Myocardial blood flow estimates from dynamic contrast-enhanced magnetic resonance imaging: three quantitative methods

Dynamic contrast-enhanced cardiovascular magnetic resonance imaging can be used to quantitatively assess the myocardial blood ow (MBF), recovering the tissue impulse response function for the transit of a gadolinium bolus through the myocardium. Several deconvolution techniques are available, using various models for the impulse response. The method of choice may in uence the results, producing differences that have not been deeply investigated yet. Three methods for quantifying myocardial perfusion have been compared: Fermi function modelling (FFM), the Tofts model (TM) and the gamma function model (GF), with the latter traditionally used in brain perfusion MRI. Thirty human subjects were studied at rest as well as under cold pressor test stress (submerging hands in ice-cold water), and a single bolus of gadolinium weighing 0.1 ± 0.05 mmol kg−1 was injected. Perfusion estimate differences between the methods were analysed by paired comparisons with Student’s t-test, linear regression analysis, and Bland–Altman plots, as well as also using the two-way ANOVA, considering the MBF values of all patients grouped according to two categories: calculation method and rest/stress conditions. Perfusion estimates obtained by various methods in both rest and stress conditions were not signi cantly different, and were in good agreement with the literature. The results obtained during the rst-pass transit time (20 s) yielded p-values in the range 0.20–0.28 for Student’s t-test, linear regression analysis slopes between 0.98–1.03, and R values between 0.92–1.01. From the Bland– Altman plots, the paired comparisons yielded a bias (and a 95% CI)—expressed as ml/min/g—for FFM versus TM, −0.01 (−0.20, 0.17) or 0.02 (−0.49, 0.52) at rest or under stress respectively, for FFM versus GF, −0.05 (−0.29, 0.20) or −0.07 (−0.55, 0.41) at rest or under stress, and for TM versus GF, −0.03 (−0.30, 0.24) or −0.09 (−0.43, 0.26) at rest or under stress. With the two-way ANOVA, the results were p = 0.20 for the method effect (not signi cant), p < 0.0001 for the rest/stress condition effect (highly signi cant, as expected), whereas no interaction resulted between the rest/ stress condition and method (p = 0.70, not signi cant). Considering a wider time-frame (60 s), the estimates for both rest and stress conditions were 25%–30% higher (p in the range 0.016–0.025) than those obtained in the 20 s time-frame. MBF estimates obtained by various methods under rest/stress conditions were not signi cantly different in the rst-pass transit time, encouraging quantitative perfusion estimates in DCE-CMRI with the used methods.