Purpose: To investigate the influence of the contrast-to-noise ratio (CNR) and temporal resolution of 3D dynamic contrast-enhanced magnetic resonance imaging on the quantification of pulmonary perfusion parameters by means of Monte-Carlo simulations and a volunteer study.
Methods: Quantification of perfusion parameters such as pulmonary blood flow (PBF) and pulmonary blood volume (PBV) was simulated using synthetic data with varying CNR (noise standard deviations ranging from 0% to 25% of the parenchymal signal maximum) and different temporal resolutions from 1 to 5 seconds. Simulation results were compared with perfusion measurements in 9 healthy volunteers (age: 18-31) using dynamic 3D gradient-echo sequences with different k-space acquisition schemes, resulting in different temporal resolutions of 1.1, 1.3, and 2.0 seconds per volume and varying CNR. Lung parenchyma was segmented using a semiautomatic technique, and PBF, PBV, as well as MTT were determined pixelwise using Tikhonov-regularized deconvolution with an optimized L-curve criterion.
Results: The simulations showed a breakdown of the deconvolution algorithm for temporal resolutions lower than 3 s. PBF was increasingly underestimated with decreasing CNR; the temporal resolution had little influence on PBF estimates in the investigated range. PBV, on the other hand, was more dependent on the temporal resolution and showed less, but noticeable influence of the CNR. For low temporal resolutions, PBV was overestimated.In the volunteer study, no significant differences of PBF between the different sequences were found; however, PBV was underestimated by the slowest sequence, in contrast to the simulation results.
Conclusion: Both the mathematical simulations and volunteer measurements showed that the temporal resolution has less influence than the CNR on the quantification of PBF. On the other hand, PBV is more influenced by temporal resolution than by CNR. This indicates that the gain in measurement speed that can be obtained by modern acceleration schemes need not be invested in ultra-high temporal resolution. Instead, sequence optimization should aim for a suitable balance of sufficient CNR on the one hand, and sufficient temporal resolution on the other hand, whereas maintaining a high spatial resolution.