Motion in MRI is a significant issue, leading to long scan times and loss of diagnostic quality. Motion in cardiovascular MRI is especially complex, as it contains both respiratory and cardiac motion in addition to general bulk motion. Cardiac motion is effectively gated with an electrocardiogram (ECG) which can bin data into specific cardiac phases, but respiratory motion presents a different challenge due to prohibitively long scan times when data acquisition is limited to end-expiration. Retrospective techniques have been developed to attempt to register frames acquired at varied respiratory phases, but these techniques are limited as they cannot correct for through-plane or exaggerated in-plane motion. Prospective slice tracking has previously been developed to attempt to follow the heart throughout the respiratory cycle, allowing for highly efficient free-breathing imaging. However, these techniques are generally applied with a generic tracking factor to correlate a respiratory signal to the position of the heart, and do not adequately represent the patient and respiratory phase-specific motion of the heart.
We have developed a patient and respiratory phase-specific three-dimensional prospective motion correction technique (PROCO) that can model and correct for respiratory motion of the heart in real-time. For each study, a short training scan consisting of a series of single heartbeat images, each acquired with a preceding diaphragmatic navigator, was performed to fit a model relating the patient-specific three-dimensional respiratory motion of the heart to diaphragm position. The resulting motion model was then used to update the imaging plane in real-time to correct for translational motion based on respiratory position provided by the navigator.
This model was initially validated by comparing against uncorrected free-breathing (FB), a generic tracking factor of 0.6 (FB-TF), navigator gating (Nav-Gate) and navigator gating combined with a generic tracking factor of 0.6 (Nav-Gate-TF). Each method was applied in a 2-chamber, 4-chamber and short-axis view in a group of 11 healthy volunteers. PROCO reduced the range/RMSE of residual motion to 4.08±1.4/0.90±0.3mm, compared to 10.78±6.9/2.97±2.2mm for FB, 5.32±2.92/1.24±0.8mm for FB-TF, 4.08±1.6/0.93±0.4mm for Nav Gate, and 2.90±1.0/0.63±0.2mm for Nav Gate-TF. Nav Gate and Nav Gate-TF reduced scan efficiency to 48.84±9.31% and 54.54±10.12%, respectively. In this study we showed that PROCO successfully limited the residual motion in single-shot imaging to the level of traditional navigator gating, while maintaining 100% acquisition efficiency.
After validating the effectiveness of the PROCO method, we applied the technique to T1 and T2 mapping, where each source image is prospectively corrected prior to image registration and parametric fitting. 10 repetitions of mid-ventricular T1 and T2 maps were acquired under breath-hold and free-breathing with and without PROCO in a group of 7 healthy volunteers. Retrospective image registration was applied to all source images prior to T1 and T2 estimation. Measurements of T1 and T2 were made in the 6 AHA mid-ventricular segments and compared using Bland-Altman analysis with breath-hold measurements as the reference standard. Free-breathing acquisitions with PROCO greatly improved the precision of parametric mapping with respect to breath-hold measurements, producing limits of agreement of -45.4 to 19.94 ms (T1) and -0.86 to 2.82 ms (T2), compared to -74.99 to 69.59 ms (T1) and -6.62 to 10.44 ms (T2) under free-breathing without PROCO. The wide variability in measurements without PROCO was likely due to exaggerated in-plane motion and through-plane motion that cannot be corrected retrospectively. In addition, 3 patients were included in this study to assess the benefits of PROCO in a population with pre-existing T1 or T2 abnormalities. Free-breathing PROCO maps retained similar quality to breath-holds and were able to consistently identify areas of pathology.
PROCO was additionally applied in a perfusion imaging sequence in a group of 3 healthy volunteers. Two separate injections of contrast agent were given for a PROCO sequence and a traditional free-breathing (FB) sequence without motion correction; images were acquired in the 2-chamber, 4-chamber and short axis views once per heartbeat with a corresponding navigator. Each acquisition was retrospectively down sampled and reconstructed with compressed sensing at R = 4. Frames were then registered (PROCO-R and FB-R) using non-rigid image registration, and an ROI was placed in the basal-anterior, basal-lateral and mid inferolateral segments for the 2-chamber, 4-chamber and short axis views, respectively, to record perfusion curves for both registered and non-registered source images. These regions are generally plagued by respiratory motion, and therefore present an optimal location to test the efficiency of the PROCO method. Perfusion curve results showed sharp spikes in signal intensity under FB conditions due to respiratory motion moving blood pool signal in and out of the ROI, whereas PROCO conditions largely removed this effect. While registration removed the larger spikes from FB acquisitions, FB-R curves still showed inconsistencies related to errors in registration, as registration cannot correct for large in-plane or through-plane motion. PROCO-R curves maintained consistency and provide improved diagnostic capability by allowing for accurate and consistent representation of cardiac anatomy. Measurements with a stationary ROI were also compared to a dynamic ROI, where the ROI was moved from frame to frame manually to follow the position of the heart; PROCO stationary ROI measurements were significantly more similar to the dynamic ROI when compared to FB measurements. In addition, preliminary results suggest that PROCO could allow for higher rates of acceleration with compressed sensing, as it stabilizes the position of the heart over all acquired frames.
Our patient and respiratory phase-specific prospective motion correction technique has allowed for highly efficient free-breathing cardiovascular MRI, improving accuracy and precision of measurements in parametric mapping and perfusion imaging. This technique was shown to be superior to a generic linear tracking factor, and similar in motion reduction to navigator gating while maintaining 100% acceptance rate. Accuracy and precision in parametric mapping with PROCO were shown to be superior to retrospective correction only, and similar to breath-holding. Finally, PROCO allowed for improved perfusion curve analysis and higher rates of acceleration with compressed sensing reconstruction in cardiac perfusion imaging.