Increased BOLD sensitivity in the orbitofrontal cortex using slice-dependent echo times at 3 T

Available online at www.sciencedirect.com

Magnetic Resonance Imaging 31 (2013) 201 – 211

Increased BOLD sensitivity in the orbitofrontal cortex using slice-dependent echo times at 3 T☆ Sebastian Domsch a,⁎, 1 , Julia Linke b, 1 , Patrick M. Heiler a , Alexander Kroll a, b , Herta Flor b , Michèle Wessa b , Lothar R. Schad a a

Department of Computer Assisted Clinical Medicine, Medical Faculty Mannheim, Heidelberg University, Theodor-Kutzer-Ufer 1–3, 68167 Mannheim, Germany b Department of Cognitive and Clinical Neuroscience, Central Institute of Mental Health, Medical Faculty Mannheim, Heidelberg University, Germany Received 13 December 2011; revised 12 June 2012; accepted 21 June 2012

Abstract Functional magnetic resonance imaging (fMRI) exploits the blood oxygenation level dependent (BOLD) effect to detect neuronal activation related to various experimental paradigms. Some of these, such as reversal learning, involve the orbitofrontal cortex and its interaction with other brain regions like the amygdala, striatum or dorsolateral prefrontal cortex. These paradigms are commonly investigated with event-related methods and gradient echo-planar imaging (EPI) with short echo time of 27 ms. However, susceptibility-induced signal losses and image distortions in the orbitofrontal cortex are still a problem for this optimized sequence as this brain region consists of several slices with different optimal echo times. An EPI sequence with slice-dependent echo times is suitable to maximize BOLD sensitivity in all slices and might thus improve signal detection in the orbitofrontal cortex. To test this hypothesis, we first optimized echo times via BOLD sensitivity simulation. Second, we measured 12 healthy volunteers using a standard EPI sequence with an echo time of 27 ms and a modified EPI sequence with echo times ranging from 22 ms to 47 ms. In the orbitofrontal cortex, the number of activated voxels increased from 87±44 to 549±83 and the maximal t-value increased from 4.4±0.3 to 5.4±0.3 when the modified EPI was used. We conclude that an EPI with slicedependent echo times may be a valuable tool to mitigate susceptibility artifacts in event-related whole-brain fMRI studies with a focus on the orbitofrontal cortex. © 2013 Elsevier Inc. All rights reserved. Keywords: Echo time; Event-related fMRI; Susceptibility gradients; BOLD sensitivity; Orbitofrontal cortex

1. Introduction In the past two decades, functional magnetic resonance imaging (fMRI) exploiting the blood oxygenation level dependent (BOLD) effect [1,2] has revealed valuable insights of the brain functions underlying psychological processes like decision making and reward processing [3]. The BOLD effect associated with oxy- and deoxyhemoglobin and perfusion changes [4] has been analytically described [5,6] and tested in phantoms [7]. Neuronal activation ☆ Conflict of interest declaration: None of these authors has any current or potential conflict of interest concerning this paper. ⁎ Corresponding author. Tel.: +49 621 383 5120; fax: +49 621 383 5123. E-mail address: [email protected] (S. Domsch). 1 These authors contributed equally to this work.

0730-725X/$ – see front matter © 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.mri.2012.06.020

in orbitofrontal regions is difficult to image as signal loss and image distortions are common in this brain region [8]. These detrimental effects result from macroscopic susceptibility gradients occurring close to air/tissue interfaces and are a general problem of the gradient-echo echo-planar imaging (EPI) sequence [9] commonly used for fMRI [10,11]. Several strategies to compensate for susceptibility effects have been proposed and include z-shimming [12–14], tailored excitation pulses [15–17] or spiral trajectories [18–20]. Either these strategies require sophisticated modifications of the pulse sequences and processing routines or they compromise spatial or temporal resolution. An alternative strategy is the optimization of echo times since the BOLD contrast depends critically on this sequence parameter particularly in subcortical brain areas [21]. Robust activation in the amygdala, also affected by susceptibility gradients, and the orbitofrontal cortex has been observed in

202

S. Domsch et al. / Magnetic Resonance Imaging 31 (2013) 201–211

many previous whole-brain fMRI studies using an EPI sequence with a short echo time of 27 ms [22–24]. However, the orbitofrontal cortex extends over several slices, and each slice is affected differently by susceptibility gradients. For a block design focusing at the amygdala, Stoecker et al. [25] showed that an EPI sequence with slice-dependent echo times improved signal detection at 1.5 T. This method leaves temporal resolution unaffected, which is important for eventrelated designs. Hence, it would be of great relevance for neuroimaging research if this approach could also improve signal detection in the orbitofrontal cortex during eventrelated fMRI compared to a standard EPI protocol with reduced echo time at field strength of 3 T. The goal of the present work was the improvement of BOLD sensitivity (BS) in the orbitofrontal cortex using slice-dependent echo times. First, T2* relaxation times and field gradient measurements were performed to simulate BS and to approximate optimal echo times in single slices. Second, a standard EPI sequence with short echo time was compared to a modified EPI sequence with slice-dependent echo times during an event-related whole-brain fMRI experiment with a focus on the orbitofrontal cortex. The practical relevance of the modified EPI was judged by comparing the number of activated voxels and the statistical significance of the activation. 2. Materials and methods 2.1. BS calculation During neuronal activation, the BOLD effect alters microscopic background fields and thus the relaxation rate T2*. In order to detect local changes of neuronal activation, the signal has to be susceptible to small changes of T2*. The BS is a measure of the local signal intensity change (dI) related to changes of local T2* relaxation times dT2* and is proportional to the derivative of the signal intensity with dI respect to T2*, BS∝ . Neglecting macroscopic backdT2* ground fields, the signal intensity is given by I = ρ:exp½−TE = ðT2 *Þ:

ð1Þ

The first-order Taylor expansion of the signal intensity evaluated at the T2* value during the baseline state (T*2B) can be written as: ! TE
* *  *2 −TE = T2B 1+ ⋅ T2 −T2B ; I = ρ⋅e ð2Þ T *2 2B

where ρ is the local spin density. Eq. (2) motivates the following definition for the BS:   2 dI  * BS : = T2B : ð3Þ  dT2* T * = T * 2

2B

[21]. Calculating the BS according to this definition yields the simple expression: BS = TE*I;

ð4Þ

which has been experimentally verified [26]. Accordingly, the maximal BS is reached when TE is equal to T2*, which is in accordance with literature results presented in the work of Gati et al. [27]. In the following, the effects of macroscopic field inhomogeneities on BS will be considered. Through-plane gradients (Gss) in the z-direction cause extra spin dephasing along the slice direction. Accordingly, the expression for the signal intensity given by Eq. (1) has to be modified to: 0 I = ρ:exp½−TE = ðT2 *Þ∫Pðz0 ÞeiγGss z TE dz0 ;

ð5Þ

where P(z) denotes the slice profile and γ the gyromagnetic ratio. Assuming rectangular slice profiles, which were used for all measurements in this study, the integral in Eq. (5) can be analytically computed and is given by sinc [γ·Gss·TE·Δz/2], where Δz denotes the slice thickness. As discussed in the work of Deichmann et al. [21], macroscopic in-plane gradients (Gsp) parallel to the EPI phase encode direction alter the sampling rate and therefore also the signal intensity. They further shift the effective echo time (TE eff ) of the central gradient echo of the EPI readout. Both effects can be described by the single variable Q, defined as: Q = 1−γ Δt FoV Gsp = ð2πÞ;

ð6Þ

where Δt denotes the time between consecutive gradient echoes and FOV denotes the field of view in the EPI phase encoding direction. Q affects the signal and the echo time in the following way: I = I0 = Q and TE eff = TE = Q:

ð7Þ

In Eq. (7), I0 is the signal intensity and TE is the effective echo time without in-plane gradients. Eq. (6) shows that Q is equal to one if the field gradients Gsp are zero. Depending on the sign of the gradients, Q becomes smaller or greater than one. According to Eq. (7), the gradient echo is shifted to smaller or to higher TE values if the in-plane gradients are parallel or antiparallel to the EPI phase encoding direction. The signal and therefore also the BS are completely lost if the gradient echo is shifted outside the EPI readout window of the duration TA: jTE−TE = Qj N TA = 2

ð8Þ

[25], where TA depends on the number of phase encoding steps and the bandwidth per pixel. According to Eq. (5) and Eq. (7), in the presence of macroscopic field inhomogeneities

S. Domsch et al. / Magnetic Resonance Imaging 31 (2013) 201–211

along the EPI phase encoding direction and along the slice direction, the signal intensity is given by: I = ρ = Q:exp½−TE = ðQT2 *Þ∫Pðz0 ÞeiγGss z TE dz0 : 0

ð9Þ

Applying the BS definition given in Eq. (3) to the first-order Taylor expansion of Eq. (9) yields the following expression for the BS in the presence of macroscopic background fields: BS = TE = Q2 ρ:exp½−TE = ðQ:T2 *Þ :sinc½γ:Gss :TE:Δz = 2:

ð10Þ

Since BS is proportional to 1/Q 2, the maximal value of the BS increases for positive gradients (Qb1) and decreases for negative gradients (QN1) (i.e., in-plane gradients are parallel or antiparallel to the EPI phase encoding direction). In the case of nonzero in-plane gradients (Gsp≠0; Gss=0), the optimal echo time is shifted to TE=QT2*. Nonzero through-plane gradients (Gss≠0; Gsp=0) shift the BS maximum to lower TE due to additional signal dephasing in slice direction. 2.2. MR measurements All measurements were performed on a 3 T whole-body MR system (Magnetom Trio, Siemens Healthcare, Erlangen, Germany) equipped with a 12-channel head coil. Written informed consent was obtained from all subjects who participated in this study. Gradient field maps were acquired from six healthy subjects (27 ± 1 years, five males) using a multi-gradient echo 3D-FLASH sequence. After applying the RF-pulse, multiple gradient echoes were generated under the readout gradient slope spanning a range of echo times. The sequence was repeated until k-space was fully sampled. The slices were aligned along the anterior and posterior commissure (AC–PC) line and the following sequence parameters were used: 40 slices; matrix size=64×64; repetition time (TR)= 30 ms; flip angle α=12°; FOV=220 mm; slice thickness Δz= 2.5 mm. From each of the contrasts, one phase image was reconstructed, yielding a set of phase images at the different TE values of 2.45 ms, 5.12 ms, 7.79 ms, 10.46 ms, and 13.13 ms. The phase difference Δφ between adjacent voxels depends on the strength of the macroscopic field gradient G (assumed to be linear) and is proportional to the voxel dimension Δy and the time t between the signal excitation and the readout, Δφ=γ·G·Δy·t. Thus, the gradient strength could be determined from the set of phase images acquired at different echo times. Phase unwrapping was performed along the temporal dimension [28] before Gss and Gsp were estimated by two separate linear least squares fit on a voxelby-voxel basis: Δφss = γ:Gss :Δz:t; and

ð10aÞ

Δφsp = γ:Gsp :Δy:t:

ð10bÞ

Afterwards, T2*-weighted intensity images were acquired at 12 different echo times between 5 ms and 140 ms using

203

the multigradient echo 3D-FLASH sequence with the following sequence parameters: TE = [5, 10, 15, 20, 25, 30, 40, 50, 70, 90, 110, 140] ms; α=25°. The remaining sequence parameters were kept constant. It has been shown that exponential T2*-fitting without accounting for through-plane dephasing leads to an underestimation of T2* [29]. Therefore, Gss resulting from the phase mapping study was inserted into Eq. (5), which then was used as the fitting function. Confidence bounds for local T2*-fitting and gradient field mapping were set to 95%. Voxels with a relative fit error of ΔT2*/T2* N 0.25 and ΔG/G N 0.25 were not further processed. The resulting T2* and gradient field maps were coregistered to the T2*-weighted intensity images, spatially normalized to standardized coordinates in the Montreal Neurological Institute space and finally resampled to a 96×96×40 matrix size. 2.3. fMRI experiment with focus on orbitofrontal cortex activation Within an interval of 2 weeks, 12 healthy participants (23±3 years, six females) performed a self-paced probabilistic reward-reversal task twice using either the modified or the standard EPI sequence (40 slices; descending slice direction; matrix size=96×96; TR=2.7 s; bandwidth=1578 Hz, flip angle=90°; FOV=220 mm; slice thickness=2.3 mm, interslice gap=0.7 mm, parallel imaging factor PI = 2 using GRAPPA [30].). Axial slices were aligned along the AC– PC. Both types of EPI sequences were counterbalanced to avoid ordering effects. The echo time of the standard EPI was set to 27 ms. The echo time of the modified EPI sequence increased linearly from TE=22 ms to TE=47 ms (Fig. 1) in ascending slice direction (i.e. from slice 1 to 27) by implementing a slice-dependent delay between the signal excitation and the EPI readout train. In the upper slices (i.e. from slice 28 to 40), a fixed delay of 25 ms was implemented, yielding constant echo times of TE=47 ms. Delaying the echo times increased the temporal resolution (i.e. TR for one EPI image stack) by the sum of the delays in all slices ∑ Δτslice . Compared to the EPI with a fixed TE of 27 slice¼1…40

ms, the actual TR of the slice-dependent TE EPI is prolonged by approximately 0.6 s (assuming the parameter settings from above). If TR is not below 2.5 s, this temporal loss can be compensated by shortening the dead time between the EPI readout and the next RF excitation depending on the matrix size, TE and the bandwidth. Linearly increasing echo times were used to avoid abrupt image intensity changes in adjacent slices, which may otherwise corrupt image coregistration in the process of spatial normalization [25]. The reward-reversal paradigm was identical to the paradigm used by Linke et al. [31]. In brief, volunteers made a forced choice between two play cards on each trial. The “correct” play card received an 8/2 ratio of positive to negative feedback, whereas the “incorrect” play card always received negative feedback. Feedback was provided in the form of winning (reward) or losing (punishment) a small amount of money.

204

S. Domsch et al. / Magnetic Resonance Imaging 31 (2013) 201–211

a voxel-by-voxel basis for every single subject (SPM first-level analysis) according to the general linear model. 3. Results 3.1. BS calculation

Fig. 1. Slice-dependent echo time data acquisition scheme. Slices are aligned along the anterior to the center of the posterior commissure (AC–PC). The orbitofrontal cortex is imaged with increasing echo times between 22 ms and 37 ms.

Imaging data were preprocessed and analyzed according to standard procedures using SPM8 (http://www.fil.ion.ucl.ac. uk/spm). First, functional scans were slice-timed and then realigned to the first scan of each run. Using the modified compared to the standard EPI sequence, the slices were not uniformly sampled in time. Therefore, the standard SPM slice timing routine was adjusted accounting for the actual acquisition timing of the modified EPI sequence. During normalization, images were resampled every 2.3 mm using sinc-interpolation and were smoothed with a 7×7×7-mm 3 Gaussian kernel to decrease spatial noise. For the rewardNpunishment condition, the BOLD signal changes were modeled on

Fig. 2 shows T2*-relaxation times below 10 ms and susceptibility gradients of more than 200 μTm −1 in the lower part (slice 1) of the orbitofrontal cortex observed in a single subject. The upper part (slice 5) is clearly less affected by susceptibility gradients leading to increasing T2* and decreasing susceptibility gradients values. The results of all subjects are summarized in Table 1. In slice 1, T2* was on average between 8±3 ms and 31±7 ms. In- (through-) plane gradients were in the range of −202±57 μTm −1 (−230±56 μTm −1) and 288±63 μTm −1 (214±17 μTm −1). In slice 5, T2* was in the range of 19±3 ms and 47±9 ms. In- (through-) plane gradients were in the range of −135±61 μTm −1 (−209±36 μTm −1) and 73±25 μTm −1 (63±25 μTm −1). Fig. 3 shows the effect of typical gradient strengths on BS observed in the orbitofrontal cortex. For example, given a T2* value of 35 ms and a susceptibility gradient of 150 μTm −1 in the slice (phase) encoding direction would shifts the BS peak from 35 ms to 4 ms (19 ms). Assuming the BOLD signal is acquired at a TE value of 27 ms, then only 87% (2%) of the maximal BS would remain. Fig. 4 depicts a simulation of the maximal TE values before total signal loss occurs due to in-plane gradients in the

Fig. 2. T2*-relaxation time and field gradient maps in slice and phase encoding direction depicted for two different axial slices. The orbitofrontal cortex is marked by the ROI. The data were obtained from a single subject.

S. Domsch et al. / Magnetic Resonance Imaging 31 (2013) 201–211 Table 1 Transverse relaxation times (T2*) and susceptibility gradients in slice and phase encoding direction (Gss, Gsp) observed in two different slices of the orbitofrontal cortex Subject

T2*/ms

Gss/μTm−1

Gsp/μTm−1

a 1 2 3 4 5 6 Mean

[8…32] [10…25] [6…29] [5…32] [11…44] [5…23] [8±3…31±7]

[−178…190] [−240…240] [−250…324] [−200…286] [−100…338] [−243…350] [−202±57…288±63]

[−124…209] [−250…186] [−255…219] [−214…212] [−274…239] [−261…216] [−230±56…214±17]

b 1 2 3 4 5 6 Mean

[17…58] [20…41] [15…35] [22…55] [20…47] [18…47] [19±3…47±9]

[−146…57] [−205…43] [−257…95] [−222…48] [−212…41] [−211…92] [−209±36…63±25]

[−34…90] [−154…79] [−180…65] [−104…27] [−204…81] [−135…94] [−135±61…73±25]

In a and b, the minimal and maximal parameter values denote the 5 and 95 percentiles measured in slice 1 and in slice 5, respectively. The standard deviation denotes the variation between the different subjects.

range of [−350…350] μTm −1. According to Eq. (8), the maximal TE (TE max) was defined as TE max≡1/2·TA/|1−1/ Q|. It shows that a TE value of 22 ms prevents total signal loss for any negative in-plane gradients in the range of [−350…0] μTm −1 (QN1). For increasing positive gradient strengths in the range of [0…180] μTm −1 (0bQb1), the maximal echo time decreases rapidly below 20 ms. In the range of [180… 350] μTm −1 (Qb0), TE max slightly increases but still remains below feasible TE values. Fig. 5 shows a BS calculation based on Eqs. (8) and (10) for typical in- and through-plane gradients. According to this simulation, a TE value of 20 ms is adequate to recover BS for

205

all possible combinations of through-plane gradients in the range of −350 μTm −1 to 350 μTm −1 and in-plane gradients in the range of −350 μTm −1 to the order of 100 μTm −1. When TE is reduced from 40 ms to 30 ms and to 20 ms, the percentage of dead pixels (i.e. BS/BS maxb0.1) decreases from 37% to 32% and to 8% in the left hemisphere of the Gsp–Gss plane (i.e., Gspb0). This echo time reduction yields nearly no effect in the right hemisphere (i.e., Gsp N0), decreasing the percentage of dead pixels from 44% to 43% and to 40%. Fig. 6 demonstrates that, in inferior slices, shortening TE effectively increases BS and reduces signal loss in the orbitofrontal cortex but decreases BS in areas not affected by susceptibility gradients. The average percentage of dead voxels in the orbitofrontal cortex decreases from 67%±15% to 39%±16% to 16%±4% in all subjects when TE is reduced from 40 ms to 30 ms to 20 ms. A calculation of the average BS in the orbitofrontal cortex in all subjects reveals that BS is significantly enhanced when TE increases with the slice number (Fig. 7). Optimal TE values simulated for all subjects are depicted in Fig. 8. The BS is maximal for increasing TE values between 17±2 ms (slice 1) and 39±2 ms (slice 9) in foot-to-head direction. In areas not affected by susceptibility gradients, an optimal TE value of 46±3 ms was measured, which is in good agreement with cortical values of about 40–50 ms [32]. 3.2. FMRI experiment with focus on orbitofrontal cortex activation Fig. 9 depicts two inferior slices acquired with the standard and the modified EPI sequence. The orbitofrontal cortex clearly shows less signal loss when TE is reduced from 27 ms to 22 ms (Fig. 9). The single-subject fMRI analysis shows orbitofrontal cortex activation in 5 out of 12 participants when the stan-

Fig. 3. Echo time dependence of the BS in the presence of typical susceptibility gradients occurring in the orbitofrontal cortex. The maximum is normalized to 1.

206

S. Domsch et al. / Magnetic Resonance Imaging 31 (2013) 201–211

Fig. 4. Simulation of maximum echo times before signal loss occurs due to macroscopic susceptibility gradients along the EPI phase encoding direction based on Eq. (8). As can be seen, the slope depends critically on the parameter Q defined in Eq. (7).

dard EPI sequence was used (Table 2). With the modified EPI sequence, activation was detected in all participants, the average number of activated voxels increased from 87±44 to 549±83, and the maximal t value, averaged over all subjects, increased from 4.4±0.3 to 5.4±0.3. At a significance level of α≤10 −4 (uncorrected), a secondlevel one-sample t-test revealed an average number of 55 voxels to be activated in the orbitofrontal cortex using the standard EPI sequence (Fig. 10). At the same threshold, 553 voxels were activated when the modified EPI sequence was used. Activation in inferior and superior slices acquired at echo times below and, respectively, above 27 ms was clearly more robust. Slices acquired at echo times of about 27 ms showed similar activation

in both EPI data sets. However, parietal activation is stronger at TE values above 27 ms, showing superiority of the modified EPI sequence. Comparing statistical inferences of both EPI sequences in a second-level paired t-test at a significance level of α≤10 −3 (uncorrected), stronger activation was detected in 303 (0) voxels of the orbitofrontal cortex when the modified (standard) EPI was used (Fig. 11).

4. Discussion The goal of the present work was to demonstrate that the adjustment of echo times in single EPI slices improves BS in

Fig. 5. BS maps in the presence of in- and through-plane gradients depicted for different echo times. The reference point at (0|0) is indicated by the white crosshair. The BS maximum is normalized to 1 and values below 10 % are set to 0. This simulation is based on Eq. (8) and Eq. (10). The following sequence parameters were assumed: T2*=25 ms, bandwidth=1578 Hz, TA=30 ms, FOV=2202 mm2, slice thickness=3 mm.

S. Domsch et al. / Magnetic Resonance Imaging 31 (2013) 201–211

207

Fig. 6. BS maps obtained from a single subject and depicted for different echo times. The orbitofrontal cortex is marked by the ROI.

the orbitofrontal cortex. First, BS simulations were performed to approximate optimal echo times for single EPI slices. Second, an EPI sequence with slice-dependent echo times was compared to an optimized EPI sequence with constant short echo time during an event-related fMRI experiment with focus on the orbitofrontal cortex. Both simulations and functional results indicate that BS is increased when slice-adjusted echo times are used. In the inferior orbitofrontal cortex, gradient strengths more than 250 μTm −1 were observed, which are in agreement with the findings of DePanfilis et al. [33]. It was demonstrated that even moderate susceptibility gradients in the order of 150 μTm −1 shift the maximal BS to echo times clearly below 30 ms. Both BS simulations and actual EPI measurements confirmed that a TE value of about 20 ms significantly reduces signal dropouts and increases BS in inferior slices of the orbitofrontal cortex. Further, we observed that T2* -relaxation times increase and field gradients decrease in foot-to-head direction. Therefore, short TE values of about 20 ms decrease the BS in superior slices of the orbitofrontal cortex. Accordingly, the numerically calculated optimal echo times averaged over all

subjects increased with the slice number and were in the range of 17±2 ms and 39±2 ms. However, echo times below 20 ms are not realizable using an EPI sequence with standard imaging parameters (e.g. matrix size, bandwidth, partial Fourier, parallel imaging factor, etc.). But using a short TE value of 20 ms potentially preserves BS in the presence of strong in- and throughplane gradients of more than 300 μTm −1. An important case-by-case analysis must be made with respect to the polarity of the in-plane gradients as described by Deichmann et al. [21,34]. Our simulations showed that a TE value of 22 ms is adequate to avoid signal loss caused by negative in-plane gradients of more than 300 μTm −1, whereas an echo time reduction is ineffective to recover signal loss due to positive in-plane gradients. Even positive in-plane gradients of only 80 μTm −1 cannot be compensated via TE reduction because the gradient echo is rapidly shifted outside the acquisition window with increasing gradient strengths. We showed that an EPI with slice-dependent echo times between 22 ms and 37 ms significantly improves statistical inference in the orbitofrontal cortex in both on the single-

Fig. 7. Average BS in ascending axial slices of the orbitofrontal cortex depicted for different echo times. The error bars denote the standard error of the mean.

208

S. Domsch et al. / Magnetic Resonance Imaging 31 (2013) 201–211

Fig. 8. Optimal echo times in ascending axial slices of the orbitofrontal cortex and depicted for six healthy participants.

subject level and on the group level compared to an optimized standard EPI with short TE of 27 ms. The effectiveness of the modified EPI sequence might be further improved by changing the slice tilt to reduce in-plane gradients with positive signs as we have observed in the orbitofrontal cortex. It has been previously shown that a tilt

angle of 30° from the axial towards the coronal orientation yields the best choice to minimize signal dropouts and thus to increase the BS in the orbitofrontal cortex [34]. However, it should be noted that, due to the shape of the brain, a transversal slice orientation yields time-efficient volume coverage. The number of slices would potentially have to be increased when tilting the slice orientation away from a transversal toward a coronal orientation [34], which would be detrimental for the temporal resolution, which is particularly important in event-related fMRI studies. It has been experimentally demonstrated that the phase encoding scheme significantly affects susceptibility-induced signal losses in EPI [33]. Hence, a slice-dependent change of the

Table 2 Single-subject fMRI results shown for the standard and the modified EPI sequence Subject

1 2 3 4 5 6 7 8 9 10 11 12 Mean Fig. 9. Representative axial slices acquired at TE of 27 ms (top) and about 22 ms (bottom). Signal dropout in the orbitofrontal cortex is marked by the arrow.

Standard EPI

Modified EPI max

N

T

201 108 526 0 67 0 0 0 77 0 50 0 87±44

3.8 4.4 6.3 3.6 5.0 3.4 4.4±0.3

N

Tmax

289 359 609 461 127 1054 377 695 630 243 873 866 549±83

4.4 4.8 6.0 5.1 3.9 6.5 4.6 4.9 5.9 4.8 6.4 7.3 5.4±0.3

N and Tmax denote the number of significant voxels and maximal t-values detected in the orbitofrontal cortex. The results are displayed at a threshold of α≤5·10−3 (uncorrected) and a minimal cluster size of 15 voxels. The error denotes the standard error of the mean.

S. Domsch et al. / Magnetic Resonance Imaging 31 (2013) 201–211

209

Fig. 11. Second-level paired t-test result overlaid on structural T1-weighted image. The results are displayed at a threshold of α≤10−3 (uncorrected) and a minimal cluster size of 15 voxels. Voxels above the threshold are stronger activated in the modified compared to the standard EPI data sets. The left hemisphere is displayed on the left.

Fig. 10. Second-level one-sample t-test results overlaid on structural T1weighted images and depicted for the standard (A) and the modified EPI (B). The parameter images were aligned along the AC–PC line to compare the activation in single slices acquired at distinct echo times. The results show orbitofrontal and parietal activation thresholded at α≤10−4 (uncorrected). The minimal cluster size was set to 15 voxels. The left hemisphere is displayed on the left.

phase encoding gradient direction additional to the TE variation could further improve fMRI in critical brain areas without affecting the temporal resolution. It seems implausible that a reduction of TE in the standard EPI sequence could outweigh the benefits of using slice-adjusted echo times. Using an echo time below 27 ms would mitigate susceptibility artifacts in inferior slices. But it has to be noted that activation detected with the modified EPI sequence was more robust in inferior and superior slices acquired at echo times below and above 27 ms. Only slices acquired at approximately 27 ms showed similar activation in both EPI data sets. It is very likely that this

effect will occur for any fixed echo time. Moreover, most psychological studies require whole-brain coverage. For a shorter echo time, BS would be nonoptimal in cortical brain areas not affected by susceptibility gradients where robust activation is detected between 30 and 50 ms [32]. Parietal activation was seen in both EPI data sets, whereas more significant voxels were found in superior slices scanned at TE values above 27 ms. This finding agrees well with the fact that parietal regions are not exposed to susceptibility gradients, and therefore, the BOLD peak should be above 30 ms. It has to be discussed whether more detected activity equates to more accurate detection of neuronal activity. Large vessels (draining veins) in the vicinity of an activated region are a known source of false-positive voxels in gradient-echo fMRI where the BOLD signal consists of an intravascular and a dominating extravascular component. Duong et al. [35] reported that for high magnetic field strengths (i.e. 4 T and 7 T), the weak intravascular BOLD signal contribution of large vessels decreases with longer echo times. Whether this effect occurred at field strength of 3 T and whether it led to more false-positive voxels in inferior slices acquired with the modified EPI sequence only or in inferior slices acquired with both sequences remain unknown. However, in the context of research with several groups such as patients and controls, this issue is of minor importance as effects of systematic errors like false-positive voxels cancel out. Yet, subtle differences might be detected more easily using a sequence more sensitive in the detection of signal change. Despite a rather large variability of magnetic field gradients between different subjects, the simulation of optimal

210

S. Domsch et al. / Magnetic Resonance Imaging 31 (2013) 201–211

echo times in single slices showed consistent results over all subjects. Still, considering that susceptibility gradients will depend notably on the person's orientation inside the magnet, which cannot be identical from day to day, we propose to implement T2*-fitting and gradient field mapping tools on the MR scanner for individual echo time optimization directly before the fMRI experiment as previously suggested by Stoecker et al. [25]. Individual gradient field maps could then additionally be used to remove EPI images distortions [36,37] to increase spatial accuracy especially in regions prone to susceptibility artifacts. Besides echo time optimization in single EPI slices, multi-echo acquisition methods [38–42] pose an alternative for whole-brain fMRI studies since the BOLD contrast is preserved in all areas. But it has to be noted that they compromise temporal resolution since the EPI readout train is prolonged. This lack of temporal resolution is detrimental for event-related fMRI studies and could only be compensated by decreasing spatial resolution since spatial coverage is indispensable in whole-brain fMRI. Recently, it has been shown that susceptibility-induced BOLD signal dropouts and distortions are reduced when spatial resolution is increased in the EPI readout direction, which does not affect temporal resolution [43]. Thus, this strategy could be combined with slice-dependent echo times to further reduce susceptibility artifacts in subcortical brain areas. In summary, we have shown that a TE value of about 20 ms reduces signal dropouts and increases BS in inferior slices of the orbitofrontal cortex. Further, in superior slices, maximal BS is reached at echo times of about 40 ms, supporting the idea of separately adjusting the echo time in single slices. Finally, during an eventrelated whole-brain fMRI experiment, a modified EPI sequence with slice-dependent echo times yielded superior statistical inference in the orbitofrontal cortex compared to a standard EPI sequence with reduced echo time. We conclude that an EPI with slice-dependent echo times between 20 ms and 40 ms may be a valuable tool for event-related whole-brain fMRI studies with focus on the orbitofrontal cortex.

[4]

[5]

[6]

[7]

[8]

[9] [10]

[11]

[12]

[13]

[14] [15] [16]

[17]

[18]

[19]

Acknowledgments

[20]

This study was supported by the Deutsche Forschungsgemeinschaft (SFB 636/C1, C6 and Z3).

[21]

[22]

References [1] Ogawa S, Lee TM, Kay AR, Tank DW. Brain magnetic resonance imaging with contrast dependent on blood oxygenation. Proc Natl Acad Sci USA 1990;87(24):9868–72. [2] Ogawa S, Lee TM, Nayak AS, Glynn P. Oxygenation-sensitive contrast in magnetic resonance image of rodent brain at high magnetic fields. Magn Reson Med 1990;14(1):68–78. [3] Diekhof EK, Falkai P, Gruber O. Functional neuroimaging of reward processing and decision-making: a review of aberrant motivational and

[23]

[24]

affective processing in addiction and mood disorders. Brain Res Rev 2008;59(1):164–84. Weber MA, Gunther M, Lichy MP, Delorme S, Bongers A, Thilmann C, et al. Comparison of arterial spin-labeling techniques and dynamic susceptibility-weighted contrast-enhanced MRI in perfusion imaging of normal brain tissue. Invest Radiol 2003;38(11):712–8. Bauer WR, Nadler W, Bock M, Schad LR, Wacker C, Hartlep A, et al. The relationship between the BOLD-induced T(2) and T(2)(*): a theoretical approach for the vasculature of myocardium. Magn Reson Med 1999;42(6):1004–10. Bauer WR, Nadler W, Bock M, Schad LR, Wacker C, Hartlep A, et al. Theory of the BOLD effect in the capillary region: an analytical approach for the determination of T2 in the capillary network of myocardium. Magn Reson Med 1999;41(1):51–62. Schroder H, Bongers A, Schad LR. Testing of model assumptions of MR oxygen extraction fraction mapping using a phantom study. Z Med Phys 2006;16(4):275–84. Czervionke LF, Daniels DL, Wehrli FW, Mark LP, Hendrix LE, Strandt JA. Magnetic susceptibility artifacts in gradient-recalled echo MR imaging. Am J Neuroradiol 1988;9:1149–55. Mansfield P. Multi-planar image formation using NMR spin echoes. J Phys C: Solid State Physics 1977;10:L55–8. Schroder J, Wenz F, Schad LR, Baudendistel K, Knopp MV. Sensorimotor cortex and supplementary motor area changes in schizophrenia. A study with functional magnetic resonance imaging. Br J Psychiatry 1995;167(2):197–201. Bottmer C, Bachmann S, Pantel J, Essig M, Amann M, Schad LR, et al. Reduced cerebellar volume and neurological soft signs in first-episode schizophrenia. Psychiatry Res 2005;140(3):239–50. Ordridge RJ, Gorell JM, Deniau JC, Knight RA, Helpern JA. Assessment of relative brain iron concentration using T2-weighted and T2*-weighted MRI at 3 tesla. Magn Reson Med 1994;32:335–41. Frahm J, Merboldt KD, Hänicke W. Direct flash MR imaging of magnetic field imhomogeneities by gradient compensation. Magn Reson Med 1988;6:474–80. Constable RT, Spencer DD. Composite image formation in z-shimmed functional MR imaging. Magn Reson Med 1999;42:110–7. Cho ZH, Ro YM. Reduction of susceptibility artifact in gradient-echo imaging. Magn Reson Med 1992;23:193–200. Konstandin S, Heiler PM, Scharf J, Schad LR. Comparison of selective arterial spin labeling using 1D and 2D tagging RF pulses. Z Med Phys 2011;21(1):26–32. Stenger VA, Boada FE, Noll DC. Three-dimensional tailored RF pulses for the reduction of susceptibility artifacts in T2*-weighted functional MRI. Magn Reson Med 2000;44:525–31. Preston AR, Thomas ME, Ochsner KN, Cooper JC, Glover GH. Comparison of spiral-in/out and spiral-out BOLD fMRI at 1.5 and 3 T. Neuroimage 2004;21:291–301. Glover GH. 3D Z-shim method for reduction of susceptibility effects in BOLD fMRI. Magn Reson Med 1999;42:290–9. Nagel AM, Laun FB, Weber MA, Matthies C, Semmler W, Schad LR. Sodium MRI using a density-adapted 3D radial acquisition technique. Magn Reson Med 2009;62(6):1565–73. Deichmann R, Josephs CH, Corfield CDR, Turner R. Compensation of susceptibility-induced BOLD sensitivity losses in echo-planar fMRI imaging. Neuroimage 2002;15:120–35. Cools R, Clark L, Owen AM, Robbins TW. Defining the neural mechanisms of probabilistic reversal learning using event-related functional magnetic resonance imaging. J Neurosci 2002;22(11):4563–7. Evers EA, Cools R, Clark L, van der Veen FM, Jolles J, Sahakian BJ, et al. Serotonergic modulation of prefrontal cortex during negative feedback in probabilistic reversal learning. Neuropsychopharmacology 2005;30(6):1138–47. Morawetz C, Holz P, Lange C, Baudewig J, Weniger G, Irle E, et al. Improved functional mapping of the human amygdala using a standard functional magnetic resonance imaging sequence with simple modifications. Magn Reson Imaging 2008;26(1):45–53.

S. Domsch et al. / Magnetic Resonance Imaging 31 (2013) 201–211 [25] Stocker T, Kellermann T, Schneider F, Habel U, Amunts K, Pieperhoff P, et al. Dependence of amygdala activation on echo time: results from olfactory fMRI experiments. Neuroimage 2006;30(1):151–9. [26] Lipschutz B, Friston KJ, Ashburner J, Turner R, Price CJ. Assessing study-specific regional variations in fMRI signal. Neuroimage 2001;13(2):392–8. [27] Gati JS, Menon RS, Ugurbil K, Rutt BK. Experimental determination of the BOLD field strength dependence in vessels and tissue. Magn Reson Med 1997;38(2):296–302. [28] Windischberger C, Robinson S, Rauscher A, Barth M, Moser E. Robust field map generation using a triple-echo acquisition. J Magn Reson Imaging 2004;20:730–4. [29] Peters AM, Brookes MJ, Hoogenraad FG, Gowland PA, Francis ST, Morris PG, et al. T2* measurements in human brain at 1.5, 3 and 7 T. Magn Reson Imaging 2007;25:748–53. [30] Griswold MA, Jakob PM, Heidemann RM, Nittka M, Jellus V, Wang J, et al. Generalized autocalibrating partially parallel acquisitions (GRAPPA). Magn Reson Med 2002;47(6):1202–10. [31] Linke J, Kirsch P, King AV, Gass A, Hennerici MG, Bongers A, et al. Motivational orientation modulates the neural response to reward. Neuroimage 2010;49(3):2618–25. [32] Fera F, Yongbi MN, van Gelderen P, Frank JA, Mattay VS, Duyn JH. EPI-BOLD fMRI of human motor cortex at 1.5 T and 3.0 T: sensitivity dependence on echo time and acquisition bandwidth. J Magn Reson Imaging 2004;19(1):19–26. [33] De Panfilis C, Schwarzbauer C. Positive or negative blips? The effect of phase encoding scheme on susceptibility-induced signal losses in EPI. Neuroimage 2005;25(1):112–21. [34] Deichmann R, Gottfried JA, Hutton C, Turner R. Optimized EPI for fMRI studies of the orbitofrontal cortex. Neuroimage 2003;19: 430–41.

211

[35] Duong TQ, Yacoub E, Adriany G, Hu X, Ugurbil K, Kim SG. Microvascular BOLD contribution at 4 and 7 T in the human brain: gradient-echo and spin-echo fMRI with suppression of blood effects. Magn Reson Med 2003;49(6):1019–27. [36] Jezzard P, Balaban RS. Correction for geometric distortion in echo planar images from B0 field variations. Magn Reson Med 1995;34: 65–73. [37] Hutton C, Bork A, Josephs O, Deichmann R, Ashburner J, Turner R. Image disstortion correction in fMRI: a quantitative evaluation. Neuroimage 2002;16:217–40. [38] Poser BA, Norris DG. Investigating the benefits of multi-echo EPI for fMRI at 7 T. Neuroimage 2009;45(4):1162–72. [39] Poser BA, Versluis MJ, Hoogduin JM, Norris DG. BOLD contrast sensitivity enhancement and artifact reduction with multiecho EPI: parallel-acquired inhomogeneity-desensitized fMRI. Magn Reson Med 2006;55(6):1227–35. [40] Posse S, Wiese S, Gembris D, Mathiak K, Kessler C, Grosse-Ruyken ML, et al. Enhancement of BOLD-contrast sensitivity by single-shot multi-echo functional MR imaging. Magn Reson Med 1999;42(1): 87–97. [41] Speck O, Hennig J. Functional imaging by I0- and T2*-parameter mapping using multi-image EPI. Magn Reson Med 1998;40(2): 243–8. [42] Weiskopf N, Klose U, Birbaumer N, Mathiak K. Single-shot compensation of image distortions and BOLD contrast optimization using multi-echo EPI for real-time fMRI. Neuroimage 2005;24(4): 1068–79. [43] Weiskopf N, Hotton C, Josephs O, Turner R, Deichmann R. Optimized EPI for fMRI studies of the orbitofrontal cortex: compensation of susceptibility-induced gradients in the readout direction. Magn Reson Mater Phy 2007;20:39–49.