Functional Fields in Human Auditory Cortex Revealed by Time-Resolved fMRI without Interference of EPI Noise

NeuroImage 13, 328 –338 (2001) doi:10.1006/nimg.2000.0683, available online at http://www.idealibrary.com on

Functional Fields in Human Auditory Cortex Revealed by TimeResolved fMRI without Interference of EPI Noise Francesco Di Salle,* ,† Elia Formisano,‡ ,§ Erich Seifritz, ¶ David E. J. Linden,㛳 ,** Klaus Scheffler,†† Claudio Saulino,‡‡ Gioacchino Tedeschi,§§ Friedhelm E. Zanella,† Alessandro Pepino,‡ Rainer Goebel,** ,§ and Elio Marciano‡‡ *Department of Biomorphological and Functional Sciences, ‡Department of Electronic Engineering and Telecommunications, ‡‡Neurosciences of Human Communication, University of Naples “Federico II,” Italy; 㛳Department of Neurology and †Department of Neuroradiology, University of Frankfurt, J. W. Goethe, and **Max-Planck-Institut fu¨r Hirnforschung, Frankfurt, Germany; §Department of Psychology, Neurocognition, University of Maastricht; ¶Department of Psychiatry, University of Basel, Switzerland; ††Section of Medical Physics, Department of Diagnostic Radiology, University of Freiburg, Germany; and §§Department of Neurological Sciences, II University, Naples, Italy Received May 30, 2000

INTRODUCTION The gradient switching during fast echoplanar functional magnetic resonance imaging (EPI-fMRI) produces loud noises that may interact with the functional activation of the central auditory system induced by experimental acoustic stimuli. This interaction is unpredictable and is likely to confound the interpretation of functional maps of the auditory cortex. In the present study we used an experimental design which does not require the presentation of stimuli during EPI acquisitions and allows for mapping of the auditory cortex without the interference of scanner noise. The design relies on the physiological delays between the onset, or the end, of stimulation and the corresponding hemodynamic response. Owing to these delays and through a time-resolved acquisition protocol it is possible to analyze the decay of the stimulus-specific signal changes after the cessation of the stimulus itself and before the onset of the EPIacoustic noise related activation (decay-sampling technique). This experimental design, which might permit a more detailed insight in the auditory cortex, has been applied to the study of the cortical responses to pulsed 1000 Hz sine tones. Distinct activation clusters were detected in the Heschl’s gyri and the planum temporale, with an increased extension compared to a conventional block-design paradigm. Furthermore, the comparison of the hemodynamic response of the most anterior and the posterior clusters of activation highlighted differential response patterns to the sound stimulation and to the EPI-noise. These differences, attributable to reciprocal saturation effects unevenly distributed over the superior temporal cortex, provided evidence for functionally distinct auditory fields. © 2001 Academic Press Key Words: auditory cortex; functional magnetic resonance imaging.

1053-8119/01 $35.00 Copyright © 2001 by Academic Press All rights of reproduction in any form reserved.

The central processing of auditory information is performed in primates by a complex and interconnected array of cortical areas. In the monkey, converging results from studies of cortical cytoarchitecture and microelectrode recordings revealed a “core” of auditory processing in the superior temporal lobe. This core extends over approximately 15 mm in length and contains two (Merzenich and Brugge, 1973; Morel et al., 1993) or three (Kaas and Hackett, 1998; Kaas et al., 1999) primary-like fields, with koniocortical organization and a specific correspondence between cortical topography and sound frequencies. This core is surrounded by a narrow region forming the auditory “belt,” operating at a higher level of processing and divided into several functional fields. In the auditory belt weak and nonspecific responses to pure tones but stronger responses to bandpassed noise bursts are reported (Rauschecker et al., 1995; Rauschecker, 1998; Merzenich and Brugge, 1973; Galaburda and Sanides, 1980; Morel et al., 1993; Kaas and Hackett, 1998; Kaas et al., 1999). Since the early 1980s, the organization of the auditory cortex has been addressed in humans by studies measuring the regional cerebral blood flow (rCBF) and glucose metabolism with positron emission tomography (PET) (Mazziotta et al., 1982; Lauter et al., 1985; Lockwood et al., 1999) and by surface recording of electroencephalographic (EEG) and magnetoencephalographic (MEG) evoked activity (Romani et al., 1982; Makela et al., 1993; Pantev et al., 1995; Verkindt et al., 1995; Gutschalk et al., 1999). These studies demonstrated a reliable temporal lobe response to tonal and verbal stimuli, but, due to the limited spatial resolu-

328

BOLD-DECAY SAMPLING FOR AUDITORY STIMULATION

tion, failed to model the details of functional organization known from microelectrode maps in animals. Blood oxygenation level dependent (BOLD) functional magnetic resonance imaging (fMRI) produces maps of brain functions based on neuronal activationinduced differences in regional cerebral blood oxygenation (Belliveau et al., 1991; Ogawa et al., 1992; Bandettini et al., 1992; Kwong et al., 1992). This method presents advantages over PET and MEG in terms of increased spatial resolution and may represent an ideal solution for the investigation of the auditory cortex (Zatorre, 1997). The BOLD response to different auditory stimuli (Binder et al., 1994; Millen et al., 1995; Formisano et al., 1998), and some indications of a tonotopic representation of sound frequencies along a mediolateral direction (Stainer et al., 1997; Wessinger et al., 1997; Talavage et al., 1997; Bilecen et al., 1998) have been demonstrated in previous fMRI studies. However, even these studies have failed, so far, to display the expected functional complexity of the auditory cortex. This might be due to a spatial resolution still inappropriate to match the dimensions of cortical fields, but also to the presence of a loud background acoustic noise in the MR scanner during functional measurements (Ulmer et al., 1998a/b; Bandettini et al., 1998). Indeed, the fast switching of high-strength gradients during echo-planar (EPI) fMRI generates an amplitude-modulated periodic sound with complex spectrum and very high sound pressure levels reaching 80 –100 dB (Hedeen and Edelstein, 1997). FMRI time-series are usually collected through a block-design scheme, alternating the acquisition of blocks of images during stimulation with blocks of images during control or baseline periods. When this scheme is applied in auditory studies, subjects undergo continuous stimulation by the scanner noise plus an alternate stimulation by experimental sound stimuli. This type of experiments assumes the stimulus-related and the scanner noise-related haemodynamic responses to add up in a linear fashion without interacting. Some kinds of interaction, however, are likely to be present. The presentation of sound stimuli during the EPI acquisitions may determine a psychoacoustic “masking effect” that depends on the overlap of spectral components of the stimulus and the sequence noise (Belin et al., 1999; Formisano et al., 1998). Moreover, subjects may be engaged in processes different from simple auditory perception because they have to extract the stimulus from the background acoustic scene (Bregman, 1990). Finally, EPI noise itself determines a consistent activation of the auditory cortex (Bandettini et al., 1998; Ulmer et al., 1998a). Even if the effects of this activation are still unclear, a partial “saturation” of vascular dynamics that results in weaker BOLD responses to other sounds is expected (Bandettini et al., 1998), and other effects, including inhibition or habit-

329

uation, have been hypothesized (Ulmer et al., 1998a; Cho et al., 1998). In the present work we propose an averaged singletrial experimental design for EPI-fMRI that allows a reliable mapping of the stimulus-related cortical activation in the auditory cortex. This approach shares with several recent techniques the aim of spoiling the effects of the background acoustic noise (Scheffler et al., 1998; Bilecen et al., 1998; Edminster et al., 1999; Hall et al., 1999; Eden et al., 1999; Belin et al., 1999). Furthermore, it allows for the analysis of the decay of the stimulus-related BOLD response after the cessation of the stimulus itself (decay-sampling technique) by means of a magnetization-subtraction technique similar to that proposed by Bandettini et al. (1998) for mapping the activation induced by the scanner noise. We applied this technique to the study of the cortical response to amplitude-modulated sine tones; the results have been compared to those obtained by a conventional block design paradigm, and allowed to distinguish, within the auditory cortex, neuronal clusters showing different models of response to the stimulus and to the EPI noise. METHODS Subjects Six right handed healthy male volunteers aged between 22 and 39 years without a history of neurological or psychiatric disorders participated in the study. All subjects gave informed consent to participate in the study and showed normal results in audiological examination. MR Imaging Image acquisition was performed with a 1.5 Tesla superconducting SIGNA MR scanner (General Electric Medical Systems) using a standard circularly polarized head coil. Functional images were acquired with a conventional gradient-echo (GRE) EPI sequence (one excitation for each volume acquired, repetition time/echo time [TR/TE] ⫽ 500 ms/66 ms, flip angle [FA] ⫽ 90°, field of view [FOV] ⫽ 200 ⫻ 200 mm 2, voxel size ⫽ 1.6 ⫻ 1.6 ⫻ 6 mm 3). Due to a limitation on the total number of images per series posed by the scanner software and in order to acquire a sufficient number of time points, a single-slice procedure was chosen. The position of the slice was chosen on a sagittal T1-weighted series in order to cover bilaterally the transverse gyri of Heschl and was controlled on a test axial EPI image. Same-session 3-D anatomical volumes were collected using an inversion-recovery (IR)prepared fast GRE T 1-weighted sequence (TR/TE/FA ⫽ 22.9/10.2/20°) that produced 120 sagittal partitions at a slice thickness of 1.3 mm. With the same spatial

330

DI SALLE ET AL.

FIG. 1. Decay-sampling acquisition protocol. For the stimulated time-series, 20 EPI images at a repetition time of 500 ms were collected after 10 s of acoustic stimulation. For the control time-series, 20 EPIs were collected without any prior auditory stimulation. Six stimulated and control time series were collected within each session. According to this technique, three main signal components were expected to contribute to the MR-signal time-courses of the voxels in the auditory areas: the longitudinal magnetization reaching the steady state (upper panel), the hemodynamic response induced by the experimental acoustic stimulus (lower panel, bold line), the hemodynamic response induced by the scanner acoustic noise during EPI acquisition (lower panel, normal line). Note different scaling between upper and lower panel (for definition of T M, T d, and T epi, see Methods).

location as the EPI sequence, spoiled fast-GRE 2-D time-of-flight (TR/TE/FA ⫽ 26 ms/6.9 ms/60°) angiograms were performed. Auditory Stimulation The system used for auditory stimulation consisted of a personal computer with a commercial soundcard, whose scaled output was connected to the amplifier and to the loudspeaker of a modified hearing aid (E22P 152891 Oticon, Denmark). A flexible rubber tube of 10 mm in diameter and ⬃1 m in length, in which the loudspeaker was inserted at the proximal end outside the bore of the magnet, was used to deliver the digitally generated acoustic stimuli (sampling rate ⫽ 44100 Hz) to the subject in the scanner. The rubber tube was connected on one side to the output of the hearing aid and on the other side, binaurally, to the subject’s ears. The stimulation system was calibrated by means of a half-inch condenser microphone (Larson Davis, 2540), a precision sound level meter (Bruel and Kjaer, 2231) and a calibrator (Larson Davis, CA250). The system accurately preserves tone frequencies in the range between 500 and 4000 Hz and does not induce artifacts in MR images. Special padded headphones and ear plugs

were used in order to provide an overall attenuation of ⬃30 dB to the environmental noise. Stimuli consisted of 10-s sine tones with the carrier frequency at 1000 Hz, amplitude-modulated with a 10-Hz square wave (rise/fall time ⫽ 5 ms, plateau ⫽ 40 ms, duty cycle ⫽ 0.5, interstimulus interval [ISI] ⫽ 50 ms). The sound pressure level of the stimulus, measured at the output of the sound delivery system, was set to 70 dB SPL for all subjects. Experimental Design Functional images were acquired according to the decay-sampling scheme shown in Fig. 1 (Bandettini et al., 1998; Formisano et al., 1999). The first time-series (referred to as “stimulated time-series”) consisted of 20 EPI measurements lasting 10 s that were collected immediately following an acoustic stimulation of 10 s. The experimental stimulation was presented to the subject in absence of the scanner acoustic noise. According to this protocol, three main signal components were expected to contribute to the MR-signal timecourses within the voxels covering the auditory areas: (1) the BOLD response induced by the experimental stimuli, (2) the BOLD response induced by the EPI

BOLD-DECAY SAMPLING FOR AUDITORY STIMULATION

noise, and, superimposed to both, (3) the decaying longitudinal magnetization approximating steady state. The time required to reach the steady state (T M) depends on the flip angle (FA), TR, and T 1 of the tissue. With a TR of 0.5 s and a FA of 90°, the T M of the brain tissue is in the range of 4 to 6 s and is comparable with the time-constant of the haemodynamic response decay following the auditory stimulation (T d ⬃6 –9 s) and with the onset of the response to the EPI-noise (T epi ⬃5–9 s) (Bandettini et al., 1998). With the conventional fMRI acquisition schemes, the first few images of the EPI time-series (i.e., the images that are collected before the longitudinal magnetization reached steadystate) are discarded and only the later images in the time-series (i.e., the images that are collected after the longitudinal magnetization reached the steady-state) are used to detect the stimulus-related signal changes. With the decay-sampling scheme, however, BOLD signals are measured under non-steady-state conditions of the longitudinal magnetization and, since the signal changes due to the magnetization decay are much higher than BOLD changes (20 – 60 vs 1–5% (Bandettini et al., 1998)), it becomes necessary to remove the magnetization effects. To this end, a second time-series (referred to as “control time-series” consisting of 20 EPIs in 10 s) was collected without any preceding auditory stimulation. Assuming that the magnetization component is the same in the “stimulated” and in the “control” time-series, the magnetization effects can be subtracted voxel-wise (stimulated total MR-signal minus control total MR-signal ⫽ net BOLD signal). In these experiments six pairs of control and stimulated time-series were collected for each subject’s session. In order for the magnetization and the haemodynamic responses to reach a stable baseline, stimulation and control time-series were separated by 30 s. Therefore, each experimental session lasted 8 min, including 2 min effective scanning time (6 repetitions ⫻ 2 timeseries ⫻ 10 s). To compare the BOLD decay measurement protocol to a conventional block-design the scanning session comprised an additional functional measurement with six alternated blocks of acoustic stimulation (10 s) and silence (10 s) during continuous (TR ⫽ 500 ms, total scanning time ⫽ 2 min) EPI acquisitions in two subjects. The anatomical position and the other parameters of the sequence were the same as in the previous measurements. Data Analysis Functional time-series were preliminarily realigned using a 2-D motion-correction algorithm (Woods et al., 1993). The linear signal drifts were then estimated using linear interpolation and were subtracted from each voxel signal time-course. Only the samples collected under a steady-state longitudinal magnetization

331

condition (i.e., images 11 through 20) were used for signal drift estimation. Then the masking effects of longitudinal magnetization were removed. Normalized “difference” time-series were created by subtracting the averaged preprocessed control time-series from the averaged preprocessed stimulated time-series and by normalizing the obtained values to the corresponding value of the control time series. To avoid large-vessels artefacts, voxels that presented a percentage signal change greater than 10% in the normalized difference time-series were masked and discarded from further analysis. Cortical areas responding to the auditory stimulation were identified by applying multiple regression analysis (Friston et al., 1995) to the difference timeseries. The time-course of the i th voxel of the difference time-series (x i(t), t ⫽ 1, . . . ,20) was modeled as x i共t兲 ⫽ ␤ 1is 1共t兲 ⫹ ␤ 2is 2共t兲 ⫹ e i共t兲, where s 1(t) is the predictor of the decay of the BOLD response induced by the external auditory stimulation, s 2(t) is the predictor of the BOLD response induced by the EPI noise, e i(t) is the error term, and ␤ 1i and ␤ 2i are the regressor coefficients. The predictor s 2(t) in the model takes into account the possibility that amplitudes of the response induced by the scanner noise in the stimulated and in the control time-series are different. A positive value of the coefficient ␤ 2i indicates a greater response in the stimulated time-series, while a negative value of the coefficient ␤ 2i indicates a greater response in the control time-series. The shapes of s 1(t) and s 2(t) were obtained by convolution of boxcar waveforms representing the experimental conditions (auditory stimulation, EPI-acquisitions) and the linear model of the impulse response proposed by Boynton et al. (1996) (Fig. 1). For each subject, activation maps were generated by associating each voxel of the difference time-series with the value of the F i statistics, resulting from the least square estimate of the coefficients ␤ 1i and ␤ 2i. Effects were only accepted as significant when the corresponding uncorrected significance value P i yielded P⬘i ⫽ P i 䡠 N ⬍ 10 ⫺2, where N was the number of unmasked intrabrain voxels in the subject. 2-D statistical maps were superimposed on the original functional scans and incorporated into the 3-D anatomical data sets through interpolation to the same resolution. Since the 2-D functional and 3-D structural measurements were performed within the same recording session, coregistration of the respective data sets (and of 2-D statistical maps) could be computed directly based on the slice position parameters of the T2*-weighted measurement (number of slices, slice thickness, distance factor, Tra-Cor angle, FOV, shift mean, off-center read, off-center phase, in plane reso-

332

DI SALLE ET AL.

FIG. 2. The MR signal in the original time-series (upper panel), and in the averaged stimulated (lower panel, solid line) and control (lower panel, dashed line) time-series from a region-of-interest (ROI) in the left auditory cortex (left) and to a ROI in the left occipital lobe (right) of subject 1. In the auditory areas the signal intensity measured during the initial samples of each trial of the stimulated time-series (s) was reliably higher than during the corresponding samples of the control time-series (c). The signal difference reduced to zero approximately 5–7 s after the stimulus cessation, when the hemodynamic response to the EPI control scans plateaued.

lution) and on parameters of the 3-D measurement (number of sagittal partitions, shift mean, off-center read, off-center phase, resolution), with respect to the initial overview measurement (scout). For confirmation of coregistration accuracy a 2-D anatomical data set was obtained by reslicing the 3-D data set using the slice position parameters and compared visually to the EPI data sets. For each subject the structural 3-D data sets and 3-D statistical maps were transformed into Talairach space using the BrainVoyager 3.⫻ software (http://www.brainvoyager.com). Talairach transformation was performed in two steps. The first step consisted in rotating the 3-D data set of each subject to be aligned with the stereotaxic axes. For this step the location of the anterior commissure (AC) and the posterior commissure (PC) and two rotation parameters for midsagittal alignment had to be specified manually in the 3-D MPR data set. In the second step the extreme points of the cerebrum were specified. These points together with the AC and PC coordinates were then used to scale the 3-D data sets into the dimensions of the standard brain of the Talairach and Tour-

noux (1988) atlas using a piecewise affine and continuous transformation for each of the 12 defined subvolumes. Finally, segmentation of 3-D anatomical data sets and surface rendering was obtained following procedures described elsewhere (Goebel et al., 1998; Dierks et al., 1999; Linden et al., 1999). RESULTS Exploratory time-course analysis of the stimulated and control time-series revealed that for regions-ofinterest (ROIs) located in the transverse temporal gyrus the signal intensity measured during the initial samples of each trial of the stimulated time-series was reliably higher than during the corresponding samples of the control time-series (Fig. 2, left upper panel). This signal difference reduced to zero 5–7 s after the cessation of the stimulus (Fig. 2, left lower panel). The effect was specifically task-related since it was not evident in nonauditory regions. For comparison purposes, the signal time-course of a mesial occipital region, supposed not to be functionally related to auditory perception,

333

BOLD-DECAY SAMPLING FOR AUDITORY STIMULATION

FIG. 3. Percentage of activation (PoA) in individual subjects. PoA is calculated as the number of activated voxels in an area divided by the total number of activated voxels in the subject (see Table 1). Bars on the left refer to left hemisphere, bars on the right refer to right hemisphere.

was systematically analyzed. This ROI did not show any difference between stimulated and control timeseries (Fig. 2, right panels). Multiple regression analysis of the difference timeseries showed that the highest significance values were obtained bilaterally in the auditory cortex, with some degree of asymmetry between the two sides in all subjects (Fig. 3; Table 1). Activation was strongly clustered into two or three distinct areas, partly corresponding to Heschl’s gyri by location, length and inclination, as recognizable from a gross anatomic analysis (Figs. 3 and 4; Table 1), but also including the superior temporal gyrus (STG) and the superior temporal sulcus (STS). These territories generally had a stripe-like morphology, and their overall inclination was rostrolateral to caudomedial. The first territory was located on the rostral bank of the anterior transverse temporal gyrus and extended to the planum polare (Fig. 4). The second territory roughly parallel to the first one, was located about 10 mm caudally, extending from the second transverse temporal gyrus (if present) to the caudal bank of the first transverse temporal gyrus (Fig. 4). An additional third territory was present in four subjects and was located more caudally on the planum temporale (Figs. 3 and 4, Table 1). The territories also showed extension to the lateral surface of the temporal lobe, including the STG and the STS. The clusters of activated pixels presented location, length and inclination clearly different from the regional blood vessels, as apparent from 2-D time-offlight vascular maps with the same spatial location

(Fig. 5). Cluster locations were approximately the same as in the block-design comparison experiment, but at the selected threshold (P⬘ ⬍ 10 ⫺2, corrected) the decaysampling paradigm showed an increased extension of activation in the auditory fields (Fig. 5). Interestingly, further analysis of supra-threshold voxels revealed the presence of two different time-series profiles. The first, found in anterior auditory regions, started from higher percentage signal change values (5.2 ⫾ 1.8%) and exactly matched the expected decay of the haemodynamic response, reaching the baseline level 4 – 6 s after the cessation of the stimulus. The second time-course profile, found in the posterior clusters, started from reduced percentage signal change values of 2.6 ⫾ 1.0% and, after reaching the baseline in about the same time as the first timecourse, presented a marked negative value of ⫺2.5 ⫾ 1.5% that lasted until the end of the sampling period without recovery to baseline (Fig. 6). The distribution of the estimated coefficients predicting the decay of the BOLD responses induced by the external auditory stimulation and by the EPI noise, corresponding to time-courses of the suprathreshold voxels of all subjects are shown in Fig. 7a. The mean values of the distributions of the ␤ 1i and ␤ 2i coefficients, for the voxels in the anterior and posterior clusters were significantly different (t test, ␤ 1i, P ⬍ 10 ⫺7; ␤ 2i, P ⬍ 10 ⫺11). Since the predictor s 2(t) in the model takes into account the possibility that the amplitudes of the response induced by the scanner noise in the stimulated and in the control time-series are different, the negative values of the coefficients ␤ 2i indicate a greater response in the control time-series. In order to clarify these differences, the signal time-courses were reanalyzed in nonsubtracted averaged stimulated and conTABLE 1 Locations, Frequency of Activation, and Percentage of Activation (PoA) of Activated Areas during Auditory Stimulation Tailarach coordinates Anatomical area

X

Y

Z

No of subjects

PoA

Anterior HG—left Anterior HG—right Posterior HG—left Posterior HG—right PT—left PT—right

⫺44 ⫾ 3 41 ⫾ 4 ⫺46 ⫾ 6 55 ⫾ 5 ⫺62 ⫾ 5 46 ⫾ 8

⫺15 ⫾ 4 ⫺14 ⫾ 5 ⫺26 ⫾ 2 ⫺13 ⫾ 3 ⫺44 ⫾ 4 ⫺25 ⫾ 7

1⫾2 3⫾2 3⫾1 2⫾2 4⫾2 4⫾1

6 5 6 4 4 3

0.27 0.18 0.36 0.12 0.03 0.04

Note. The position of each area is given as the mean (⫾SD calculated over the corresponding number of subjects) of the Tailarach coordinates of the center of mass of each cluster as selected in the individual multiple regression maps (P⬘ ⬍ 10 ⫺2, corrected). The range of Z coordinate is limited by the single slice acquisition mode. The percentage of activation indicates the mean fraction of the suprathreshold voxels. For PoA in single subjects see Fig. 3 (HG, Heschl’s gyrus; PT, planum temporale).

334

DI SALLE ET AL.

FIG. 4. 3-D representation of the auditory cortex activation determined by the 1000-Hz stimulus. FIG. 5. Block-design activation map (left upper panel), decay sampling activation map (right upper panel), and vascular map (left lower panel) of subject 2. Note the bigger extent of activation in the decay-sampling map and the different location of activation clusters and large vessels (white).

trol time-series. The negative value in the difference time-courses of the posterior clusters was indeed explained by a marked EPI-induced activation, which peaked at a latency of 6 s and was only recognizable in the control time-series (Fig. 7b).

DISCUSSION The presented data indicate that an averaged singletrial experimental design, which samples the decay of the BOLD response to acoustic stimuli, can be success-

BOLD-DECAY SAMPLING FOR AUDITORY STIMULATION

335

FIG. 6. Averaged “difference” time-courses of the anterior (left upper panel) and posterior (left lower panel) gyrus of Heschl after a pulsed 1000-Hz stimulus. Bars indicate one SD. Dashed lines indicate the predicted time-courses. Surface-rendered (right upper panel) and 2-D representation (right lower panel) of activated gyri of Heschl: red and yellow colours indicate the prevalence of the type a and type b decay patterns (see Methods), respectively. The two patterns appear clearly clustered in the anterior and posterior areas of activation.

fully applied to the mapping of the auditory cortex. The paradigm was designed to avoid any interference between experimental stimulus and scanner noise and to yield a temporal separation between the corresponding BOLD responses. This goal was achieved through EPI acquisitions that started after the completion of each preceding sound stimulation period. Given the temporal dynamics of the BOLD signal, the first EPI measurements after the end of the sound stimulation period are influenced only by the decay of the stimulusrelated response and are not yet affected by the EPI noise-related activation. The morphology of the activation in response to pure tones provided by our approach is in good agreement with the known regional anatomy of the auditory cortex. In humans, as in other primates, the cortical structures devoted to the auditory processing are located on the temporal bank of the Sylvian fissure. This region consists of the first gyrus of Heschl (or first transverse temporal gyrus) and, in some individuals, of a second transverse temporal gyrus, and includes several surrounding fields extending to the planum polare, the planum temporale, and the lateral surface of the temporal lobe. While the first gyrus of Heschl receives

direct ascending projections from the medial geniculate body and is the putative site of the primary auditory cortex, the anterior and posterior areas are thought to correspond to the auditory belt, the higher auditory processing area of nonhuman primates (Kaas et al., 1999). Accordingly, the activation in our experiments was strongly clustered to the gyri of Heschl and the surrounding temporal surface, and three main domains were distinguishable, which corresponded well to the regional gyral anatomy (Penhune et al., 1996) and to the description by Gaschler-Markefski et al. (1997) of three main stripe-like clusters (T1–T3) of auditory activation. The analysis of the difference time-series of the BOLD signal led to the identification of two clearly distinct patterns of haemodynamic response. In the first pattern (type a, Fig. 6), a stable baseline was reached at about 4 –5 s after the end of stimulation, in the second (type b, Fig. 6) the decay continued and showed a prolonged undershoot lasting until the end of the sampling period. The strong and prolonged undershoot of the type b response is likely to reflect different responses to the EPI-noise in the stimulated epochs

336

DI SALLE ET AL.

FIG. 7. (a) Distributions of the estimated regressor coefficients in the anterior and posterior gyri of Heschl. The mean value of the distribution of the coefficients that were mainly determined by the response to the stimulus (␤ 1i) was significantly higher in the anterior gyrus of Heschl’s (P ⬍ 10 ⫺7, t test). The mean value of the distribution of the coefficients that were mainly determined by EPI-noise (␤ 2) was approximately 0 in the anterior areas and negative in the posterior areas (P ⬍ 10 ⫺11, t test). (b) Time-courses in the stimulated (solid line) and control (dashed line) time-series for anterior (left) and posterior (right) gyrus of Heschl. Note, in the posterior areas, the higher BOLD response to the scanner noise during the control EPIscans.

(where EPI-noise stimulation is preceded by the sound stimulus) compared to control periods. The voxels presenting this undershoot would then be supposed to respond more strongly to the EPI-noise if it is not preceded by another acoustic stimulation.

The interpretation of the type b pattern of hemodynamic response as arising from a modulation of the response to the EPI-noise acoustic stimulation by the preceding sound stimuli is confirmed by the analysis of the nonsubtracted stimulated and control time-series in the areas which showed the type b pattern in the difference time-courses. This analysis highlighted the presence of a marked EPI-induced activation which peaked at a latency of 6 s and was only recognizable in the control time-series. This finding, absent in voxels presenting the type a decay, suggests the presence of a saturation effect exerted by the acoustic stimuli over the scanner noise (analogous to that exerted by the EPI-noise over experimental acoustic stimuli, cf. Bandettini et al., 1998). Such a saturation could plausibly be expected in nontonotopic areas of auditory cortex, but not in tonotopic areas, because the narrow frequency content of the experimental 1000-Hz sine tone would not saturate the wider spectral distribution of the EPI noise. Correspondingly, the type b BOLD decay was mainly localized to the posterior stripe of the auditory activation area, which is likely to include functional fields corresponding to the posterior portion of the monkey auditory belt complex (Rauschecker et al., 1995). In the auditory belt, multiple and weakly specific tonotopic representations are thought to be interspersed among functional fields responsive to different sound features. The BOLD signal pattern of the posterior area is likely to be produced by frequency-unspecific interferences in neuronal firing and regional hemodynamics between two considerably different auditory stimuli. This area would therefore be a plausible candidate for a portion of the human analogue of the auditory belt. Even if our single-slice approach permits only a limited exploration of the complex regional anatomy (Zatorre, 1997; Wessinger et al., 1997; Steinmetz et al., 1989; Penhune et al., 1996) our data indicate an asymmetry in auditory cortex response which is consistent with previous reports (Lauter et al., 1985; Binder et al., 1994; Millen et al., 1995; Scheffler et al., 1998). This asymmetry is in agreement with recent indications of a hemispheric specialization in the processing of auditory tonal and rhythmic patterns (Penhune et al., 1999) and is unevenly distributed in the three territories activated in the auditory cortex (Fig. 3, Table 1). The consistency of our results across subjects corroborates the validity of the decay-sampling method even with a signal-to-noise ratio much less favorable than in previous similar studies, where a higher magnetic field strength and/or a lower spatial resolution were used (Bandettini et al., 1998; Talavage et al., 1999). The smaller voxel size constitutes an additional advantage because it allows for a more detailed analysis of the functional parcellation of the human auditory cortex and to discard voxels containing large veins more specifically.

BOLD-DECAY SAMPLING FOR AUDITORY STIMULATION

In sum, the combination of a high spatial and temporal resolution with the suppression of the EPI-noise influence and a detailed analysis of the different patterns of BOLD signal behaviour, which form the basis of the present study, helped to identify a spatial segregation of neural activities in human auditory cortex, which might correspond to frequency-specific and frequency-unspecific functional fields. CONCLUSIONS The application of a single-trial experimental design which is sensitive to the decay of the BOLD response and spoils the interference of the EPI noise-related effects, together with the use of high spatial resolution acquisitions, can open insightful perspectives in the study of the auditory cortex in humans and laboratory animals. The first application of this experimental design to the analysis of the auditory cortex response to single frequency sine tone stimuli provided evidence for a differentiation of neuronal clusters with functional behaviors which might suggest frequency-specific and frequency-unspecific activities. Furthermore, their anatomical location closely resembles the core and part of the belt components of the central auditory network identified in nonhuman primates. Among the many complex features of the auditory cortex which might be revealed by this method, a clear understanding of the tonotopic organization, while exceeding the purposes of the present paper, still remains a main issue and will deserve to be specifically addressed in future studies. We expect that an extensive application of the decaysampling procedure should eventually clarify the details of the tonotopic distribution of sound frequencies along the primary-like functional fields as well as the representation of other sound features, ultimately leading to a deeper understanding of the functional architecture of the human auditory cortex. ACKNOWLEDGMENTS E.S. is supported by Swiss National Science Foundation Grants 3200-052194.97 and 63-58040.99 and holds a University of Basel habilitation stipend and an SNF-professorship.

REFERENCES Bandettini, P. A., Wong, E. C., Hinks, R. S., Tikofsky, R. S., and Hyde, J. S. 1992. Time course EPI of human brain function during task activation. Magn. Reson. Med. 25: 390 –398. Bandettini, P. A., Jesmanowicz, A., Van Kylen, J., Birn, R. M., and Hyde, J. S. 1998. Functional MRI of brain activation induced by scanner acoustic noise. Magn. Reson. Med. 39: 410 – 416. Belin, P., Zatorre, R. J., Hoge, R., Evans, A. C., and Pike, B. 1999. Event-related fMRI of the auditory cortex. Neuroimage 10: 417– 429.

337

Belliveau, J. W., Kennedy, D., McKinstry, R. C., Buchbinder, B. R., Weisskoff, R. M., Cohen, M. S., Vevea, J. M., Brady, T. J., and Rosen, B. R. 1991. Functional mapping of the human visual cortex by magnetic resonance imaging. Science 254: 716 –719. Bilecen, D., Scheffler, K., Schmid, N., Tschopp, K., and Seelig, J. 1998. Tonotopic organization of the human auditory cortex as detected by BOLD-FMRI. Hear. Res. 126: 9 –27. Binder, J. R., Rao, S. M., Hammeke, T. A., Yetkin, F. Z., Jesmanowicz, A., and Bandettini, P. A. 1994. Functional Magnetic Resonance Imaging of Human Auditory Cortex. Ann. Neurol. 6: 662– 672. Boynton, G. M., Engel, S. A., Glover, G. H., and Heeger, D. J. 1996. Linear systems analysis of functional magnetic resonance imaging in human V1. J. Neurosci. 16: 4207– 4241. Bregman, A. 1990. Auditory Scene Analysis. MIT Press, Cambridge, MA. Cho, Z. H., Chung, S. C., Lim, D. W., and Wong, E. K. 1998. Effects of the acoustic noise of the gradient systems on fMRI: A study on auditory, motor, and visual cortices. Magn. Reson. Med. 39: 331– 335. Dierks, T., Linden, D. E. J., Jandl, M., Formisano, E., Goebel, R., Lanfermann, H., and Singer, W. 1999. Activation of Heschl’s gyrus during auditory hallucinations. Neuron 22: 615– 621. Edmister, W. B., Talavage, T. M., Ledden, P. J., and Weisskoff, R. M. 1999. Improved auditory cortex imaging using clustered volume acquisitions. Hum. Brain. Mapp. 7: 89 –97. Eden, G. F., Joseph, J. E., Brown, H. E., Brown, C. P., and Zeffiro, T. A. 1999. Utilizing haemodynamic delay and dispersion to detect fMRI signal change without auditory interference: The behavior inter-leaved gradients technique. Magn. Reson. Med. 41: 13–20. Formisano, E., Pepino, A., Bracale, M., Di Salle, F., Saulino, C., and Marciano, E. 1998. Localization and characterization of auditory perception through functional magnetic resonance imaging. Technol. Health Care 6: 111–123. Formisano, E., Di Salle, F., Pepino, A., Saulino, C., Morrone, R., La Montagna, D., Tedeschi, G., and Bracale, M. 1999. Tonotopic mapping in absence of scanner acoustic noise [Abstract]. Neuroimage 9: S645. Friston, K. J., Holmes, A. P., Worsley, K. J., Poline, J. B., Frith, C. D., and Frackowiak, R. S. J. 1995. Statistical parametric maps in functional imaging: A general linear approach. Hum. Brain Map. 2: 189 –210. Galaburda, A. M., and Sanides, F. 1980. Cytoarchitectonic organization of the human auditory cortex. J. Comp. Neurol. 190: 597– 610. Gaschler-Markefski, B., Baumgart, F., Tempelmann, C., Schindler, F., Stiller, D., Heinze, H. J., and Scheich, H. 1997. Statistical methods in functional magnetic resonance imaging with respect to nonstationary time series: Auditory cortex activity. Magn. Reson. Med. 38: 811– 820. Goebel, R., Khorram-Sefat, D., Muckli, L., Hacker, H., and Singer, W. 1998. The constructive nature of vision: Direct evidence from functional magnetic resonance imaging studies of apparent motion and motion imagery. Eur. J. Neurosci. 10: 1563–1573. Gutschalk, A., Mase, R., Roth, R., Ille, N., Rupp, A., Hahnel, S., Picton, T. W., and Scherg, M. 1999. Deconvolution of 40 Hz steadystate fields reveals two overlapping source activities of the human auditory cortex. Clin. Neurophysiol. 110: 856 – 868. Hall, D. A., Haggard, M. P., Akeroyd, M. A., Palmer, A. R., Summerfield, A. Q., Elliot, M. R., Gurney, E. M., and Bowtell, R. W. 1999. “Sparse” temporal sampling in Auditory fMRI. Hum. Brain Map. 7: 213–223. Hedeen, R. A., and Edelstein, W. A. 1997. Characterization and prediction of gradient acoustic noise in MR imagers. Magn. Reson. Med. 37: 7–10. Kaas, J. H., and Hackett, T. A. 1998. Subdivisions of auditory cortex and levels of processing in primates. Audiol. Neurootol. 3: 73– 85.

338

DI SALLE ET AL.

Kaas, J. H., Hackett, T. A., and Tramo, M. J. 1999. Auditory processing in primate cerebral cortex. Curr. Opin. Neurobiol. 9: 64 – 170. Kwong, K. K., Belliveau, J. W., Chesler, D. A., Goldberg, I. E., Weisskoff, R. M., Poncelet, B. P., Kennedy, D. N., Hoppel, B. E., Cohen, M. S., Turner, R., Cheng, H. M., Brady, T. J., and Rosen, B. R. 1992. Dynamic magnetic resonance imaging of human brain activity during primary sensory stimulation. Proc. Natl. Acad. Sci. USA 89: 5675–5679. Lauter, J. L., Herscovitch, P., Formby, C., and Raichle, M. E. 1985. Tonotopic organization in human auditory cortex revealed by positron emission tomography. Hear. Res. 20: 199 –205. Linden, D. E. J., Prvulovic, D., Formisano, E., Vo¨llinger, M., Zanella, F. E., Goebel, R., and Dierks, T. 1999. The functional neuroanatomy of target detection: An fMRI study of visual and auditory oddball tasks. Cereb. Cortex 9: 815– 823. Lockwood, A. H., Salvi, R. J., Coad, M. L., Arnold, S. A., Wack, D. S., Murphy, B. W., and Burkard, R. F. 1999. The functional anatomy of the normal human auditory system: Responses to 0.5 and 4.0 kHz tones at varied intensities. Cereb. Cortex 9: 65–76. Makela, A., Ahonen, M., Hamalainen, M., and Hari, R. 1993. Functional differences between auditory cortices of the two hemisphere revealed by whole-head neuromagnetic recordings. Hum. Brain Map. 1: 48 –56. Mazziotta, J. C., Phelps, M. E., Carson, R. E., and Kuhl, D. E. 1982. Tomographic mapping of human cerebral metabolism: Auditory stimulation. Neurology 32: 921–937. Merzenich, M. M., and Brugge, J. F. 1973. Representation of the cochlear partition on the superior temporal plane of the macaque monkey. Brain Res. 50: 270 –296. Millen, S. J., Haughton, V. M., and Yetkin, M. D. 1995. Functional magnetic resonance imaging of the central auditory pathway following speech and pure-tone stimuli. Laryngoscope 105: 1305– 1310. Morel, A., Garraghty, P. E., and Kaas, J. H. 1993. Tonotopic organization, architectonic fields, and connections of auditory cortex in macaque monkeys. J. Comp. Neurol. 335: 437– 459. Ogawa, S., Tank, D., Menon, R., Ellermann, J. M., Kim, S. G., Merkle, K., and Ugurbil, K. 1992. Intrinsic signal changes accompanying sensory stimulation: Functional brain mapping using MRI. Proc. Natl. Acad. Sci. USA 89: 5951–5955. Pantev, C., Bertrand, O., Eulitz, C., Verkindt, C., Hampson, S., Schuierer, G., and Elbert, T. 1995. Specific tonotopic organizations of different areas of the human auditory cortex revealed by simultaneous magnetic and electric recordings. Electroencephalog. Clin. Neurophysiol. 94: 26 – 40. Penhune, V. B., Zatorre, R. J., MacDonald, J. D., and Evans, A. C. 1996. Interhemispheric anatomical differences in human primary auditory cortex: Probabilistic mapping and volume measurement from magnetic resonance scans. Cereb. Cortex 6: 661– 672. Penhune, V. B., Zatorre, R. J., and Feindel, W. H. 1999. The role of auditory cortex in retention of rhythmic patterns as studied in patients with temporal lobe removals including Heschl’s gyrus. Neuropsychologia 37: 315–331.

Rauschecker, J. P., Tian, B., and Hauser, M. 1995. Processing of complex sounds in the macaque nonprimary auditory cortex. Science 268: 111–114. Rauschecker, J. P. 1998. Cortical processing of complex sounds. Curr. Opin. Neurobiol. 8: 516 –521. Robson, M. D., Dorosz, J. L., and Gore, J. C. 1998. Measurements of the temporal fMRI response of the human auditory cortex to train of tones. Neuroimage 7: 185–198. Romani, G. L., Williamson, S. J., and Kaufman, L. 1982. Tonotopic organization in human auditory cortex. Science 216: 1339 –1340. Scheffler, K., Bilecen, D., Schmid, N., Tschopp, K., and Seelig, J. 1998. Auditory cortical responses in hearing subjects and unilateral deaf patients as detected by functional magnetic resonance imaging. Cereb. Cortex 8: 156 –163. Stainer, J. C., Ulmer, J., Yetkin, F. Z., Haughton, V. M., Daniels, D. L., and Millen, S. J. 1997. Functional MR of the primary auditory cortex: An analysis of pure tone activation and tone discrimination. AJNR Am. J. Neuroradiol. 18: 601– 610. Steinmetz, H., Fu¨rst, G., and Freund, H. J. 1989. Cerebral cortical localization: Application and validation of the proportional grid system in MR imaging. J. Comput. Assist. Tomogr. 13: 10 –19. Talairach, J., and Tournoux, P. 1988. Co-planar stereotaxic atlas of the human brain: 3-dimensional proportional system: An approach to cerebral imaging. Thieme, Stuttgart. Talavage, T. M., Ledden, P. J., Sereno, M. I., Rosen, B. R., and Dale, A. M. 1997. Multiple phase-encoded tonotopic maps in human auditory cortex [Abstract]. Neuroimage 5: S8. Talavage, T. M., Edmister, W. B., Ledden, P. J., and Weisskoff, R. M. 1999. Quantitative assessment of auditory cortex responses induced by imager acoustic noise. Hum. Brain Map. 7: 79 – 88. Ulmer, J. L., Biswal, B. B., Zerrin Yetkin, F., Mark, L. P., Mathews, V. P., Prost, R. W., Estkowski, L. D., McAuliffe, T. L., Haughton, V. M., and Daniels, D. L. 1998a. Cortical activation response to acoustic echo planar scanner noise. J. Comput. Assist. Tomogr. 22: 111–119. Ulmer, J. L., Biswal, B. B., Mark, L. P., Mathews, V. P., Prost, R. W., Millen, S. J., Garman, J. N., and Horzewski, D. 1998b. Acoustic echoplanar scanner noise and pure tone hearing thresholds: The effects of sequence repetition times and acoustic noise rates. J. Comput. Assist. Tomogr. 22: 480 – 486. Verkindt, C., Bertrand, O., Perrin, F., Echallier, J-F., and Pernier, J. 1995. Tonotopic organization of the human auditory cortex: N100 topography and multiple dipole analysis. Electroencephalogr. Clin. Neurophysiol. 96: 143–156. Wessinger, C. M., Buonocore, M. H., Kussmaul, C. L., and Mangun, G. R. 1997. Tonotopy in human auditory cortex examined with functional magnetic resonance imaging. Hum. Brain Map. 5: 18 – 25. Woods, R. P., Cherry, S. R., and Mazziotta, J. C. 1992. Rapid automated algorithm for aligning and reslicing PET images. J. Comput. Assist. Tomogr. 16: 620 – 633. Zatorre, R. J. 1997. Functional neuroimaging in the study of the human auditory cortex. AJNR Am. J. Neuroradiol. 18: 621– 623.