1.2.5 - Number of Excitations (NEX)

For orthopedic applications, the required spatial resolution is typically determined by the size of the structures being imaged, e.g. a much higher spatial resolution would be required to assess a meniscal tear than to assess the extent of a bone tumor. It is therefore not usually possible to trade off spatial resolution to achieve higher SNR because of the small structural details that must be resolved. The image SNR is instead improved by lengthening the scan time, usually by increasing the NEX after an appropriate RBW for the intended clinical application is chosen.

The NEX is an imaging parameter that allows repeated acquisition and averaging of the k-space data required to form an image. The process of averaging this data increases SNR in the reconstructed image by increasing the total sampling time. Image SNR increases with the square root of the NEX. The loss of SNR that occurs for smaller voxel sizes can to some degree be compensated for by acquiring the image with increased NEX; however, this is at the cost of additional scan time. Increasing NEX from 1 to 4 increases the SNR in an image by a factor of 2 (sqrt(4) = 2), but entails a factor of 4 increase in scan time.

A fractional value less than 1 can also be chosen for NEX. For fractional NEX = 0.5 (this includes SSFSE), a half-Fourier acquisition is used. This means that slightly over one half of the phase encode lines in k-space are actually collected, and a homodyne reconstruction algorithm1 is used to reconstruct an image. (Fig. 1.6) Because k-space is complex-conjugate symmetric for real objects, it is possible to reconstruct an image without completely filling k-space. The danger of using a half-Fourier method, however, is that any errors in the collected data will propagate directly into the image. In a full Fourier method, many such errors will tend to be self-corrected through the collection of redundant information. For fractional NEX larger than 0.5, zero-filling is used to completely fill k-space (as discussed in Section 1.2.8).

Figure 1.6

The homodyne reconstruction for partial filling of k-space uses (a) a step function, or (b) a ramp function to weight the acquired data such that the high k-space data has twice the weighting as the low frequency data.