The modulation in k-space data which causes the motion artifacts in MRI occurs only in one direction (i.e., the phase encode direction). Previous works have shown that combining two MRI images acquired over the same field-of-view (FOV) with their readout and phase encode directions interchanged can reduce motion artifacts.^1,2 However, these studies used the same slice-select directions. The concept of using two MRI images acquired over the same region of interest using different slice-select directions (i.e., orthogonal views) has not been explored. In most small animal imaging studies both long axis (coronal or sagittal) and short axis (axial) images of the region of interest are obtained. It can be useful to study whether combining these orthogonal images will help in reducing the motion artifacts. The goal of this study is to explore whether combining two orthogonal views obtained with different slice-select directions could improve the quality of the image by reducing artifacts due to motion. The advantages of this method are that no a priori knowledge about the type of motion is required for their implementation and no external hardware is needed and it doesn’t increase the acquisition time over conventional scans. The evaluation of this motion correction algorithm was performed in phantom and in vivo mouse images.

The long axis and short axis images have motion artifacts in different (phase encoding) directions, and have different in-plane FOVs and slice-select directions. The flowchart for the implementation of the algorithm is shown in the Figure 1. Consider axial image stack $$$I_a(X,Y,z)$$$ and coronal image stack $$$I_c(x,Y,Z)$$$ acquired orthogonally to each other. To correct for motion in the axial view, the coronal view is upsampled along the x-axis and the axial view is upsampled along the z-axis. $$$\hat{I_a}(X,Y,Z)$$$ and $$$\hat{I_c}(X,Y,Z)$$$ represents the upsampled axial and coronal image stacks respectively. The complex conjugate product is obtained using the two upsampled views. The square root of the magnitude of the complex conjugate can then be downsampled in either direction to obtain a motion corrected image.

Simulated motion in biological phantom: A paraformaldehyde fixed E17.5 embryo was used as a biological phantom as it possesses anatomic structures similar to those observed in live animals but does not suffer from motion artifacts observed for in vivo imaging. Coronal and axial views were acquired using T1-weighted FLASH imaging sequence (T_R=519.5ms, T_E=4ms, FA=30.0, FOV =2.2*2.2cm, slice thickness=0.26mm, matrix=512*512, navgs=1, acquisition time=3min, number of contiguous slices=46) in 500 MHz 11.7T magnet (Bruker Biospin). Artificial motion as observed in in vivo case is simulated along the phase encode direction in both the images.³ The bulk misalignment is simulated in these images by rotating the coronal image by 3°.⁴

In vivo experiment: For quantification of fat volume, coronal and axial image stacks of the abdominal region were acquired using respiratory triggered RARE T1 weighted sequence (T_R=1200 ms, T_E=7.5ms, FA=180.0, FOV=3*3cm, matrix=256*256, slice thickness=0.75 mm, acquisition time=12 min) in 400 MHz 9.4T magnet. Both qualitative (visual inspection) and quantitative measures are used for the evaluation of the algorithm.

Simulated motion in biological phantom: In Figure 2a motion artifact caused by the simulated motion is evident in spinal cord region of coronal image. Figure 2c shows the result of motion correction algorithm which has less ghosting artifacts when compared to Figure 2a.

In vivo experiment: The coronal image of in vivo mice abdomen with visible motion and the result of the correction algorithm are shown in Figure 3. The ghosting caused by motion (highlighted in the coronal image by red arrow) is reduced by the correction algorithm. Due to the presence of motion in the coronal view, the anatomical details inside the kidneys are blurred and those details are visible in the upsampled and rotated axial image stack, and by using the correction algorithm, some of these structures are recovered (highlighted by white arrow). The correction algorithm improved signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR) and mean edge width in both the experiments [Table 1,2].

The effective resolution achieved in the corrected image is lower than the original image. This resolution compromise is due to the combination of the in-plane view from an image stack with the through-plane view of the orthogonal image stack. This motion correction algorithm is useful when orthogonal images are already acquired as part of the normal workflow of imaging studies, and the original image data have moderate motion artifacts. The results show that it is advantageous to use the available orthogonal image(s) to improve the image quality by reducing ghosting artifacts caused by motion.