Contrast, resolution, and detectability in MR imaging.

Abstract

With the introduction of fast scan techniques and high field imagers, the ability to achieve very high resolution MR images in reasonable imaging times is now possible. Increased resolution allows for better detection of small, high contrast pathological features, but at some cost. Increasing resolution leads to a nonrecoverable decrease in signal-to-noise ratio per pixel and a loss of low contrast detectability for constant imaging time. This article examines the tradeoffs between image resolution, signal-to-noise ratio, and low contrast detectability in MR imaging. Contrast detail curves are presented for images collected in a constant imaging time, with constant field of view and bandwidth but at different resolutions, and these are compared with theoretical curves. The problem of measuring contrast levels in magnitude images, with different resolutions and receiver attenuation values, is discussed and a definition that accommodates these parameters developed. In addition, a clinical example is shown demonstrating a decrease in soft tissue differentiation with increasing resolution, again for fixed imaging time. The results indicate that moving to high resolution imaging matrices requires consideration be given to the sacrifice in low contrast detectability that occurs. Most importantly, it is shown that filtering a high resolution image to a lower resolution image, through nearest neighbor averaging, does not regain the detectability lost in initially collecting the high resolution image.