AUTOMATIC CORRECTION OF MOTION ARTIFACTS IN MR IMAGES

In many applications of MRI, motion is the most important factor limiting image quality. Recent advances in fast scanning, navigator echoes, and other techniques have addressed the issue of motion in many applications, but it remains a significant problem whenever there is a need for maximum image sharpness, imaging of moving structures, high spatial resolution, or if patients are predisposed to motion. The objective of this proposal is the development and validation of "autocorrection" methods for the retrospective correction of motion artifacts in MRI. These novel methods, based on techniques used in synthetic aperture radar and seismic data processing, can deduce and correct for patient motion during image acquisition from the raw data alone, with no need for special pulse sequences, motion tracking techniques, patient preparation, or specialized hardware. They have been shown to reduce motion artifacts nearly as well as navigator echoes in certain types of images with no a priori knowledge of the patient motion. Their basic principle is the iterative optimization of an image quality metric by searching over trial motion corrections. The research plan has three specific aims: (1) further development of autocorrection methodology, (2) investigation of image quality metrics, and (3) implementation and validation of autocorrection in five specific clinical applications: rotator cuff imaging, spine imaging, cerebral and renal angiography, 3D head imaging, and diffusion imaging. These applications are clinically important, pose widely different technical requirements, and are representative of a wide range of future applications of autocorrection techniques. The overall significance of this work is that autocorrection has potential for wide applicability to the reduction of motion artifacts in many different kinds of MR imaging. All classes of MR image acquisition techniques, including many applications of fast scan methods, may be degraded by artifacts, contrast loss, and blurring due to motion during acquisition. The techniques proposed here are simple, flexible, and require only the raw data from the MR scanner. They are applicable to both conventional and fast scan applications, and may be adaptable to many non- motion sources of artifacts as well.