Influence of the arterial input sampling location on the diagnostic accuracy of cardiovascular magnetic resonance stress myocardial perfusion quantification

Abstract

Background Quantification of myocardial blood flow (MBF) and myocardial perfusion reserve (MPR) by cardiovascular magnetic resonance (CMR) perfusion requires sampling of the arterial input function (AIF). While variation in the AIF sampling location is known to impact quantification by CMR and positron emission tomography (PET) perfusion, there is no evidence to support the use of a specific location based on their diagnostic accuracy in the detection of coronary artery disease (CAD). This study aimed to evaluate the accuracy of stress MBF and MPR for different AIF sampling locations for the detection of abnormal myocardial perfusion with expert visual assessment as the reference. Methods Twenty-five patients with suspected or known CAD underwent vasodilator stress-rest perfusion with a dual-sequence technique at 3T. A low-resolution slice was acquired in 3-chamber view to allow AIF sampling at five different locations: left atrium (LA), basal left ventricle (bLV), mid left ventricle (mLV), apical left ventricle (aLV) and aortic root (AoR). MBF and MPR were estimated at the segmental level using Fermi function-constrained deconvolution. Segments were scored as having normal or abnormal perfusion by visual assessment and the diagnostic accuracy of stress MBF and MPR for each location was evaluated using receiver operating characteristic curve analysis. Results In both normal (300 out of 400, 75 %) and abnormal segments, rest MBF, stress MBF and MPR were significantly different across AIF sampling locations (p < 0.001). Stress MBF for the AoR (normal: 2.42 (2.15–2.84) mL/g/min; abnormal: 1.71 (1.28–1.98) mL/g/min) had the highest diagnostic accuracy (sensitivity 80 %, specificity 85 %, area under the curve 0.90; p < 0.001 versus stress MBF for all other locations including bLV: normal: 2.78 (2.39–3.14) mL/g/min; abnormal: 2.22 (1.83–2.48) mL/g/min; sensitivity 91 %, specificity 63 %, area under the curve 0.81) and performed better than MPR for the LV locations (p < 0.01). MPR for the AoR (normal: 2.43 (1.95–3.14); abnormal: 1.58 (1.34–1.90)) was not superior to MPR for the bLV (normal: 2.59 (2.04–3.20); abnormal: 1.69 (1.36–2.14); p = 0.717). Conclusions The AIF sampling location has a significant impact on MBF and MPR estimates by CMR perfusion, with AoR-based stress MBF comparing favorably to that for the current clinical reference bLV.

Free‐breathing motion‐informed locally low‐rank quantitative 3D myocardial perfusion imaging

“…28,41,57 The arterial input function (AIF) is preferably recorded in the ascending aorta in an interleaved fashion. 58 The present study proposes and validates a 3D motion correction approach for locally low-rank (LLR) image reconstruction of Cartesian pseudo-spiral in-out k-t undersampled single-bolus first-pass perfusion data. The method is referred to as respiratory motion-informed locally low-rank reconstruction (MI-LLR).…”

“…For quantification of perfusion data, a dual‐sequence, single‐bolus approach is desirable 28,41,57 . The arterial input function (AIF) is preferably recorded in the ascending aorta in an interleaved fashion 58 …”

Impact of Temporal Resolution and Methods for Correction on Cardiac Magnetic Resonance Perfusion Quantification

“…Perfusion images first underwent motion correction and were corrected for coil sensitivity using the acquired proton density maps. 35 Then, pixelwise MBF was quantified using an automated pipeline that detects the left ventricle in the basal slice, exports the AIF, and prepares the curves for quantification. 35 MBF was estimated using Fermi function‐constrained deconvolution alone (see Supporting Information).…”

“… 35 Then, pixelwise MBF was quantified using an automated pipeline that detects the left ventricle in the basal slice, exports the AIF, and prepares the curves for quantification. 35 MBF was estimated using Fermi function‐constrained deconvolution alone (see Supporting Information). The mean (global) MBF in all three short‐axis slices was measured at both rest and stress, which also allowed estimation of global MPR as the stress‐to‐rest ratio of MBF.…”

Quantification of balanced SSFP myocardial perfusion imaging at 1.5 T: Impact of the reference image

“…Myocardial blood flow is estimated using Fermiconstrained deconvolution, restricted to the first pass in the blood pool, as described previously 11,30,31 using inhouse software developed in MATLAB. The Gd estimates from the AIF were estimated using a similar approach as described previously for the myocardium but considering a typical native-blood T 1 at 1.5 T of 1600 ms. No spatial smoothing nor temporal smoothing was applied to the data before deconvolution.…”