Perfusion MRI With Radial Acquisition for Arterial Input Function Assessment Eugene G. Kholmovski * and Edward V.R. DiBella Quantification of myocardial perfusion critically depends on accurate arterial input function (AIF) and tissue enhancement curves (TECs). Except at low doses, the AIF is inaccurate be- cause of the long saturation recovery time (SRT) of the pulse sequence. The choice of dose and SRT involves a trade-off between the accuracy of the AIF and the signal-to-noise ratio (SNR) of the TEC. Recent methods to resolve this trade-off are based on the acquisition of two data sets: one to accurately estimate the AIF, and one to find the high-SNR TEC. With radial k-space sampling, a set of images with varied SRTs can be reconstructed from the same data set, allowing an accurate assessment of the AIF and TECs, and their conversion to con- trast agent (CA) concentration. This study demonstrates the feasibility of using a radial acquisition for quantitative myocar- dial perfusion imaging. Magn Reson Med 57:821– 827, 2007. © 2007 Wiley-Liss, Inc. Key words: dynamic MRI; perfusion; radial sampling; arterial input function; myocardium The accuracy of quantitative analyses of tissue perfusion derived from dynamic T 1 -weighted contrast-enhanced (CE)-MRI critically depends on knowledge of the arterial input function (AIF) and tissue enhancement curve (TEC) (1–5). The true AIF is equivalent to the function that de- scribes changes in contrast agent (CA) concentration in the arterial blood with time. The CA concentration cannot be directly measured in perfusion MRI studies. As a result, the AIF is typically derived from the mean signal intensity of the blood in a region of interest (ROI). A linear relation- ship between the change in blood signal intensity due to the CA and the CA concentration is assumed. However, in many practical cases such an assumption is not valid because of the high concentration of CA and/or the long saturation recovery time (SRT) of the applied pulse se- quence (6,7). Low CA doses and/or very short SRTs for which the linear relationship is valid are not optimal for perfusion studies due to relatively low enhancement of the tissue of interest. This makes it more difficult to reliably discrimi- nate tissue regions with normal and reduced perfusion. To improve the accuracy of the AIF estimate for the CA con- centration range relevant to perfusion studies, two meth- ods have recently been proposed (8 –11). The first method (8,9) uses two scans (with separate injections of very low and high CA concentrations) separated by a short time interval. The very-low-dose “prebolus” scan is used only to estimate the AIF of the high-dose scan. In the second technique (10,11), the AIF is found from an additional low-resolution slice acquired with a much shorter SRT than the SRT used for the other, higher-resolution slices used for TEC evaluation. Both techniques potentially al- low more reliable AIF assessment than the conventional approach in which both AIF and TECs are derived from the same data set. However, the prebolus techniques re- quire increased imaging time and complexity, and the AIF of the high-dose scan can be accurately derived from the AIF of the low-dose scan only if the patient’s physiological parameters are the same during both scans. Furthermore, such an approach is logistically demanding for pharmaco- logical stress studies when vasodilator agents with a short action time, such as adenosine, are used. In the second method, the limited time interval available for the acqui- sition of diagnostic slices is further reduced to accommo- date the low-resolution scan that is exclusively used to estimate the AIF. This method also may suffer from T * 2 effects, especially at 3T (10,12,13). In addition, both of these methods use different data sets to estimate the AIF and TECs. For a quantitative analysis it is preferable to obtain the AIF estimate and TECs from the same data set to guarantee complete time and spatial correspondence be- tween them. In this paper we present a method to obtain accurate AIF estimates and high-SNR TECs without the complexities and time penalties of the methods discussed above. The AIF and TECs can be obtained from the same data set if the data are acquired using a T 1 -weighted pulse sequence with radial sampling of k-space. THEORY For a turbo-FLASH pulse sequence with saturation recov- ery magnetization preparation, the signal of the n-th read- out can be presented as: M xy  n = M 1 - E D  a n-1 + M 1 - E 1  1 - a n-1 1 - a , [1] M = M o sine -TE T * 2 , E D = e -TD T1 , E 1 = e -TR T1 , a = cosE 1 , where TD is the time interval between the center of the saturation (90°) RF pulse and the first excitation (°) pulse. Ideal (complete) saturation by the RF pulse is assumed in the derivation of the Eq. [1]. Assuming fast exchange be- tween the protons inside erythrocytes and plasma protons, one can readily modify the equation to include the depen- Utah Center for Advanced Imaging Research, Department of Radiology, University of Utah, Salt Lake City, Utah, USA. Presented in part at the 8th Annual Meeting of SCMR, San Francisco, CA, USA and the 13th Annual Meeting of ISMRM, Miami Beach, FL, USA. *Correspondence to: Eugene G. Kholmovski, UCAIR, Department of Radiol- ogy, University of Utah, 729 Arapeen Drive, Salt Lake City, UT 84108. E-mail: ekhoumov@ucair.med.utah.edu Received 29 January 2006; revised 9 December 2006; accepted 12 January 2007. DOI 10.1002/mrm.21210 Published online in Wiley InterScience (www.interscience.wiley.com). Magnetic Resonance in Medicine 57:821– 827 (2007) © 2007 Wiley-Liss, Inc. 821