High Precision Measurement of Mean Transit Time for Pharmacological MRI
Introduction Pharmacological MRI based on BOLD contrast is difficult since drug induced signal intensity changes are small. An alternative is bolus track MRI using a paramagnetic contrast agent which, however, has limitations regarding quantitation, one reason is the poor definition of the arterial input curve (AIC). Although AIC in principle can be determined from contrast agent induced signal intensity changes in larger arteries, the conversion into concentration time courses is inprecise. In this study a new approach was employed correcting the AIC before deconvolution. For evaluation cerebral perfusion was assessed before and after application of L-Arginine (L-Arg). Methods Bolus Track MRI (TE 54 ms; TR 0.8 s; 128x128x8) was performed after i.v. administration of 0.2 mmol/kg GdDTPA in four volunteers before and after L-Arg (500 mg/kg, i.v.). In one slice a white matter ROI was selected and AIC was taken from 6 different groups of arterial pixels and averaged. Concentrations were calculated and part of first passage in AIC was replaced with a fitting function that has been varied in an optimization procedure. The goal was to find the best prediction of time courses in tissue by model-dependent convolution in a transport-diffusion-model. ”Improved“ AIC determined in this way were used as most probable AIC at high concentrations for MTT determination. Results For the ”standard“ AIC, the MTT values were 2.7 ± 0.6 s before and 2.3 ± 0.5 s after L-Arg, the ratio of MTT in the same tissue regions after/before L-Arg was 0.8 ± 0.2. The MTT shortening indicated a CBF increase but not significantly. Using model-dependent deconvolution, the ”improved“ MTT values were similar, but less noisy. MTT values were 3.0 ± 0.5 s before and 2.5 ± 0.5 s after L-Arg. The improvement in SNR was particularly evident when assessing the after/before ratios of MTT which were 0.83 ± 0.02. Conclusion The improved definition of the input function significantly reduced the noise in our data as indicated by smaller standard deviations. As illustrated by the measurement of L-Arg-induced CBF increase, this improvement should make the assessment of rather small pharmacologically induced perfusion changes more precise.