Development and application of a visual perception model for the objective assessment of fast magnetic resonance imaging

With the development of interventional and high-speed applications, there has been an increase in the need for, and consequently the development of, high-frame rate, fast magnetic resonance imaging (MRI) techniques. Often, the increase in frame rate is achieved at the expense of image quality. For novel fast MR techniques, image quality is often measured by the subjective judgment of the sequence designer or some simple measure of image gray level, such as signal to noise ratio. A definitive method for evaluating the quality of these methods, which include keyhole, radial, and spiral imaging, among others, has not been previously accepted. The goal of this research is to use new techniques from human vision modeling to develop a quantitative tool for the objective assessment of fast MR imaging.; We have developed and validated a human vision model called the perceptual difference model (PDM). PDM results correlated well with human observer ratings of image quality in the presence of degradation due to noise, blur, contrast changes, and image artifacts. Additionally, we have shown that PDM results can be used to predict interventional radiologists' ratings of the acceptability of fast MR images for certain clinical applications.; We quantified the image quality degradation for novel fast MRI acquisition and reconstruction strategies. In keyhole imaging, we examined the tradeoffs of keyhole stripe size and orientation with objective PDM scores of the resulting image quality. In spiral MR, we have quantified the effects of 672 different acquisition and reconstruction combinations. Results show that there are preferred combinations of acquisition and reconstruction techniques for superior image quality, and there are advantages to using conventional regridding with Voronoi density compensation.; We also used the PDM to evaluate fat suppression imaging. We evaluated the keyhole Dixon technique for fat-suppressed MR images, and showed that images perceptually equivalent to current methods can be created while saving 20–30% of the acquisition time.; These results have demonstrated a new methodology for evaluating fast MR imaging that will aid the development of improved MR sequences. We believe that the PDM will be applicable to many other medical imaging, as well as industrial, applications.