295.67||(48.22)||0.06|
|L + R||668||404||633||741||568.33||(145.47)|
|L + R||5.34||3.23||5.06||5.93||4.55||(1.15)|
|Measure||Side||Subject 1||Subject 2||Subject 3||Subject 4||Group||Symmetry Index|
|Number of voxels||L||341||245||301||371||295.67||(48.22)||0.06|
|L + R||668||404||633||741||568.33||(145.47)|
|L + R||5.34||3.23||5.06||5.93||4.55||(1.15)|
Note: L = left; R = right; Ave = (L + R)/2; SD = standard deviation. Symmetry Index: (L − R)/[0.5(L + R)]
In the present study, first we outlined the location and course of the stem portion of the MdLF using the DT-MRI segmentation method. As shown in Figure 1 (upper panel), in 4 representative rostrocaudal coronal sections taken at the levels depicted in the lateral view of the hemisphere, we were able to outline the location of the MdLF within the core of the STG. The lower panel in Figure 1 portrays the T2-weighted EPI sections corresponding to the DT-MRI of the upper panel. Figure 2 comprises 3D reconstructions of the rostrocaudal extent of the MdLF projected on a lateral view in the 8 hemispheres of the 4 subjects in the present study. Thus the extent of the major portion (stem) of the MdLF in the white matter of the STG (T1 or STG) is obtained. Although there was some variation in the course of the MdLF fibers, the basic profile of MdLF in 8 hemispheres was similar. Furthermore, the composite dataset of each individual was placed in the Talairach coordinate space as shown in Figure 3, in which the stem portion of the MdLF is represented in the 3 cardinal planes. The extent of the MdLF in each case was depicted in a color-coded fashion. Talairach coordinates are listed in the inset of Figure 3D representing the center of mass of MdLF in the coronal section within the STG for the 8 hemispheres of the 4 subjects studied. Talairach coordinates of the entire extent of the MdLF stems are portrayed in the axial and sagittal dimensions as well (Fig. 3). It is striking to see the resemblance of the extent and topography of MdLF in all subjects. Thus the utilization of the segmentation DT-MRI method provided and confirmed the general extent and location of MdLF (Figs 1, 2, 3).
We used the DT-MRI tractographic technique to further elucidate the trajectory of the MdLF. As shown in Figure 4 the trajectory of the MdLF in a parasagittal plane is similar in the 8 hemispheres (right and left in 4 subjects) and seems to extend from the temporal pole to the IPL (angular gyrus). This is the case whether we perform tractography using as seeds the entire segmented stem (Fig. 4A) or a seed in a single coronal section (Figs 3D and 4B). Figure 5 shows the trajectory of the MdLF in 3 planes (parasagittal, A; axial, B; and coronal C, rostral; D, middle; and E, caudal) and its connecting cortical regions. We also used tractography to differentiate the MdLF from other neighboring long fiber tracts. Thus as shown in Figure 6, the trajectory of MdLF (in green) is quite distinct from the SLF II-AF (in yellow). Figure 7 compares the trajectory of the MdLF with that of SLF II-AF in a representative case and demonstrates their respective connecting cortical areas, which were derived using the cortical parcellation technique by Caviness et al. (1996)Figure 7A indicates the trajectory of SLF II-AF and MdLF. It shows that MdLF courses more rostrally and caudally and is located medial to SLF II-AF. The rostral portion of SLF II-AF is located dorsal to MdLF, whereas the incurred segment of SLF II-AF (i.e., the AF) is located ventral to MdLF. Figure 7B depicts the connecting cortical areas of MdLF, which reaches the posterior portion of the angular gyrus caudally and extends up to the mid-anterior portion of the STG rostrally. Figure 7C,D, in addition to showing the rostrocaudal connections of MdLF, depicts the cortical connections of the AF subcomponent of the SLF II-AF in the caudal part of the superior and middle temporal gyri. Figure 7E shows parietal connections of the SLF II subcomponent of SLF II-AF (in the anterior part of the angular gyrus). Figure 7F is a frontal view depicting the trajectory of MdLF and SLF II-AF. Specifically, the MdLF is located medial to the SLF II-AF. MdLF runs within the core of the STG white matter from the temporal pole (Brodmann's area (BA) 38) to the caudal end of STG and then extends caudally to the angular gyrus (BA 39).
Quantitative Analyses of the MdLF
We also derived measurements of the volumes and average FA values as well as the symmetry coefficient for the stem of MdLF in 4 normal subjects, which are shown in detail in Table 1 for each individual subject as well as for the group. The group mean left/right value for the FA was 0.34, volume was 4.55 cm3, and the symmetry coefficient was 0.02 (rightward) for the FA and 0.06 (leftward) for volume, which did not reach statistical significance. Inter-rater reliability for MdLF was 85% and intrarater reliability was 86% for overlap. This was comparable to the reliabilities of 2 other pathways, the corticospinal tract and the cingulum bundle, which we tested for the purpose of comparison. Specifically, for the corticospinal tract, inter-rater reliability was 76% and intrarater reliability was 79% for overlap. For the cingulum bundle inter-rater reliability was 84% and intrarater reliability was 86% for overlap.
Experimental studies in the monkey have shown that apart from local and projection fibers, the core of the white matter of the STG contains a distinct long association fiber pathway. This pathway has been identified using anterograde isotope tracer injections in the IPL of the macaque monkey. It extends from the IPL up to the temporal pole and has been termed MdLF to distinguish it from the SLF II-AF and the inferior longitudinal fascicle (Seltzer and Pandya 1984). Recently, the MdLF has been delineated using diffusion imaging in a study comparing the results of isotope injections in monkeys with those of the diffusion spectrum imaging technique (Schmahmann et al. 2007). The experimental study also has shown that this pathway carries fibers to multimodal areas (TPO) in the cortex of the upper bank of the superior temporal sulcus as well as to auditory related areas in the Sylvian fissure and the STG. Thus it seems that the MdLF fibers connect the IPL (areas PG-Opt), with the superior temporal region. In the present study using DT-MRI segmentation and tractography, we are able to discern a long association fiber pathway coursing between the IPL (angular gyrus) and the STG in 4 human brains (8 hemispheres). Furthermore, we combined DT-MRI tractography with cortical parcellation to better delineate the MdLF in terms of the cortical areas it connects (Makris et al. 2005). Considering the limitation of the DT-MRI tractographic technique to allow precisely the tracking of the origins and terminations of fiber bundles (Makris et al. 1997; Mori et al. 1999) cortical parcellation presents the advantage to complement DT-MRI tractography in approximating the cortical regions of origin and termination of corticocortical association fiber tracts (Makris et al. 2005). Moreover, extrapolation from the experimental data in monkeys regarding the origins and terminations of fiber pathways provide additional information regarding the origin and terminations of a fiber bundle. Following this approach, and also guided by a description of MdLF derived from human histological material (Makris 1999), we observed that the MdLF fibers extend caudally up to the angular gyrus, whereas they reach rostrally the caudal, middle and rostral portions of the superior temporal cortices. Our results indicate that the MdLF is distinct from SLF II-AF fibers and that its fibers are located medial to the AF fibers (Figs 7 and 8). Additionally, the MdLF fibers extend more rostrally and caudally than the SLF II-AF fibers. Furthermore, the MdLF fibers are dorsally located within the STG, whereas the AF fibers remain in a more ventral position. Their differentiation, based on criteria of topography of the voxels, and the orientation of diffusion properties of the tissue pertaining to these fiber systems, elucidated the distinct topographic anatomy of MdLF and SLF II-AF in the context of clearly anatomically segmented and parcellated human T1-MRI datasets.
On the basis of the above described topographic location and trajectory of the MdLF, as well as the known functional role of the cortical areas that are interconnected by this fiber pathway, some general annotations regarding the possible functional role of this fiber system related to the language and attention functions can be made. The IPL is a heteromodal association area related to such cognitive functions as language and attention (Riddoch 1935; Brain 1941; Paterson and Zangwill 1944; McFie et al. 1950; Denny-Brown and Banker 1954; Sperry 1961; Geschwind and Kaplan 1962; Geschwind 1965a, 1965b, 1987; Critchley 1966; Heilman et al. 1970; Mesulam 1981; Posner et al. 1984; Caplan and Evans 1990; Caplan 1992; Goodglass and Wingfield 1993). Lesions in the left angular gyrus usually are associated with such type of language impairment as a naming deficit in humans (Geschwind and Kaplan 1962; Geschwind 1965a, 1965b; Caplan and Evans 1990; Caplan 1992; Goodglass and Wingfield 1993). Alternatively, damage in the right IPL (BA 39 and 40), is usually associated with severe impairment in spatial attention, referred to as hemi-inattention (Heilman and Van Den Abell 1979, 1980; Mesulam 1981; Posner et al. 1984), which is a principal symptom of the neglect syndrome (Heilman and Van Den Abell 1980).
Based on our observations, the angular gyrus is connected with language-related areas in the mid-portion of the STG (Wernicke's area) independently of the AF by the MdLF (Figs 7 and 8). This finding may have implications related to the language circuitry. Both the superior temporal plane and the angular gyrus have been discussed in numerous models of language processing, all of which would necessitate communication between these 2 key areas (Geschwind 1965a, 1965b; Dronkers et al. 2004; Hickok and Poeppel 2004; Catani et al. 2005). With this being the case, the MdLF could be a principal conduit of linguistic information between the angular and superior temporal gyri. This notion could be an addition or alternative to the schema of language, in which the angular gyrus (“Geschwind's territory”) would be connected with the STG (“Wernicke's territory”) via the AF as recently purported by Catani et al. (2005) using DT-MRI tractography (Fig. 8A).
It is worth noting that whereas in the dominant hemisphere the role of the MdLF may be related to language function, in the nondominant hemisphere the MdLF may have a role in attentional processing. The attention network in the cerebral cortex is within a distributed system including the angular (BA 39) and supramarginal (BA 40) gyri, the middle and superior lateral frontal gyri (BA 8, 9, and 46) and the cingulate gyrus (BA 23 and 24) (Critchley 1966; Heilman and Van Den Abell 1980; Heilman et al. 1983; Heilman and Valenstein 1985; Goldman-Rakic 1988; Heilman et al. 1970; Mesulam 1990, 1998; Posner and Petersen 1990; Cabeza and Nyberg 2000; Duncan and Owen 2000; Corbetta and Shulman 2002) and the cortices of the banks of the superior temporal sulcus are also part of the broader attention network both, in monkeys (Watson et al. 1994) and humans (Karnath et al. 2001). This notion has received further support based on physiological and anatomical lesion experimental studies in monkeys (Goldberg and Robinson 1977; Lynch 1980; Bushnell et al. 1981; Motter and Mountcastle 1981; Mountcastle et al. 1975, 1981; Robinson et al. 1978; Watson et al. 1994). Thus the MdLF, by virtue of interconnecting the angular gyrus with the STG could well play a role in attention processing.
We performed measurements of biophysical parameters for FA, and of volume and of symmetry for the stem portion of the MdLF. FA ranged from 0.32 to 0.36 on the left and from 0.32 to 0.39 on the right. The FA values for the stem of the MdLF cannot be compared with other studies given that currently these are not available in humans or in monkeys. The FA values, however, were similar to the FA reported in other studies of different fiber tracts in the normal human brain. Pierpaoli and Basser reported FA values of 0.46 in subcortical white matter (Pierpaoli and Basser 1996). Klingberg et al. (2000) showed mean unscaled FA values in temporoparietal white matter that ranged from 0.38 to 0.59. In previous studies of the cingulum bundle and the SLF we have shown that FA values of the CB ranged from 0.45 to 0.54 (Makris, Pandya, et al. 2002), the overall FA for the SLF (all 4 subcomponents combined) was 0.44, and that SLF I in particular showed FA of 0.36 to 0.37 (Makris et al. 2005). Thus, it seems that our results of FA for the stem of the MdLF are in general agreement with data on other fiber pathways.
The overall volume of the MdLF stem portion was 4.55 cm3. Compared with the stem of the OFF, which is 1.81 cm3 (Makris et al. 2007), the MdLF is approximately 2 and a half times bigger. It is also almost 3 times smaller than the stem of the cingulum bundle, which is 12.06 cm3 (Makris, Pandya, et al. 2002). Although the MdLF is considerably smaller than the total SLF stem (all 4 subcomponents combined), which is 32.21 cm3, when compared with the individual subcomponents of the SLF, the MdLF is approximately equal to both SLF III, which measures 4.29 cm3, and the vertical component of the AF, which is 4.34 cm3 (Makris et al. 2005). The SLF II is 18.78 cm3 in size, approximately 4 times larger than the MdLF. Furthermore, these values should be compared with measurements derived from topographically constrained methods such as “white matter parcellation” (Makris et al. 1999). As measured in 20 healthy adult subjects using T1-weighted MRI, the white matter of the STG has been estimated to be 5.91 cm3. Moreover, the MdLF corresponds to 1.1% of total cerebral white matter, which was estimated to be 432.47 cm3 (Makris et al. 1999). Given the small number of subjects in this report, however, care should be taken in the interpretation of the population variance of these volumetric observations. The stem of the MdLF showed leftward volumetric asymmetry, which was not statistically significant. Given that the MdLF may play a prominent role in attention processing on the right hemisphere and the low number of subjects studied in our investigation, based on this leftward, however, not significant asymmetry we cannot ascertain whether there is a structural lateralization related to language function for the MdLF. Future studies using a larger number of subjects may bear insight in this interesting issue.
The angular gyrus and its connections with the STG are key components of the language system in the dominant hemisphere (Dejerine 1895, vol. II, p. 247, Fig. 248; Geschwind and Galaburda 1987; Catani et al. 2005). In the present study we delineated the MdLF as a major fiber pathway between the angular gyrus and superior temporal region. We suggested that this ipsilateral long association fiber tract could be an important conduit of linguistic information between these 2 cortical centers. Future studies using specific language and attention paradigms, implementing a combination of functional imaging techniques (such as functional MRI and/or magnetoencephalography) as well as structural MRI (such as T1-MRI and DT-MRI) may allow further the elucidation of the language circuitry as well as the role of MdLF in linguistic and attention processing.
In conclusion, in this study we have delineated and measured the MdLF in the human brain. The present study is the first to delineate and quantify this fiber pathway in humans in vivo. According to our observations the MdLF is a long association fiber pathway that courses within the entire STG and the IPL. The MdLF is distinct from other adjacent fiber tracts such as the SLF II and the AF. It is a mid-size fiber tract, interconnecting the angular gyrus with the superior temporal cortical regions, which emphasizes its centrality to such associative or higher brain functions as language and attention. The elucidation of its anatomy is relevant to correlate structural and functional parameters more precisely in cognitive and clinical studies.
National Institutes of Health National Center for Complementary and Alternative Medicine (PO1AT002048-05) to N.M.; and Fairway Trust to D.K.
We gratefully acknowledge Dr Verne Caviness, Dr Edward H. Yeterian, Dr Larry Seidman, John Bruyere, Steve Hodge, and Dr Andre van der Kouwe for their valuable contributions to the preparation of this manuscript. We also thank the anonymous reviewers for providing useful comments on the manuscript. Conflict of Interest: None declared.