Information content of SNR/resolution trade-offs in three-dimensional magnetic resonance imaging

In MRI, a trade-off exists between resolution and signal-to-noise ratio, since different fractions of the available scan time can be used to acquire data at higher spatial frequencies and to perform signal averaging. By comparing a wide variety of 3D isotropic MR scans with different combinations of SNR, resolution, and scan duration, the impact of this trade-off on the image information content was assessed. The information content of mouse brain, mouse whole-body, and human brain images was evaluated using a simple numerical approach, which sums the information contribution of each individual k-space data point. Results show that, with a fixed receiver bandwidth and field of view, the information content of trade-off images is always maximized when the SNR is equal to about 16. The optimal imaging resolution is dependent on the scan duration, as well as certain MR system properties, such as field strength and coil sensitivity. These properties are, however, easily accounted for with the acquisition of a single scout MR image, and the optimal imaging resolution can then be calculated using a simple mathematical relationship. If the imaging task is approached with a predetermined resolution requirement, the same scout scan can be used to calculate the scan duration that will provide the maximum possible information. Using these relationships to maximize the image information content is an excellent technique for guiding the initial selection of imaging parameters.