Acoustic radiation force imaging (ARFI) is a non-invasive tool to localize the ultrasound (US) focus during high intensity focused ultrasound (HIFU) therapies. In MRI-ARFI the US-induced tissue displacement is detected with motion encoding gradients (MEG) with high amplitude and long duration. The MEG can lead to artifacts which are mainly caused by bulk motion and eddy currents. An effective correction method applied during ARFI post processing has been presented¹, which extracts a background phase for later subtraction after masking out the focal spot. This method requires manual selection of the focal spot, which increases post-processing time, introduces operator bias, and is severely hampered when the uncorrected displacement data exhibit strong artifacts.

In this work we propose an alternative method to correct artifacts in ARFI images that acquires a short FID navigator signal which is used later on for correction. The navigator correction does not require any additional interaction and can be incorporated into the MR-system’s image reconstruction process.

For the ARFI measurements, a spoiled gradient echo (FLASH) sequence was modified to include bipolar MEGs with interleaved polarities². An FID navigator that collects only 32 sampling points (duration 160 μs) was inserted after the slice selection rewinder gradient prior to any encoding steps (Fig. 1), i.e., the navigator signal does not contain any spatial encoding and represents the integral signal over the entire excited volume.

With this sequence an ARFI experiment was conducted at a clinical 3T MR system (Siemens PRISMA) in a milk-gelatin phantom³ (gelatin bloom value 175). The slice-selection direction and the direction of the MEG were parallel to the direction of ultrasound, and the following parameters were used: TE/TR = 20/60 ms, resolution 1.2×1.2×5 mm³, MEG duration 10 ms, MEG amplitude 40 mT/m. The fixed-focus HIFU transducer (1.7 MHz, focal length 6 cm) was triggered by the MR pulse sequence. Sonication started 4 ms after MEG start and lasted for 6 ms until the end of MEG.

During image reconstruction the 32 data points within the FID navigator data of each repetition m were averaged, and the phase φ_m was calculated. Since artifacts are mainly caused by phase variations, the mean phase over all repetitions (N – total number of repetitions) was subtracted,

$$\triangle\varphi_m = \varphi_m - \frac{\sum_{m=1}^N\varphi_m}{N}, (1)$$

and the remaining phase difference Δφ_m was subtracted in complex domain from the corresponding k-space data acquired during the m-th imaging readout. Corrected imaging data was Fourier transformed and ARFI displacement maps were reconstructed from the phase difference between the two different MEG polarities.

The measured phase difference is shown as a function of the k-space line index in Fig. 2. As can be seen, the phase change during data acquisition is on the order of 20 mrad with standard deviation of 5 mrad. In Fig. 3 a comparison of magnitude images and ARFI displacement maps before and after correction is shown. Without correction, signal intensity varies substantially, and even signal voids are present, whereas with correction barely any artifacts can be seen. Furthermore, with correction the SNR near the focal spot increases by a factor of about 2.5. In the ARFI maps the effect of the correction is even more pronounced: without correction, the ARFI map does not show any distinct displacement information, whereas with correction a clear and well-defined displacement is visible at the location of the US focus. At the center of the focus, the measured displacement with FID correction amounted to 3μm.In this work, a short FID navigator is introduced into a FLASH sequence and used for artifact correction in ARFI displacement images. Without correction the artifacts are very pronounced and entirely overlay the focal region, so that image-based correction methods as proposed by Chen et al¹ become difficult, because the focal spot cannot be unambiguously identified. After navigator correction, the focal spot is clearly visible and displacements can be reliably detected. The proposed FID navigator has to be acquired prior to any encoding gradients, so that the pulse sequences become slightly longer than their conventional counterparts. However, in this implementation the additional time per TR interval was only about 1 ms, which is a small fraction and extends the overall acquisition time by only 1 s. As the navigator correction method does not require any user input, it might be preferable over manual correction methods and could be implemented in nearly any ARFI sequence without substantially prolonging the acquisition duration.