Accelerated 3D Multispectral MRI with Robust Principal Component Analysis for Separation of On- and Off-Resonance Signals

We have developed a method to acquire diagnostic MR images in 2-3 minutes to help millions of people who have complications related to metal devices, commonly hip replacements, knee replacements or spinal fixations.

Our group has led the development of 3D methods for MRI in patients with implanted metal devices, which avoid the severe distortions and image artifacts that metal causes in conventional MRI.  These methods are routinely used in clinical practice throughout the world, but require scan times of 6-10 minutes and compromise image resolution.  This research uses advanced acquisition and image reconstruction methods to reduce these scan times to 2-3 minutes.  Alternatively, the methods can be used to improve the spatial resolution of images. 

The novel approach uses a constrained reconstruction that decomposes the image into the content that is disrupted by the metal, and the content that could be imaged with normal MRI.  This is combined with an advanced data sampling scheme using compressed sensing methods, but tailored specifically to the task of imaging near metal.  Graphically this is shown below – most excited slices are flat (as expected).  However, near metal the shape of the slice is distorted because the metal causes variations in the magnetic fields used. By constraining the signal in these categories to either be “on-resonance” or of limited extent, the reconstruction accurately reproduces an image with much less of the original data, enabling faster image acquisition.

Levine E, Stevens K, Beaulieu C, Hargreaves B. Accelerated three-dimensional multispectral MRI with robust principal component analysis for separation of on- and off-resonance signals. Magn Reson Med. 2018 Mar;79(3):1495-505.

Online Journal Article

Principal component analysis of the matrix constructed from bin profiles corresponding to each column in the image allows for separation of on- and off-resonance signals.

Associate Professor of Radiology (Musculoskeletal Imaging) and, by courtesy, of Orthopaedic Surgery
Professor of Radiology (Musculoskeletal Imaging) and, by courtesy, of Orthopaedic Surgery
Professor of Radiology (Radiological Sciences Laboratory) and, by courtesy, of Electrical Engineering and of Bioengineering

Evan Levine is an alumnus of the BMR group