At higher field strengths (3T and above), T2*-weighted images acquired with gradient-echo (GRE)-based sequences, e.g. FLASH, SWI and BOLD EPI, are often impaired by local signal loss induced by susceptibility effects. In the human brain, these artifacts typically appear near the orbital frontal and inferior temporal cortex and affect the imaging of brain activity with fMRI or the diagnosis of diseases, e.g. stroke or hemorrhage. The effect worsens with increasing echo time (TE) and slice thickness, and is often dominated by the through-plane signal loss. Proposed methods (1, 2) mitigate through-plane dephasing using parallel transmission without increasing scan time or changing the slice coverage.

In this work, we implemented the pTX z-shim including B1 inhomogeneity correction (2) on a clinical 3T scanner equipped with two whole-body transmit channels. Furthermore, slice-specific delay-times were calculated automatically inline to ensure user-friendly operation as well as maximal signal gain in artifact regions. Finally, human studies were performed to demonstrate the potential benefit for clinical GRE and fMRI imaging.

To summarize the methods (1, 2), the excitation pulse is applied on separate transmit channels with a time shift to impose a linear “prewinding” phase to compensate the phase cancellation induced by the through-plane susceptibility. Our prototype sequences included an inline pTX pulse optimization (2) to mitigate the B1 inhomogeneity.

Five human subjects were studied on a 3T MAGNETOM Skyra (Siemens Healthcare, Erlangen, Germany) using prototype gradient-echo (FLASH) and EPI-based sequences with image resolutions of 0.7x0.7x6 mm3 and 1.7x1.7x6 mm3, respectively. Studies with standard RF excitation pulses (reference) were compared to slice-specific pTX z-shim pulses, which were calculated fully automatically from inline B0 and B1 maps and designed for maximal signal gain in artifact regions.

For both gradient-echo and EPI images, difference images (corrected vs. reference) were calculated in % of the reference mean intensity. Temporal-signal-to-noise ratio (tSNR) maps were generated from EPI time series of 100 repetitions (ratio of the mean signal intensity to the standard deviation of the noise: $$$ tSNR =\overline{S}_{d}/\sigma_{N} $$$). tSNR provides information on the detection ability for fMRI and is particularly suitable to highlight the benefit for fMRI applications. In all series, noise regions were masked and time series were detrended voxelwise by a second-order polynomial. To evaluate the actual signal or tSNR gain, respectively, artifact regions-of-interest (ROI) were defined in affected slices, and mean or maximum values were calculated.

Exemplary signal gain in FLASH studies is shown in Fig. 1. Some image slices acquired with the standard excitation pulse suffered from strong subject-induced susceptibility artifacts (left) in the frontal orbital and temporal cortex. Commonly, only few slices were affected (e.g. slices 7-11 of 20) and corrected by the pTX z-shim, and the rest remained untouched. In contrast, images with a slice-specific correction showed significantly signal gain in artifact regions at the cost of global SNR penalty. Difference images (right) revealed that a gain relative to the reference mean could be achieved up to 102% in the areas suffering from through-plane susceptibility. Fig. 2 summarizes the signal gain in GRE studies for all subjects. On average over all subjects, the frontal (temporal) orbital cortex had a maximal signal gain of 88+-13% (74+-21%) and a mean signal gain of 31+-11% (31+-12%).

Exemplary signal gain in gradient-echo EPI time series is shown in Fig. 3. A similar signal recovery in artifact regions compared to GRE could be achieved. In addition, tSNR maps from EPI time series (Fig. 4) show a remarkable increase with the pTX z-shim. A mean tSNR increase up to 121% in artifact regions can be seen from difference maps. Fig. 4 summarizes the tSNR gain for all subjects. On average, the frontal (temporal) orbital cortex had a mean tSNR gain of 88+-38% (86+-43%).

In the slices with signal and tSNR gains in artifact regions (ROI), a reduced SNR and fMRI detection power is visible outside the ROI (unaffected areas). In order to be able to adapt gains/losses for specific application protocols, an additional user factor to scale the pTX z-shim correction was implemented (data not shown).

The proposed method and implementation on a clinical scanner enables an easy-to-use and automated compensation for slices with local signal loss induced by susceptibility effects. Moreover, significant tSNR gain could be achieved in susceptibility-induced artifact regions in fMRI image series. Hence, this application is of high interest to enhance the visualization of brain activity and the diagnosis of diseases in former dark artifact regions.