A limitation of fMRI in the investigation of affective symptomatology is signal dropout in EPI, especially pronounced in ventromedial prefrontal cortex (vmPFC), which is a region of special interest in affective neuroscience and psychiatry¹. Standard fMRI at 3 Tesla leads to significant signal and BOLD sensitivity loss in vmPFC and other orbitofrontal and temporal regions affected by signal dropout². Multi-echo EPI (MEPI) is known to have several advantages^3-9 over EPI. To date, studies have indicated^3-8 tends better activation detection, also in vmPFC. However, to our very best knowledge, no study has reported a statistical comparison of MEPI with the standard fMRI using an appropriate task that specifically elicits activity in vmPFC.To test if MEPI, with the weighted sum combination approach⁵, outperforms standard EPI in detecting task induced activation in vmPFC using a task that robustly activates this area.

Thirty-six young and healthy subjects (age 25.0±3.2 years, 17 female) were included, four data sets had to be excluded (e.g. incomplete data).

In the fear-conditioning task, three visual stimuli (geometric shapes) were presented 10 times each with 12-14s inter-stimulus-intervals. Two of the stimuli were paired with mild electric shocks at stimulus offset, administered to the wrist at an individually titrated level that was uncomfortable but not painful. The two conditioned stimuli were paired with shocks in 100% (100% reinforcement) and 60% (60% reinforcement) of all stimulus presentations and can be summarized as CS+. The third stimulus was never paired with the shock (safety stimulus – CS-). The comparison CS->CS+ generally shows the vmPFC¹⁰. The reverse contrast shows an increase of activity in dorsal anterior cingulate (dACC) and bilateral insula¹⁰.

The data were acquired on a GE Discovery MR750 3T scanner with a 32-channel head coil. The fMRI data were acquired using a MEPI pulse sequence with 3 echoes (E1, E2, E3) in an orientation aligned with the AC-PC. Acquisition parameters were: 64x64 matrix, 3.4mm isotropic resolution, 36 slices, acceleration-factor 2, TR=2.56s, TE1/TE2/TE3=12/29/46ms. Additionally, a 3D high-resolution T1w image was acquired.

The fMRI analysis was performed with SPM8. Data were first corrected for slice timing differences and realigned. At this stage, the weighted sum combination⁵ was calculated resulting in a weighted echo Ec. Then, the MEPI data were co-registered to the 3D-T1, the 3D-T1 was normalized to the MNI space and the spatial transformation applied to the fMRI data. To be able to statistically compare the methods, a linear detrending followed by a z-score transformation was applied to all voxel time courses of MEPI (E1, E2, E3 and Ec) before spatial smoothing (Gaussian FWHM 6x6x6 mm3).

First level models comprised hemodynamic response function convolved stimulus regressors, motion coefficients as nuisance regressors were estimated, and the contrast CS ->CS+ was computed. At the second level a full factorial analysis of variance was used to compare the different methods (i.e. E1, E2, E3, Ec) with a repeated measures factor of four levels (method). The primary contrast of interest was the comparison of E2 (standard fMRI) with Ec. To control for multiple tests, familywise error (FWE) correction were used.

All methods showed the expected task activation but at different levels of statistical significance (see Figure 1). Figure 2 shows the main effect of method to highlight the regions in which the methods differ, together with the contrasts estimates for three exemplary differentially activated regions. Figure 3 depicts the Ec>E2 (standard EPI) comparison where both the vmPFC and the bilateral insula demonstrate significant differences. Figure 4 illustrates the contribution (the weights) of each echo in the brain regions of interest. Figure 5 shows the extension of the clusters in Ec compared to the others while showing the regions affected by the dropout. Our results show that MEPI outperforms standard EPI in detecting task-induced vmPFC activation, even with a basic weighted sum. Further, results show the contribution of each echo in different brain regions (Figure 4, 5). It is worth noting that MEPI does not only improve the contrast in brain regions affected by signal dropout where E1 contributes significantly, but also in homogeneous brain regions where E3 contributes significantly because it is closer to the optimal TE than E2. This is in accordance with previous studies^2-10 and a strong advantage of MEPI. Recent methodologies like ME-ICA^7-8, multiband and coil development are expected to take these results further. We have demonstrated that, in a fear-conditioning task, MEPI allow an overall better detection of fMRI activity in vmPFC and other involved brain regions. Future work will include advanced analyses^7-8, other tasks and resting state fMRI.