Variable-density (VD) spiral k-space acquisitions are used to acquire high-resolution (0.78 mm), motion-compensated images of the coronary arteries. Unlike conventional methods, information for motion compensation is obtained directly from the coronary anatomy itself. Specifically, periods of minimal coronary distortion are identified by applying the correlation coefficient template matching algorithm to real-time images generated from the inner, high-density portions of the VD spirals. Combining the data associated with these images together, high-resolution, motion-compensated coronary images are generated. Because coronary motion is visualized directly, the need for cardiac-triggering, breath-holding, and navigator echoes is eliminated. The motion compensation capability of the technique is determined by the inner-spiral spatial and temporal resolution. Results indicate that the best performance is achieved using inner-spiral images with high spatial resolution (1.6 -2.9 mm), even though temporal resolution (four to six independent frames per second) suffers as a result. Image quality within the template region in healthy volunteers was found to be comparable to that achieved with cardiac-triggered breath-hold scans, although Key words: variable-density; template matching; coronary artery; motion compensation; real-time Diagnostic-quality MR coronary images must possess submillimeter spatial resolution. While the theoretical limits of MR do not preclude the attainment of such resolutions, respiratory-(1) and cardiac-(2) induced displacement and distortion (3) of the arteries can significantly degrade image quality. To counteract these effects, motion compensation schemes have been developed (4,5). With the evolution toward higher-resolution imaging, however, a number of concerns are arising with respect to their accuracy. Most of these approaches use indirect measures such as the position of bellows placed over the chest, ECG waveforms, and diaphragm position determined by navigator echoes to infer coronary motion. Recent studies have indicated that while indirect measures may correlate with coronary displacement, they do not give a precise characterization of the actual motion (1,6,7). Additionally, indirect measures generally do not give any indication of the degree of distortion associated with the coronary motion (3). Furthermore, arrhythmias and/or difficulties in breathholding, commonly found in patients with coronary disease, give rise to added difficulties in the application of these techniques (8).A technique that makes use of direct visualization of the coronary anatomy for motion compensation was developed recently by Hardy et al. (9). In this "adaptive averaging" technique, a series of interleaved high-resolution echo-planar images are acquired. From each of the individual interleaves, aliased "subimages" are formed. Since these subimages are generated from every interleaf, they provide real-time visualization of the coronary anatomy. Each subimage is used to evaluate the motion present during the...