Diffusion imaging is a valuable tool in the identification of neurological diseases, especially in an acute setting. Large high-field systems can be challenging to site in an acute setting. Mid-field systems have various siting advantages but historically can suffer from sub-par imaging performance. In this work we compare diffusion weighted imaging performance on a 0.5 T system with a high-performance gradient system to a typical 1.5 T clinical scanner. We demonstrate the ability to achieve comparable imaging performance both analytically as well as through imaging examples.
Assuming a similar receive coil behavior for the two systems, relative SNR performance will be dictated by the relative signal levels. Using the values in Table 1 the predicted diffusion imaging performance can be evaluated using an SNR efficiency metric defined as $$ SNR_eff = {M0 * exp^{-TE \over T2} \over \sqrt{minScanTime}} $$
This expression relates the signal polarization (M0), echo time of the diffusion prescription (TE), the tissue T2 relaxation time and the minimum scan time (minScanTime) for whole brain coverage.
In addition, diffusion imaging is prone to geometric distortions in the presence of magnetic field inhomogeneity. The scale of this inhomogeneity is proportionally lower at lower field strengths, which results in significantly reduced geometric distortions. This distortion can be evaluated qualitatively on in vivo images.
The signal polarization is 3-times higher at 1.5 T. However the longer inherent T2 at 0.5 T2,3 and shorter echo time achievable with the high-performance gradient, means the relative signal at the echo time is just over 2-times higher at 0.5T. Assuming a minimum slice coverage for routine neuro protocols of 24 slices, the minimum time to acquire all 24 slices is 1.7 sec on the 0.5 T system and 3.2 sec at 1.5T. Note that these times are both greater than 3 times the T1 of white matter at the corresponding field strength, thereby allowing sufficient signal recovery between repeats and the effect of T1 recovery can be ignored. Using these values the ratio SNReff1.5T / SNReff0.5T = 1.06. Thus the SNR efficiency of a standard DWI scan on the 1.5 T system is only 6% different than what is expected on the 0.5 T system described here.
This in vivo behavior is illustrated in Figure 1 which shows mean diffusion-weighted images obtained at 1.5 T and 0.5 T on a subject with a dental implant. Significant geometric distortion in the anterior portion of the brain can be seen at 1.5 T which is not seen at lower field. The signal to noise behavior can also be seen to be qualitatively similar. Some additional image blurring can be seen in the 1.5 T images, likely a result of the effect of greater signal decay over the longer echo train duration compared to the 0.5 T system.