title
home divider head and neck divider brain divider face divider spine divider chest divider abdomen divider genitourinary divider extremities
Appendix B: Image Quality - What We See and What We Can Do

SNR: Due to the relatively small size of the fetus, fetal MR is commonly limited by SNR. There are two factors that affect the SNR in an MR image: slice thickness and matrix size.
      The SNR varies directly with the size of the voxels in an image. For example, if the thickness of slices in a 2D image were halved, thereby doubling the resolution in the slice direction, the SNR of that image would be reduced by a factor of two. Most structures in the fetal body are well visualized with 4 mm slice thickness (Fig. 1). Small or thin structures surrounded by fluid may not be visible by MR (Fig. 2).
 
figure 1a figure 1b
 
Fig. 1. Slice thickness. A fetus at 19 weeks gestation with ventriculomegaly. In (A) slices were 4 mm thick. In (B) slices were 3 mm thick. Note increased signal but increased blur in (A) compared to (B).
 
figure 2
 
Fig. 2. Difficulty in visualizing thin structures surrounded by fluid Axial MR in a fetus with an L4 neural tube defect at 26 weeks gestation . Although the soft tissue defect is well visualized by MR (arrow), the sac covering the defect is not seen. The sac was visualized on US (not shown). Non-visualization of the sac on MR is due to the thin size of the sac wall and the partial volume averaging that occurs due to cerebrospinal fluid inside the sac and amniotic fluid outside of the sac.
 
With the exception of reducing slice thickness in a 2D acquisition, resolution in MR is usually increased by increasing the number of encoding steps (either phase encoding or frequency encoding steps) acquired in (at least) one direction (ie, increasing the matrix size). The SNR of an image varies as the square root of the number of encoding steps. If resolution in the frequency encode direction is increased by doubling the number of frequency encoding steps acquired (and keeping all other imaging parameters the same), there are two competing effects on the SNR of the resulting image. The halving of the voxel size results in a halving of the SNR, but the doubling of the number of encoding steps results in an increase in SNR by a factor of the square root of two. In combination these two effects result in a reduction in SNR of only a square root of two, rather than the factor of two that might be expected from the reduction in voxel size (Fig. 3). It is important to realize that on some magnets, the longer reconstruction times associated with large matrices will allow for more fetal motion between sequence acquisitions.
 
figure 3a figure 3b
 
Fig. 3. Sagittal views of the fetal head in fetus with tuberous sclerosis at 36 weeks gestation. Image (A) is taken with a 128 x 256 matrix and (B) is taken with a 256 x 512 matrix. Both images show the small nodules (subependymal tubers, arrows) projecting into the ventricle. Image (B) has better resolution than (A) but has the same diagnostic information
 
A final consideration in maximizing SNR is FOV. To ensure the best resolution possible, it is important to keep the FOV as small as possible. But unlike typical abdominopelvic imaging, wrap around artifact into the peripheral maternal anatomy is not a problem (Fig. 4).
 
figure 4
Fig. 4. A fetus with tuberous sclerosis at 34 weeks gestation. The oblique image plane with respect to maternal anatomy (patient imaged in lateral decubitus position) gives bright wrap-around artifact. This artifact does not overlie the area of interest in the fetal brain, as the subependymal tubers (arrows) are well visualized.
 
Patient body habitus and use of the surface coil: Since patients in the late stages of pregnancy tend to have larger and more protuberant abdomens that the typical nongravid patient, patient body habitus must be considered. Because the wall of the abdomen can come very close to the outside of the magnet bore, the surface coils generally used for abdominal and pelvic imaging may be placed closer to the body coil used for RF pulse transmission. The proximity of the surface coil array to the body coil can de-tune the surface coil, resulting in failure of the magnet to complete its prescan calibration steps. In these cases it will be necessary to remove the surface coils and use the body coil alone for imaging. Surface coils are helpful to increase signal to noise, but if a patient is too large to tolerate a surface coil and still fit in the magnet, imaging can be performed with the body coil (Fig. 5).
 
figure 5
 
Fig. 5. Coronal image of the maternal abdomen obtained with a body coil at 26 weeks gestation. This image was obtained during an examination for atypical abdominal pain performed with the body coil, rather than a phased array surface coil. The fetal anatomy is still well-visualized.
 
Signal inhomogeneity when using a phased array surface coil: It is often advantageous to use a phased array surface coil, instead of the magnets built in the body coil, for signal reception. Phased arrays give much better SNR than the body coil, in part because the coil array can be placed much closer to the anatomy of interest. However, unlike the body coil, the signal intensity of an image produced by phased array is not uniform and drops off with distance from the array. Thus, a phased array image will have very high signal intensity (SI) at the abominal wall, but the intensity will be significantly lower near the center of the abdomen. This decay in SI results in an inhomogenous appearance to the image, and the varying signal intensity can make the image difficult to interpret. Most magnets have an option to make the image appear more homogenous by applying a correction (suface coil intensity correction) that will decrease signal intensity near the array and increase it farther away from the array. It should be noted that such corrections cannot eliminate the falloff in SNR with distance from the array, so that the correction may result in increased conspicuity of noise at large distances from the array.
 
Fat Sat: Fat saturation is generally of little use in fetal imaging because the fetus has so little fat. While suppressing the abdominal fat of the mother may be of some use for reducing the intensity of aliasing artifacts in small FOV imaging, this must be reconciled with the increased acquisition time that is usually required for fat suppression.