PURPOSE To compare an (independent) high-temporal-resolution arterial input function (AIF) to an AIF obtained from low-temporal-resolution data in the context of pharmacokinetic parameter estimation. The two methods are compared across a range of temporal resolutions to obtain insight into estimation errors on Ktrans and ve for data acquired at suboptimal temporal resolution. METHOD AND MATERIALS DCE-MRI data of an animal tumor model (Copenhagen rats, N=6, AT6.1) was acquired at high temporal resolution (T = 5 s). Via a k-space-based downsampling approach, low temporal resolution equivalents were obtained for T = 15/30/45/60 s. For all the data sets, AIFs were derived from the contrast uptake in muscle using literature values for Ktrans and ve. A basic two-compartment model was fit to each individual pixel within the tumor rim, and to the mean uptake across the tumor rim. For method 1, the 5-s AIF was used for all derived data sets; for method 2, the AIFs corresponding to the low temporal resolution data sets were used. RESULTS Method 1 (5-s AIF) demonstrated progressive underestimation of Ktrans and progressive overestimation of ve with decreasing temporal resolution; e.g. at T = 60 s, the average estimation errors were, -17.9% and +6.4%, respectively. The progressive trends were statistically significant for all six cases (multiple sign tests used, all p-values <0.00001). Method 2 (data-equivalent AIF) resulted in smaller, but less consistent errors. Again for T = 60 s, the absolute errors on the estimation of Ktrans and ve were 2.2% and 1.3%, respectively. This time, the effect of temporal resolution on the parameter estimation was not always statistically significant. CONCLUSION Even though commonly advised against, AIF-extraction from low temporal resolution data (method 2) resulted in unexpectedly small errors. This is likely the result of deriving the AIFs from the uptake in muscle, because this uptake can be sampled more slowly than the bolus passage in an artery. We conclude that the use of an AIF extracted from low temporal resolution data is feasible, but the stability and timing aspects require further investigation. CLINICAL RELEVANCE/APPLICATION Despite the sometimes low temporal resolution (~1 min) of clinical DCE-MRI, we could estimate pharmacokinetic parameters using a low temporal resolution AIF within a reasonable error margin.
|Title of host publication||Proc. RSNA 2009|
|Place of Publication||United States, Chicago|
|Publication status||Published - 2009|