Applying a level set method for resolving physiologic motions in free-breathing and non-gated cardiac MRI

In cardiac MRI, ECG triggering is used or patients are required to hold their breath, to alleviate motion artifacts and deterioration of image quality. However, ECG signal quality is often suboptimal and patients may not be able to adequately hold their breath. Alternative solutions for tracking breathing and cardiac beating can open the way for robust free-breathing and ECG-less cardiac MRI. Herein, we present a novel approach that isolates the effect of breathing, as well as computes both the breathing and cardiac beating waveforms directly from real-time MRI sequences. It turns a challenge into an opportunity to guide the reconstruction of high temporal resolution images. The proposed method is based on a level-set method to segment the left ventricle from a real-time MR sequence collected with free breathing and without ECG triggering. The algorithm extracts an evolving surface area, which captures the heart's systolic contraction and diastolic expansion in real-time. The computed time series of the heart's dynamic area is subjected to wavelet analysis, where the breathing and pulsation components are separated. The method was investigated on 12 real-time cardiac MRI acquisitions. We demonstrate that the left ventricular area, as computed by the level set method, produces breathing and cardiac waveforms similar with those extracted by cardiac MR experts (ground-truth). This proof-of-concept work demonstrates the capabilities of the proposed methodology paving the way for incorporation into real-time or retrospective reconstruction of high resolution cardiac MR.