Heschl gyrus and its included primary auditory cortex: Structural MRI studies in healthy and diseased subjects

Authors

  • Ihssan A. Abdul-Kareem MD, MSc,

    Corresponding author
    1. School of Health Sciences, Division of Medical Imaging and Radiotherapy, University of Liverpool, Magnetic Resonance and Image Analysis Research Centre, University of Liverpool, Liverpool, UK
    • School of Health Sciences, Room 1.09, Whelan Building, The Quadrangle, Brownlow Hill, Liverpool L69 3GB, United Kingdom
    Search for more papers by this author
  • Vanessa Sluming TDCR, MSc, PhD

    1. School of Health Sciences, Division of Medical Imaging and Radiotherapy, University of Liverpool, Magnetic Resonance and Image Analysis Research Centre, University of Liverpool, Liverpool, UK
    Search for more papers by this author

Abstract

Despite the fact that the Heschl gyrus (HG) is a crucial brain structure as it contains the primary auditory cortex (PAC), relatively few structural MRI studies have concentrated upon it. We propose that this may be attributed in part to the considerable variability of this structure and, most importantly, to the lack of unified criteria for defining the extent of the PAC along the MRI-determined landmarks of the HG, which ultimately affects the reliability and reproducibility of these studies. This review highlights three aspects: first, the standard and variant anatomy of the HG and PAC with particular focus on MRI definition of these regions; second, the importance of studying the HG and PAC in health and disease using structural MRI; and, third, the problem of MRI localization of the PAC. The scientific community should be aware that the HG and its included PAC are not synonyms. Additionally, owing to the great complexity and variability of these regions, future MRI studies should be cautious when using single brain-based atlas or maps generated by simply averaging across individuals to localize these regions. Instead, and while waiting for future in vivo microstructural localization of the PAC, the use of probabilistic and functional maps is advantageous but not without shortcomings. J. Magn. Reson. Imaging 2008;28:287–299. © 2008 Wiley-Liss, Inc.

THE HESCHL GYRUS (HG) HAS EXCITED neuroscientists who study it as it contains a very crucial region, the primary auditory cortex (PAC). Tremendous interhemispheric and interindividual variations have been reported in this gyrus by both postmortem (1, 2) and MRI studies (3, 4). Adding more complexity is the finding that the extent of the cytoarchitectonic PAC along the HG are also subjected to interhemispheric and interindividual variations (2, 5). These facts would greatly question both the validity of structural MRI studies concentrating on these regions and the use of single brain-based atlas in localizing them. This is especially true for the PAC as it, unlike the HG, lack MR-defined macroanatomical landmarks. This may, in part, explain why the HG (and its included PAC) has not been studied as extensively as other regions of the superior temporal gyrus (STG) such as the planum temporale (PT). Consequently, these difficulties will also affect interpretation of functional MRI findings. Therefore, the need for adequate knowledge of the standard and variant anatomy of these regions and particularly the PAC seems to be of extreme importance. Equally important is to correlate these anatomical variations with the diverse results obtained by previously published structural MRI studies of these regions and to suggest unified definitions and criteria that future studies can make reference to. To our knowledge, these issues have not been addressed previously.

The review is divided into three main parts: in the first, the standard and variant anatomy of the HG and its included PAC is discussed, paying particular attention to MRI definition of these regions. This is followed by highlighting the previously published morphometric/volumetric studies of these regions in health and disease; and, finally, the problem of in vivo localization of the highly-variable PAC is discussed together with suggested MRI-based methods to localize it. The review concludes by making recommendations for future MRI studies of these regions.

Anatomy of the HG

The HG lies diagonally (transversely) across the superior temporal plane (STP) in the depth of the sylvian (lateral) fissure; separating the STP into rostral and caudal parts, the planum polare (PP) and the PT, respectively (5–7). On MRI images (Figs. 1 and 2), it has been suggested that the HG can be readily identified because of its characteristic shape and location (3, 8, 9). On transverse sections, the HG appears to be a gyrus (or gyri) transversing anterolaterally from a point posterior to the insula (which is surrounded by a deep circular sulcus that separates it from the frontal, parietal, and temporal lobes) to the convexity (9, 10). In a landmark study, Penhune et al (3) described the boundaries of the HG comprehensively in T1-weighted MRI images (Fig. 1): the HG is bounded anteriorly by the first transverse sulcus (TS) that unites medially with the circular sulcus of insula (Fig. 2a) and posteriorly by the Heschl sulcus (HS) or sulcus intermedius (SI) (the incidence of which was reported to be 41% in postmortem studies (2)) when it extends at least half the length of the HG (1, 3). The posteromedial boundary is a line drawn from the medial end of the TS to the medial end of the HS. The anterolateral boundary is defined either by the visible ending of the gyrus or by extending the lines defined by the TS and HS to the lateral border of the temporal plane (3). On sagittal or coronal sections, the HG appears as a protrusion on the STP, with variable shapes, including a single Ω (9), a mushroom (11), a heart (4, 9), a bow-tie (4), or two separate esses (Ss), depending on the number of HGs and on the absence or presence, depth, and length of the SI (9). It is bounded inferiorly by a line drawn from the depth of the HS to the notch created by the meeting of the superior surface of the HG and its stem (3).

Figure 1.

The boundaries of the HG. a: Axial T1-weighted MRI section showing a left single and a right CPD of the HG (CPD). b: Coronal section. c,d: Sagittal sections of the left and right HG, respectively. 1 = first HG; 2 = second HG; 3 = first TS; 4 = HS; 5 = second HS; 6 = insula; and 7 = PT. The left HG is bounded by the first TS (3) anteriorly and the HS (4) posteriorly. The posteromedial boundary is (8), the anterolateral boundary is the visible termination of the gyrus in this case and the inferior boundary is (9) (see text for details). Note that in case of CPD, the boundaries are made for the first HG as it contains the PAC; the second HG is considered to be part of the PT.

Figure 2.

CSD of HG. a: Axial T1-weighted MRI section showing a right CSD of the HG. b: Sagittal section of the right HG. c: Coronal section. 1 = first HG; 2 = second HG; 3 = SI; 4 = first TS; 5 = HS; 6 = insula; 7 = PT; and 8 = circular sulcus of insula. Note that in this type of duplication, the posterior boundary of the first HG is the SI. In coronal and sagittal sections, it is evident that this sulcus does not reach the base of the HG (compare with Fig. 1). Note also the bifurcated HS on the left side.

In clinical practice, the most widely used MR weighting is the transverse relaxation or T2-weighting, as it reveals pathology to advantage. In certain pathological cases, the characteristic morphology of the HG would be blurred out, and hence, there is great demand to localize the HG using T2 signal intensity properties. To address this, Yoshiura et al (8) examined 60 hemispheres and reported that the HG is first hypointense to the middle temporal gyrus (MTG), is either iso- or hypointense to the STG, and is isointense to the white matter in 55% of hemispheres in coronal T2-weighted MR images. These results were confirmed in cadaveric T2-weighted images of the temporal lobe in the same study. In agreement, these findings were replicated in another—objective—study that measured T2 relaxation time and reported that the HG has T2 measurement lower than both the MTG and STG (12). Accordingly, beside its characteristic orientation and shape, it is possible to differentiate the HG from other temporal structures based on its intrinsic T2 tissue properties, which could be useful when examining pathologically-disturbed MRI images.

Anatomy of the PAC

PAC is a central core of koniocortex that occupies the HG and is surrounded by several (non-PAC) belts of parakoniocortex rostrally, laterally, and caudally; it is separated from the insula medially by a narrow strip of prokoniocortex (13). Histologically, PAC can be distinguished from the surrounding areas because of its dense granular cytoarchitecture, relatively small pyramidal cells in layer III, dense myelination, a prominent width of layer IV, and a relative cell-sparse layer V, as well as its histochemical characteristics (1, 2, 5, 13–15). Depending on histological criteria, the PAC has been classified as one area by Brodmann (termed BA 41) (16) surrounded by areas 42 and 22 laterally and area 52 medially. Other histological definitions include two areas: (TC and TD) (17) and (KAm and Kalt) (13) or three areas: (Te1.0, Te1.1, and Te1.2) (5) extending along the mediolateral axes of the HG.

Functionally, the PAC contains neurons with receptive fields sharply tuned to the spectral frequency, which is represented in mirror-symmetric tonotopic maps (18, 19). The PAC is crucial for initial processing of auditory stimuli (20) and is hypothesized to detect and extract basic acoustic signal features, such as frequency, lateralization, onset time synchronicity, modulation coherence, loudness, and harmonicism (21), to be used in other auditory areas as clues for the construction of separate object streams (22).

Topography of the PAC Along the HG

The fact that the PAC is largely confined to the HG was derived from a variety of studies; including: direct recordings of primary evoked responses (23), cytoarchitectonic (1, 2, 5, 13–15), myeloarchitectonic (1, 14), histochemical (14, 24, 25), immunocytochemical (15, 26), structural (27), and functional MRI studies (18, 21). Within the HG, many of these studies described the PAC as occupying largely either the posteromedial portion or the medial two-thirds while non-PAC areas are located more laterally (1, 2, 15, 18, 23–25, 28). Two cytoarchitectonic studies, on the other hand, included cortex on the anterolateral HG within the PAC (5, 13). Generally speaking, when there are two or more transverse gyri, the PAC tends to occupy the first one (1, 2, 5, 14).

Few structural MRI studies have attempted to localize the PAC in vivo. It was shown that the T2 signal intensity varies between the anterior and posterior halves of the first HG; T2 hypointensity is especially prominent in the posterior half of the first HG (8) and one of the suggestions given to this finding is that it is due to the presence of the PAC in the posterior half of the HG (8). The authors hypothesized that this difference in signal intensity is mainly due to the different cytoarchitecture between the PAC in the posterior half and non-PAC areas in the anterior half of the HG. However, a study that examined the visual cortex using in vivo T1 and postmortem T2 MRI hypothesized that the lamination pattern observed in the cerebral cortex in MRI images originates mainly from myeloarchitecture with a weaker influence of cytoarchitecture (29). In accordance, it can be argued that the intrinsic T2 properties of the PAC, being a primary sensory region with histological criteria similar to the visual cortex, are mainly derived from its myeloarchitecture rather than its cytoarchitecture.

In a recent study that examined only five subjects (27), the heavy myelination of the PAC, as compared to adjacent auditory areas, has claimed to localize the PAC in vivo: it was shown that regions of high longitudinal relaxation rate (R1) (a property related to myelin content), were always found overlapping the medial two-thirds of the HG, which was reminiscent of the koniocortex of the PAC.

Taken together, it could be suggested that it is the dense myeloarchitecture of the PAC that is mainly responsible for differentiating it from other non-PAC areas within the HG based on intrinsic tissue properties (transverse and longitudinal relaxation rates). These studies can offer a promising lead in the future to properly localize the PAC; however, they require further support in particular to covary these regions with high R1 or low T2 signal intensity with the variable location of the PAC along the HG; i.e., to rule out whether these regions correlate with the cytoarchitectonic PAC extent along the HG or they merely follow a fixed distribution in the medial two-thirds (27) or the posterior half of the HG (8).

Apart from studying auditory cortical structure, the connectivity patterns of PAC were also suggested to be important in defining its extent. Studies in nonhuman primates revealed that the PAC has unique contralateral and ipsilateral connectivity patterns in comparison with primary visual and somatosensory fields (30). Upadhyay et al (19) were the first to characterize and quantify structural ipsilateral connectivity within the human PAC in vivo using diffusion tensor imaging (DTI) and functional MRI (fMRI) (19). Following the localization of the PAC functionally (two high- and one low-frequency regions), DTI revealed two sets of fibers within it: high-frequency fibers connecting the high-frequency regions and low-high–frequency fibers connecting the low-frequency region with both high-frequency regions. These fibers are organized in a caudomedial–rostrolateral orientation along the HG. The low-high–frequency fibers were found to possess a lower probability, higher total number, and shorter length than the high-frequency fibers. It was hypothesized that this abundance of the shorter low-high–frequency fibers suggests that the majority of projections within the human PAC remain intrinsic to that auditory region.

Anatomical Variations of the HG and Its Included PAC

Both the HG and its included PAC are subjected to great interhemispheric and intersubject variability as evident from previous MRI and postmortem studies. These variations have significance for understanding the results of MRI studies conducted on these regions. In the following subchapters, previously published anatomical variations of the HG and PAC in healthy subjects will be summarized.

Variations in the Number of HG

A single transverse gyrus per hemisphere is the most frequent finding in both MRI and postmortem studies with a reported frequency range of 70% (2) to 75% (8). The reported frequency range for two transverse gyri was 24% (2) to 33% (9). Three gyri were reported per 6% of hemispheres (2) and up to five gyri per hemisphere were also reported (31). Further, previous studies described variable combinations of single and double gyri in both hemispheres of the same brain (1, 17, 31, 32). The most frequent combination reported in MRI studies was a single gyrus in each hemisphere as described by Penhune et al (3) (70% in 20 subjects) and Leonard et al (4) (70% in 105 subjects when measures were taken at X = 34). These data clearly disclose that there are wide interhemispheric and interindividual variations regarding the number of HGs. This could be partly due to various methods of examination (postmortem and MRI) and subjective observations but also due to the complexity of the HG and its sulcal landmarks.

For duplicated HG, two patterns have been described: in the first, the HG is duplicated by SI indenting the crowns of an initially single gyrus (common stem duplication [CSD]) (Figs. 2 and 3). In the second pattern, there is a complete posterior duplication (CPD) (Figs. 1 and 3) that is bounded by a second HS (H2S) (1, 4). The morphology of these two duplications is different: in CSD, the SI does not extend into the base of the HG initially and the two divisions remains connected, with a heart-shaped appearance. A CPD, by contrast, forms a fully separate structure shaped like a rectangle or a square (4). Leonard et al (4) reported that the frequency of HG duplication is unstable; it ranged from 20% to 60% depending on distance from the midline. They found that CSDs were more frequent than CPDs, particularly in the right hemisphere. For the CPD variant, the cytoarchitectonic PAC was reported to be limited to the first transverse gyrus (7) and, consequently, the second gyrus was considered to be part of the PT and not included in HG measurements (1, 2, 33, 34). However, for the CSD variant, the demarcation is less clear; previous studies have reported that the PAC can be either limited to the first gyrus (1, 2) or to span both gyri (2, 14, 15).

Figure 3.

Simplified drawings of the right HG and its duplications adapted from axial MRI sections. a: Single HG. b: CSD. c: CPD. Yellow area = single HG or first HG; green area = second HG; red line = first TS; dark blue line = HS; light blue line = SI; and pink line = second HS.

HG duplications have been proven to be an important finding in clinical studies. Increased frequency of HG duplication was reported in populations with learning disabilities (35, 36) and in males with resistance to thyroid hormone (37). Leonard et al (38, 39) described an anatomical risk factor index consisting of seven brain measures including the presence of duplicated left HG and the size of a single left HG to separate subjects with phonologically based learning disability (impairment in phonological decoding; the automatic cross-modal substitution of phonemes for graphemes) from those with non-phonologically-based learning disability (problems with language comprehension and/or production rather than phonological decoding). They also stated that this anatomical index can predict reading skill in normal children (study 3) (38). HG duplications can, therefore, be used as markers for learning ability and could differentiate between the different types of learning disabilities, which necessitates detailed understanding of the morphology of such duplications in clinical studies. An important methodological point of view is that as the landmarks that defines duplications are not constant but increase with distance from the midline in stereotaxic space, restricting the definition of HG duplications to medial slices where the frequency is low in normal adults would exaggerate the differences between normal and patient populations (4). Since the presence of HG duplications runs in families (37), investigation of its heritability was suggested to be important in providing an informative phenotypic marker for genetic studies of dyslexia and other learning disorders (35).

Interhemispheric Morphometric Variations

Most previous studies have reported volume (3, 40–42), surface area (43), and length (32) asymmetry favoring the left HG. Penhune et al (3) reported that this larger left HG volume was attributed to larger white matter (WM) but not gray matter (GM) (3). Also, it was shown by R1 mapping that there is leftward GM myelination asymmetry in the HG (27). Cytoarchitectonically, several studies reported leftward PAC asymmetry: the left PAC has been shown to be larger (1, 5) (although a larger study revealed that there is no clear pattern, with leftward asymmetry in 44% and rightward asymmetry in 26%) (2) and its cell columns to be both wider and more widely spaced than those of the right PAC, and individual columns on the left were contracted by more afferents than comparable cell columns on the right (44–46). Hutsler and Gazzaniga (47) reported leftward asymmetry in the size of layer III pyramidal cells in the PAC and other auditory areas. These cytoarchitectonic findings have been further supported by fMRI data showing leftward asymmetry in the PAC (10, 48).

This asymmetry of the HG (and its included PAC) favoring the left side can be linked to the left hemispheric lateralization of language and a preferential role for the left PAC in processing temporal aspects of speech (3, 49). It was hypothesized that the left HG is crucial in language learning following the finding of higher WM density and volume of the left HG in faster compared with slower phonetic learners of foreign speech sounds (50). Similarly, native English-speaking adult subjects who learned to incorporate foreign pitch patterns in word identification revealed that successful learners have larger left HG volume (especially GM volume) relative to less successful learners (20). These two studies pointed to the importance of neuroanatomic difference of the left HG volume (and its included PAC) in predicting foreign language learning. Taken together, it can be concluded that, in addition to the PT, the HG also serves as a marker of hemispheric dominance for language.

In contrast to studies reporting leftward HG and PAC asymmetry, some studies revealed no interhemispheric asymmetries in the surface area of most anterior HG (49) or in the area of PAC (13, 17) and a study reported rightward HG asymmetry (31). This inconsistency could be due to variable definitions or measurement methods of these regions. For instance, Campain and Minkler (31) found multiple gyri more commonly on the right, which, in effect, would lead to the biased finding of rightward HG asymmetry. Also, Kulynych et al (49) used MR planimetry to measure the HG surface area, which, in contrast to volumetry, would underestimate a large proportion of hidden cortex and may lead to their finding of no asymmetries.

Gender Variations

Although most studies that examined gender differences in the HG have revealed no differences (3, 4, 51–53), some studies revealed that males exhibited larger right HG GM volume than females but no gender differences were exhibited in the size of the left HG (11, 54). This resulted in rightward asymmetry in males and leftward asymmetry of HG GM volume in females (11). Conversely, a voxel-based morphometric (VBM) study revealed increased leftward asymmetry of HG GM in males compared to females (55). Gender differences were also demonstrated in the PAC volume: the PAC was proved to be larger in females than in males bilaterally; inverse asymmetry toward the right side was more frequent in females; however, laminar cell volume densities of PAC showed no gender effect (56). These data suggest gender differences in the cerebral organization of the PAC (56). From the above, it could be suggested that future studies should be careful when recruiting subjects belonging to one gender (15) as generalization to the whole population would not be appropriate.

Variations in Position in Three-Dimensional Space and the PAC Extent Along the HG

There is great interhemispheric and intersubject variability in the sulcal landmarks of the HG (3, 4). Penhune et al (3) created a probabilistic map defining the location of PAC based on MRI examination of 20 subjects in stereotaxic space. However, they considered the whole HG (the first gyrus if there is more than one) as equal to PAC, which, according to cytoarchitectonic findings (2), would overestimate the PAC. Therefore, the probabilistic map mentioned in their study could be argued as accurately representing the HG rather than the PAC. Extending and partially replicating this study, Leonard et al (4) examined MRI images of 105 subjects and described the normal variation in location and frequency of major and minor sulci associated with HG in standardized three-dimensional (3D) space. Both these studies stated that there is a significant range in the normalized location of HG boundaries; in addition, there were consistent left-right asymmetries in landmark position. As a result, the spatial layout of HG (and its included PAC) is more anteriorly located in the right hemispheres relative to the left (3, 4). In the same way, cytoarchitectonic studies stated that the left and right PAC are not mirror images and that the right PAC is located anterolaterally in relation to the left in stereotaxic space (2, 5). Consequently, these instabilities in location of HG and PAC in stereotactic space would adversely affect the use of a single-brain-based atlas like the Talairach and Tornoux in locating these areas, as coordinates may not correspond to the same anatomical structure in each participant (for further discussion, see Ref.57), which is particularly true for the HG and PAC. In functional studies, such left-right asymmetry would create serious outcomes if attempts are made to localize the data symmetrically. Doing so will, for instance, localize an area of activation at a site outside the HG in the right hemisphere and, as a result, lead to the interpretation that such activation is lateralized to the left side.

In addition to variable location in stereotaxic space, there are great interhemispheric and intersubject variations regarding the extent and relation of PAC to anatomic landmarks of HG (1, 2, 5, 14), which add more complexity to this region. The PAC may occupy a variable percentage of HG volume from 16% to 92% and it may even extend into the PP or PT (2) or the superior temporal cortex (5) to varying degrees. The smaller the distance to the transverse sulci, the greater the probability of finding non-PAC areas (2). Consequently, inaccurate results will occur when the cytoarchitectonic borders of PAC are estimated from MR-defined macroanatomical landmarks of HG (2). The crucial piece of knowledge is, therefore, that the volume of the cytoarchitectonically-defined PAC does not correlate with the volume of HG and it was hypothesized that when the volume of HG is taken as equal to the volume of PAC, an error as large as 100% can be introduced in about 40% of hemispheres (2).

MR Volumetric/Morphometric Studies of HG (and Its Included PAC) in Health and Disease

In the following subchapters, we highlight previously published MR volumetric/morphometric studies of HG (and its included PAC) in four groups: musicians, deaf, stutterers, and schizophrenia patients. Apart from stressing the importance of studying these crucial structures in health and disease, the challenging effect of interhemispheric and interindividual variability of HG and PAC will be addressed.

Musicians

Several structural MRI studies have been concentrated on musicians, looking for structural neuroplastic changes that represent use-dependent adaptation in response to long-term musical training and skill acquisition. For instance, musicians were reported to have increased corpus callosum size (58), increased GM density in Broca's area (59), increased cerebellar volume (60) compared to nonmusicians, and musicians with absolute pitch revealed increased left-sided asymmetry of the PT as compared to musicians lacking this ability (61). However, as compared to the PT, few studies elucidate the importance of the HG as a marker of structural brain reorganization in response to skill acquisition (musical training). In this context, Schneider et al (62) reported that GM volume of the anteromedial portion of the HG is 130% larger in professional musicians than in nonmusicians (62). These results were replicated in another study that used an objective MRI technique (VBM) in which a positive correlation (highest in professional, intermediate in amateur musicians, and lowest in nonmusicians) was found between GM volume in both HGs and the musician status (63). Compared with fundamental pitch (f0) listeners, spectral pitch (fSP) listeners possessed a pronounced rightward, rather than leftward asymmetry of GM volume within the lateral HG (which is sensitive to rapid temporal processing (64)). This lead to the conclusion that the relative hemispheric lateralization of the lateral HG size reflects the type of pitch processing irrespective of musical aptitude, whereas the absolute size of the HG depends on musical expertise (33). This asymmetry of the lateral HG and the related perceptual mode were suggested to have an impact on the preference of timbre, tone, and size of particular musical instruments and in particular on musical performance (65). Accordingly, the perceptual mode may be predicted by MR volumetry of the HG (65).

Schneider et al (33) defined objective criteria based on their own findings on lateralization, prior structural knowledge of PAC extent, and functional separation of adjacent pitch-sensitive areas to delineate the boundaries of the first HG and PAC in musicians and controls. They created a grand-average auditory cortex map of 87 subjects in stereotaxic space and suggested that there is a symmetric organization in all subjects with respect to angulation, extent, and transition from the HG to the anterior STG and that the medial two-thirds of the anterior HG (defined by a line perpendicular at the mediolateral two-thirds of the HG) is a reliable approximation of PAC extent. The fact that this map is based on a relatively large sample of subjects and the use of both structural and functional criteria for its generation deserve attention by future studies; however, using such a map would oversimplify a complex structure like the HG and would neither correct for the great intersubject variability in PAC extent along the HG nor it would account for the asymmetrical interhemispheric spatial layout of the PAC.

Congenitally Deaf vs. Hearing subjects

It had been shown that congenitally deaf and hearing subjects did not differ in the total or GM volume of the HG (40, 41), suggesting that auditory deafferentation does not lead to cell loss within the PAC in humans. Considering the finding in musicians of an increase in the HG GM volume in response to prolonged and extensive auditory (musical) practice, one would expect that when there are no auditory stimuli at all, then there will be a reduction in the HG GM volume according to the hypothesis of “use it or lose it.” The authors hypothesized that this lack of HG GM volume reduction can be attributed to the fact that neurons within the HG of deaf may have been responding to nonauditory inputs, as it was shown that tactile (66) or visual stimuli (67) can stimulate the auditory cortex in deaf subjects.

On the other hand, deafness would result in less WM in the HG, suggesting that auditory deprivation from birth results in less myelination and/or fewer fibers projecting to and from the auditory cortices (41). Most importantly, it was shown that the leftward asymmetries for the volumes of the HG and PT were preserved in deaf subjects (40, 41); and it was consequently hypothesized that there is a genetic component (40) in the development and maintenance of auditory cortical asymmetries. These latter findings would contradict the conventional belief that these asymmetries depend on auditory language experience (3). In another way, such asymmetries could be thought of as the cause rather than the result of language learning and acquisition.

Stutterers

Stuttering is characterized by involuntary, audible or silent, repetitions or prolongations of sounds or syllables that usually evolves before puberty without evident cause (68). Using VBM, stutterers revealed increased GM density mainly in the right PAC (but also in the left) extending into BA22 and BA21 as compared to nonstutterers (69) indicating that structural differences in stutterers not only include the PT (70, 71), as previously shown, but also extend to the PAC. These structural differences were suggested to indicate that atypical processes in this cortical region critically differentiate between stutterers and nonstutterers (69). Another VBM study revealed that stutterers have increases of WM volume in the right auditory cortex (HG and PT) resulting in the lack of the typical leftward WM asymmetry in those subjects (72). Using DTI, Chang et al (73) failed to replicate these findings, which could be partly explained by the suggestion that the VBM technique has reduced sensitivity in detecting WM differences (63) but also to the variable definitions of HG and PAC in these studies. An important methodological point regarding the VBM studies is that authors should declare whether they used probabilistic maps or whether they merely used the Talairach and Tornoux atlas in localizing brain regions; a point of extreme importance when studying highly variable regions such as the PAC. It is well known that many of such probabilistic maps are already installed as options within the software that is used for VBM data analysis (such as Statistical Parametric Mapping [SPM]; www.fil.ion.ucl.ac.uk/spm). Doing this would make it possible for the audience to validate the results obtained from these studies.

Schizophrenia Patients

Schizophrenia is a complex psychotic disorder with a wide range of symptoms including delusions, hallucinations, formal thought disorder, altered affect, and cognitive functioning (74). Using MRI, structural abnormalities in different portions of the left STG have been reported in schizophrenic patients. Fewer MR volumetric/morphometric studies were concentrated on the HG as compared to the PT in those patients (34, 42, 75, 76). VBM studies of schizophrenic patients revealed areas of GM volume reduction in the left STG including the HG as compared to controls (77). Patients with schizophrenia demonstrated bilateral HG and left PT GM volume reduction (75), and small total HG volume (78) compared with controls and other psychotic patients. The volume reduction was proved to be progressive in the GM of the left HG (34, 76) and the left PT (34) in those patients during the 1.5-year period between the scans. Consequently, it was concluded that schizophrenia is characterized by a postonset progression of GM volume loss in these regions (34). The severity of auditory hallucinations (74, 79) and delusions (79) in these patients was found to be significantly correlated with the volume loss in the left HG (74). These results addressed the importance of studying the HG to mark both the progression and the severity of disease in schizophrenic patients. This HG involvement in schizophrenic patients (42) and also in patients with schizotypal personality disorder (80) was hypothesized to be gender-specific, with only males affected.

The complex structure of the HG might in part be responsible for the diversity of results obtained in schizophrenia. For instance, although some structural MRI studies reported HG volume reduction (34, 75), others did not (81–84). This latter finding was even supported by postmortem pathological studies (85). Additionally, the presence of HG duplication created a significant diversity in measuring the HG volume: although most studies agreed not to include the CPD of the HG as part of the HG (34, 84, 86), a controversy existed, however, for including the volume of the CSD. While some studies included the whole CSD in measuring the HG volume (34, 75, 84, 86), others included only the common WM stem plus the anterior transverse gyrus in this measurement and discounted the posterior one (6, 42). The need for a unified definition of HG duplications and their measurements is, as a result, crucial for enhancing the comparability of these studies.

The use of templates derived from healthy controls to study patients is highly criticized. Instead, some authors suggested the use of probability maps constructed from patients (87, 88). Park et al (88) created a probability atlas based on mapping 16 regions of interest (ROIs) in 11 to 28 schizophrenic patients. In this map, schizophrenic patients compared with controls showed a significantly lower overlap probability in the dorsoposterior regions of HG and PT in both hemispheres, suggesting a greater heterogeneity in the spatial distribution of these regions. These results indicate that future schizophrenia studies should be cautious in normalizing data to a stereotaxic coordinate system derived from healthy controls. The use of such a map would have the added advantage over single-ROI mapping, showing the relative distribution of (ROIs) in 3D space in relation to neighboring regions (88).

MR Localization of the PAC: The Great Challenge

The already-mentioned interhemispheric and intersubject variations of the HG and of the extent of the PAC along this gyrus clearly show that studying the PAC is not clear-cut. It is important not to consider the PAC and the HG as synonymous. PAC volume does not correlate with HG volume and an overestimation of PAC volume is likely to occur when it is simply estimated by measuring the HG volume (3, 40, 41). The extremely variable extent and spatial layout of the PAC imply that the use of an atlas with a single-brain–based template like the Talairach and Tournoux to locate this region would lead to serious outcomes.

Morosan et al (5) used an observer-independent cytoarchitectonic method for the reliable definition of the PAC and its areal borders, which, in contrast to visually-based classic architectonic studies, used quantitative data and multivariate statistics to provide reproducibility, reliability, and precision of the results. They mapped the position and extent of three cytoarchitectonically-defined primary auditory areas, Te1.2, Te1.0, and Te1.1 (in 10 brains), to a spatial reference system based on a brain registered by in vivo MRI. However, the use of this map or other maps (33) that simply average MRI images of subjects into a static template would not correct the existing intersubject variabilities in the absolute location of the PAC.

As an alternative to the use of single-brain–based atlases or static templates, probabilistic atlases are thought to provide information about the neuroanatomic complexity and interindividual variability in a common stereotaxic coordinate system (89, 90). Mapping the probability across a group that a structure will occur at a given voxel location allows statistical evaluation of the location of an ROI in any brain image that has been transformed into stereotaxic space (3). In the largest-ever performed cytoarchitectonic study of the PAC, Rademacher et al (2) examined 27 brains and generated an atlas-based 3D cytoarchitectonic probabilistic map representing the location and extent of the PAC (in 10 left and 10 right hemispheres) in which the degree of intersubject overlap in each stereotaxic position is quantified. Similar cytoarchitectonic probabilistic maps were formulated for other brain regions, such as the motor (91), visual (92), and superior parietal cortices (93), as well as the hippocampus, amygdala, and entorhinal cortex (94); and have been used widely in scientific research. Many of them are implemented in the anatomy toolbox of SPM (http://www.fz-juelich.de/ime/spm_anatomy_toolbox) and have often been used in VBM studies. Integrated into functional imaging experiments, they can improve structure-function investigations of the human brain regions (for commentary, see Ref.95). Especially for the PAC, such maps would likely be effective due to the fact that cytoarchitectonic borders do not consistently coincide with sulcal contours, creating a structure-function mismatch and, also, due to the presence of other areas with different cytoarchitecture, chemoarchitecture, and connectivity sharing the HG. However, the use of such maps is not without disadvantages. They are based on brains that have already suffered volume reduction due to loss of vascular space or fixation and, consequently, they can not be considered to represent the living anatomy. In addition, although the use of such maps is more valid than single-brain–based atlases, it is inappropriate to consider them as representatives of the entire population.

Functional MRI techniques can also provide great assistance in localizing the PAC. Areas that are differentiable histologically have also been shown to be distinct functionally. fMRI studies have shown that there are mirror symmetric reversing tonotopic maps sharing a low-frequency border (high-low-low-high) and oriented in a (caudomedial-rostrolateral) direction along the HG. These maps were hypothesized to represent two subfields within the PAC (18, 19, 96). The study of Formisano et al (18) was the first to identify such maps clearly in humans, as they used a high field strength (7T) MR scanner. The spatial layout of the tonotopically-defined areas was suggested to be similar to cytoarchitonically-defined PAC (5) and to be reproducible in the six subjects studied. However, a recent study performed by Upadhyay et al (19), which was performed using a larger sample (eight subjects), described that the clear presence of these maps only occurred in one hemisphere and, in all but one subject, they were seen clearly mostly in the left hemispheres. These data challenge reproducibility and use of such maps in localizing the PAC in future studies. Another critique is that the production of such maps in a clear way requires a sufficiently high technical demand (7T scanner) (18), which is not commonly available in research centers; this problem was addressed by Upadhyay et al (19), who identified these maps using a 3T scanner. Other examples for the use of fMRI to properly localize the PAC and differentiate it from other non-PAC areas include the following: the PAC specifically lacks activation during sound imagery (97); the maximum of event-related responses seemed to be attained earlier in the PAC (98); the PAC has been found to respond to a broad range of auditory stimuli, while other areas seems to respond preferably to stimuli with sufficiently complex spectral dynamics (99); the PAC area is sensitive to faster temporal modulations and narrower spectral modulations than non-PAC areas (100, 101); the PAC is topographically organized with respect to physical stimulus properties (lateralization and frequency) (21); the PAC (T1b) showed the most robust sound-level–dependent increase in blood oxygen level dependent (BOLD) signal intensity (102); the PAC has been found to respond to sound stimuli in a more sustained fashion than in the surrounding areas with transient responses (103); and, finally, the PAC is not activated by changing the spatial layout of sound (104). Many of these studies, however, considered the PAC to span the entire HG, a consideration that can be misleading. They compared activations within the HG (which was considered to be equal to the PAC) to activations within the PT (which was considered to be the site of non-PAC areas), a comparison that would oversimplify the topography of these areas, as we know from cytoarchitectonic data that non-PAC areas are also present in the HG and that the PAC itself can extend into the PT (2).

Despite these shortcomings, these (functional) segregations or maps of the PAC are advantageous over the cytoarchitectonic probabilistic maps, as they are derived from living rather than postmortem specimens and they can map the PAC individually rather than predicting its location from a preprepared probabilistic template, a fact that addresses the problem of interindividual variability of the PAC to advantage.

Finally, the connectivity pattern within the PAC, coupled with functional tonotopic localization, can provide some aid in defining the PAC extent (19). Previous studies hypothesized that regional connectivity patterns are largely preserved across individuals, that those afferent and efferent pathways strongly influence the information processing properties of individual brain regions (105), and that they can be used to reliably identify cortical regions (106), even when these areas lack clear macroanatomical landmarks as in the case of the PAC. Therefore, it is possible (although technically difficult) by collecting both connectivity and functional data, to map the cortical and subcortical structure of the PAC as well as to directly relate the PAC structure and function in individuals.

In conclusion, we address that HG and its included PAC are subjected to great interhemispheric and interindividual variations; yet, they are important regions to study both in health and disease. Their complex anatomy has adversely affected the number, reliability, and reproducibility of structural MRI and fMRI studies (the latter is outside the scope of this work). The crucial finding of this review is that the volume of the cytoarchitectonically-defined PAC does not correlate with the volume of the HG. The need for unified criteria in studying and defining these regions (especially the PAC) seems mandatory. The following paragraphs summarize the main points that should be kept in mind whenever future MRI studies of these regions are planned.

Regarding the HG, it can be readily identified on MRI images based on its characteristic shape and orientation. For proper MR definition of its boundaries in future studies, the work of Penhune et al (3) is recommended. The remarkable interhemispheric and interindividual variations of the HG imply that the use of probability atlases (3, 4) are advantageous over single-brain–based atlases when attempting to normalize data to stereotaxic coordinates of the HG. Regarding HG duplications, especially the CSD subtype, there is a need to agree on unified criteria for their measurement, as this will increase the reproducibility and comparability of future studies. Finally, the use of probability atlases based on templates derived from patients rather than healthy subjects could provide more valid results in clinical studies, such as in schizophrenia.

Regarding the PAC, the tremendous interhemispheric and interindividual variations regarding the size, topography, and relation of the PAC to anatomic landmarks of the HG would similarly entail that the use of single-brain–based atlases or maps constructed by averaging across individuals could lead to serious shortcomings. Unlike the HG, there are no MR-defined macroanatomical boundaries for the PAC and, while waiting for extending the work of in vivo recognition of other cortical regions like the visual cortex (107) into the PAC, there are few measures to localize this region. Cytoarchitectonically-based probabilistic and functional maps were suggested to enhance localization but each has its limitations and it must be kept in mind that, even with these measures, the localization is not exactly accurate but is merely an approximation.

Finally, the scientific community should be careful when describing the HG and PAC. These regions are not synonymous and terms/phrases such as (HG = PAC), (HG, i.e., PAC), etc. should be avoided. Instead, authors should consider the use of terms that represent the real anatomic nomenclature of these regions such as the HG and its included PAC.

Ancillary