Oral Abstract Session
SCMR 22nd Annual Scientific Sessions
First-pass perfusion CMR is a common technique used to assess myocardial ischemia. Source frames are acquired over multiple heartbeats during normal respiration and in-plane breathing motion is retrospectively corrected (1) to support both visual interpretation as well as quantitative measurement of tissue perfusion. However, retrospective correction cannot account for through-plane motion, and can fail if breathing motion is extreme. Respiratory gating is incompatible with this technique, so navigator based prospective slice translation in the head-foot direction has been developed to partially account for the respiratory motion; however, this is generally an oversimplification of the motion of the heart. In this work we describe a patient specific prospective motion correction method (PROCO) based on a single diaphragmatic navigator to allow for continuous imaging while correcting for respiratory motion. This is a step towards improving visual interpretation, enabling quantitative perfusion, and highly accelerated compressed sensing reconstruction.
A short training scan is performed to characterize patient-specific motion, and a modified pulse sequence reads model parameters and updates the scan plane in real-time based on the current navigator position, as shown in Figure 1. Following model definition, PROCO was applied in a modified perfusion imaging protocol in a single healthy volunteer (image matrix: 120x160, FOV 390 mm, slice thickness: 8 mm, temporal window: 208.1 ms, bandwidth: 1490 Hz/Px, TR/TE: 1.4/1.1 ms) on a clinical scanner MAGNETOM Avanto 1.5T (Siemens Healthcare, Erlangen, Germany). A volume selective saturation recovery pulse was placed over the heart to avoid saturating the navigator pulse. Gadavist was administered over 50 consecutive frames.All data were retrospectively downsampled at R = 4 using variable density pseudo random sampling and were reconstructed using a parameter-free compressive sensing method (2). Because the scan plane was updated in real-time, we used temporally varying sensitivity maps in the reconstruction, which were estimated using ESPIRiT (3). Frames from PROCO datasets were then retrospectively registered, and time-intensity curves were evaluated in a septal region in the short axis and 4 chamber views, and an anterior segment for the 2ch view for both retrospectively registered and non-registered images (4).
Resultant perfusion curves are shown in Figure 2. Curves with and without retrospective non-rigid image registration are comparable, implying that prospective slice correction is accurately predicting the motion of the heart and allowing for more accurate analysis of perfusion. Average percent error was limited to 1.81±3.92. Sample frames are shown in Figure 3.
PROCO significantly reduces through and in-plane residual motion of the heart; this is expected to allow for more accurate qualitative and quantitative analyses of perfusion compared to retrospective in-plane registration.