For most applications, the Field of View (FOV) should be chosen to completely contain the anatomy of interest, with the anatomy filling as much of the FOV as possible. The matrix size and slice thickness should be chosen so that the size of the voxel created is appropriate for the size of the anatomical details to be resolved.

For 2D sequences, the choices of FOV and matrix size together determine the in-plane spatial resolution, and the slice thickness determines the spatial resolution in the slice direction. For example, a 2D sequence with FOV = 14 cm,* matrix size = 256 x 256 pixels and slice thickness = 3 mm corresponds to a voxel size of 0.55 mm x 0.55 mm x 3 mm (0.55 mm = 140 mm/256). This means that every pixel in the image displays a grey-scale value that corresponds to the signal intensity from a tissue voxel with these dimensions.

The amount of signal emitted by a tissue voxel is directly proportional to the size of the voxel: the larger the voxel, the larger the emitted signal. Choosing a smaller matrix size and increasing slice thickness yields an image with lower spatial resolution, but higher signal from each voxel. For the example given above, reducing the matrix size from 256 x 256 to 128 x 128 for the same FOV and slice thickness will increase the voxel size by a factor of four. (Fig. 1.1)

Figure 1.1 Schematic showing change in voxel size with changing matrix sizes.

This results in a decrease of spatial resolution but causes a four-fold increase in the signal amplitude from each voxel. Conversely, changing the matrix size from 256 x 256 to 512 x 512 decreases the signal from each voxel by a factor of four. Increasing the slice thickness by a factor of two to 6 mm results in a factor of two increase in signal from each voxel, whereas reducing the slice thickness to 1.5 mm (i.e. by one half) reduces the signal from each voxel by one half.

The MR signal that is mapped to a voxel arises only from that voxel, but the noise for that voxel arises from the entire volume of tissue to which the receiving RF coil is sensitive. The signal from a voxel therefore scales with the voxel size, but the noise does not. Thus, increasing the voxel size increases the signal from each voxel, but the background noise level is unaffected.

Figure 1.2 A-D illustrates the dependence of SNR on matrix size and slice thickness with all other user-specified parameters held constant. The SNR in Figure 1.2 B is half that in Figure 1.2 A due to the decrease in slice thickness by a factor of two. However, note that the actual image SNR in Figures 1.2 C and D does not scale directly with the voxel size in the simple way that we have suggested. This is because changing the matrix size while holding the other user-selectable parameters constant also affected a second important parameter for determining image SNR: the total signal sampling time.

Figure 1.2 Images of a GE quality assurance phantom acquired using: (A) matrix size=256 x 256, RBW=32kHz, slice thickness=10mm, 1NEX; (B) matrix size=256 x 256, RBW=32kHz, slice thickness=5mm, 1NEX; (C) matrix size=256 x 128, RBW=32kHz, slice thickness=10mm, 1NEX; (D) matrix size=512 x 512, RBW=32kHz, slice thickness=10mm, 1NEX.

---

*Throughout this text, unless explicitly stated, the FOV is assumed to be a SQUARE FOV, not rectangular.