Title: Computer-vision based real-time adapative motion-correction in MRI
Abstract: Magnetic resonance imaging (MRI) is a non-invasive medical imaging modality, which provides excellent soft tissue contrast. With approximately 35MM procedures performed annually in the US alone, it is amongst the most frequently used diagnostic imaging methods. Sadly, the sequential nature of MRI data acquisition makes MR images susceptible to patient motion. Particularly scans from certain patient groups, such as those from children, elderly, or patients with certain medical conditions (e.g. stroke, Parkinson’s disease), can be corrupted or even rendered non-diagnostic. In such cases, methods to correct or compensate for motion have to be incorporated into the MR imaging protocol to assure adequate image quality. With increased demands for higher resolution structural scanning, time-resolved examinations (e.g. functional MRI), or diffusion tensor imaging (DTI) examination times can also increase considerably. Here, even willing patients might have trouble staying still during the lengthy course of the examination and again motion correction or motion compensation methods are warranted.
Two types of motion correction/compensation strategies can be used in MRI:
- retrospective correction, that is, correction for motion – after the MR data were acquired – during image reconstruction, and
- prospective correction, that is, real-time compensation of motion by adapting the slice orientation and location relative to the pose changes of the patient.
The MR data acquisition can be designed so that information about pose changes can be detected directly from the imaging data itself, which warrants redundant sampling, or by the acquisition of extra ‘navigator images.’ Both strategies have been deployed successfully but are often limited to certain MR pulse sequences. Moreover, by acquiring extra navigator or redundant data, overall scan time, scan efficiency, and perhaps contrast can suffer considerably. Several other approaches to detect 3D motion have been proposed in the past, such as accelerometers or pickup coils. In addition, optical methods (time-of-flight cameras, stereovision) have been introduced for motion detection at high precision and accuracy and high temporal resolution.
In this presentation, examples of both retrospective and prospective methods as well as MR-based and optical tracking methods will be presented. First, a retrospective scheme, which uses a specialized data acquisition to detect & correct for patient head motion will be introduced, and the mathematical formalism underlying the post-processing in the presence of motion will be presented. Second, a prospective motion correction system that uses an external MR-compatible optical tracking device will be demonstrated. Using both methods, superior MR image quality will be demonstrated for structural and diffusion tensor imaging (DTI) MRI scans.
Speaker: Dr. Bammer, is an expert on MR acquisition and reconstruction methods with a special focus on clinical neuroimaging methods. He holds a Ph.D. in Electrical Engineering and Biomedical Engineering from the Technical University of Graz, Austria, as well as a habilitation for Medical Physics and Biophysics from the Medical University of Graz, Austria. Currently, Dr. Bammer is the director of pediatric radiology research at Stanford University and member of the radiological sciences lab at the Lucas MR Center at Stanford. He is member of the MRM editorial board and has published more than 110-peer reviewed papers, more than 500 conference abstracts and proceedings, 10 book chapters, and numerous U.S. patents.