In recent years, ultrashort echo-time imaging has been explored for the study of short T2* tissues inside the body ^1,2. While the technique has been implemented in a number of different pulse sequences, most ultrashort TE sequences are based on 3D radial center-out k-space sampling that involves the acquisition of data along large numbers of evenly spaced radial projections. These sequences include UTE ³, PETRA ⁴ and ZTE ⁵, with PETRA and ZTE being different from UTE in that the readout gradient is switched on before RF excitation, resulting in silent scanning and shorter encoding time. Though these sequences have been shown to be potentially useful for important clinical and research applications, they are susceptible to various image artifacts that may significantly degrade the image quality. Our purpose is to describe the appearances, causes and mitigations of some potential artifacts.

The images used for illustrations in this paper were acquired using UTE and PETRA sequences obtained from Siemens Healthcare as work-in-progress sequences.

1) Image blurring due to long readout length

The readout length in each radial projection equals the time between successive data points multiplied by the number of acquired data points in the projection. It is inversely related to the readout bandwidth per pixel. While a longer readout length allows higher image resolution, it may lead to image blurring due to T2* decay. The optimal readout length is approximately the T2* value of the desired tissue ⁶. Severe image blurring can occur if the readout length is several times longer than T2*, which may be reduced by increasing the bandwidth per pixel (Fig. 1).

2) Image blurring and non-uniform signal due to high readout gradient field

For sequences such as PETRA and ZTE that have the readout gradient field turned on before RF excitation, the RF pulse must fully and uniformly cover the large bandwidth generated by the gradient ⁷. Otherwise, there will be reduced excitation to the outer FOV at high spatial frequencies, causing image blurring and signal non-uniformity in those regions (Fig. 2). Therefore, the readout gradient should be set low to avoid these artifacts. However, for a given image resolution, a lower gradient translates to lower bandwidth and longer readout length that may lead to its own artifacts as mentioned earlier. Consequently, a trade-off among the related parameters may be needed.

3) Imaging blurring due to insufficient number of radial projections

Insufficient number of radial projections leads to image blurring (Fig. 3). In radial sampling, the image resolution is determined by the number of sample points per radial projections and the number of radial projections. The reconstructed image matrix size does not necessarily represent the intrinsic image resolution. To satisfy the Nyquist criterion for 3D radial sampling, large numbers of radial projections are often required ^8,9. For example, 237,739 radial projections are needed for a 256³ image matrix ⁹. This may result in long scan time, even with the minimum TR. However, depending on the specific applications, satisfactory results may be obtained by acquiring only a fraction of the theoretical number of projections ^4,9.

4. Artifacts caused by off-centered imaging

For applications that require the FOV to be off-centered in the magnet, the shift from iso-center is corrected through image reconstruction. However, there are associated artifacts that increase with the amount of off-center shift, and they become severe if the shift is more than 50% of the FOV in length (Fig. 4). To minimize the artifacts, the FOV should be kept as close to iso-center as possible.

5. Wrap-around artifacts

In order to achieve the shortest possible TE, ultrashort TE sequences often employ 3D nonselective excitation that is vulnerable to wrap-around artifacts caused by signal outside the FOV. Sequences such as UTE utilize radial data oversampling and avoid wrap-around artifacts in all directions. Sequences that do not use oversampling or selective excitation, such as PETRA, are subjected to wrap-around artifacts (Fig. 5a). Besides peripheral body parts, RF coil materials can also produce signals (Fig. 5b) that may wrap around into the imaged tissues. When phased array coils, such as a spine coil, are used with these sequences, wrap-around artifacts may be controlled by limiting the signal coverage through coil element selection.

While this paper does not cover all possible artifacts associated with ultrashort TE imaging, the artifacts discussed represent some of the commonly encountered ones. An understanding of these artifacts will help in the setting up of protocols and the identification of related imaging problems. This paper should benefit the users of ultrashort TE imaging techniques.