Phase-contrast (PC) MRI is a powerful tool for quantitative assessment of blood flow, and it has been widely used in clinical practice ^{[1, 2]}. However, the conventional approach relies on electro-cardiogram (ECG)-synchronized cine acquisition, which is not well suited to assess hemodynamic variabilities. Additionally, its practical utility can often be constrained by cardiac arrhythmias. Real-time evaluation of blood flow has shown great potential to overcome these issues ^{[3, 4]}. Recently, we have developed a novel method ^[5] for real-time PC-MRI based on low-rank modeling ^[6] (without ECG gating and respiration control). Here, we evaluate its accuracy with flow phantom experiments and illustrate its key advantages over the conventional cine-based approach. Moreover, we present a preliminary assessment of in vivo real-time flow imaging of thoracic aorta.All the experiments were performed on a 3.0T whole body system (Achieva, Philips Medical System, Best, The Netherlands) with a 32-channel cardiovascular coil. We first performed three sets of flow phantom experiments with different flow profiles to evaluate the performance of the cine and real-time acquisitions. A 15 mm-diameter stiff tube was used to simulate the aorta. The flow rates were set by a computer-programed pump (CompuFlow 100 MR, Toronto, Canada). Here we designed three classes of flow profiles to simulate different blood flow variations. The first one was generated by a repetitive pump protocol with a Gaussian profile (with duration of 0.5 s), followed by a constant flow for a total cycle time of 1 s (shown in Fig. 2a). Note that this mimics the flow profile of healthy subjects. The second one consists of two Gaussian profiles with different maximum flows, followed by a constant flow with the same duration to imitate flow variability. This results in a total repetition time of 2 s (Fig. 2b). The third one consists of four Gaussian profiles, each of which is followed by a constant flow with different durations (Fig. 2c). This simulates the flow profile of patients with cardiac arrhythmia. Here the first and second flow profiles were used for both the PC-MRI cine and real-time acquisitions, while the third one was used for real-time imaging because the period irregularity is not well suited for cine acquisition. The relevant imaging parameters are listed in Table 1. Lastly, in vivo data of thoracic aorta were acquired from a healthy volunteer with no symptoms of cardiovascular disease and informed written consent obtained. Both the PC-MRI cine and real-time acquisitions were performed in the transverse orientation, covering a FOV of 240×220 mm² at an in-plane resolution of 2.2 mm and with a slice thickness of 5 mm.For the flow phantom experiments, the reconstructed magnitude and phase images at a peak flow time instant are shown in Fig. 1a and 1b. For the first set of experiments, the consistency between the flow profiles from PC-MRI cine (Fig. 2d) and real-time (Fig. 2f) acquisitions is observed. Further, notice that both derived peak flows are close to the pre-designed one (Fig. 2a). For the second group, the reconstructed real-time flow profile (shown in Fig. 2g) well captures the high peak flow and low peak flow. In contrast, the flow profile from cine acquisition (in Fig. 2e) is only able to capture an averaged peak flow. With respect to the third group, the reconstructed real-time flow profile (Fig. 2h) reveals the similar flow profile variations in both period duration and amplitude to the pre-designed one. In terms of the in vivo experiments, the reconstructed magnitude and phase images at a systolic time frame are shown in Figs. 3a and 3b. The mean flow in the ascending aorta for 13 consecutive heartbeats using real-time PC-MRI is shown in Fig. 3c. By averaging these 13 consecutive heartbeats, the derived mean flow profile closely matches the one derived from cine acquisition (Fig. 3d).In this work, we demonstrated via phantom experiments that real-time PC-MRI with low-rank modeling is capable to well capture the hemodynamic variabilities and irregular flow variations. The excellent correlation between PC-MRI cine and real-time method has been observed in vivo experiments in healthy subject as well. For future study, we are going to evaluate the feasibility of applying the proposed real-time PC-MRI technique for patients with arrhythmia.