Efficacy and Utility of Image Quality Metrics in Magnetic Resonance Image Reconstruction

Abstract

Quantitative measurement of MRI image quality is a difficult, but important problem that can have direct and indirect benefits to MRI research. Objective image quality metrics (IQMs), such as root-mean-squared-error (RMSE) and structural similarity (SSIM), are routinely employed in a research context to evaluate MRI image quality. Here, work is presented that investigates the efficacy of 10 common IQMs including RMSE and SSIM at measuring MRI image quality as perceived by radiologists. While some metrics demonstrated high correlation with the radiologists’ scores, RMSE and SSIM were not among the top performing metrics. SSIM did not perform statistically superior to RMSE. In a second study, an IQM-based heuristic is developed to determine an optimal temporal footprint in dynamic-contrast-enhanced (DCE) MRI for 𝐾𝑡𝑟𝑎𝑛𝑠 measurements. In simulations, the heuristic was able to consistently recover 𝐾𝑡𝑟𝑎𝑛𝑠 within 10%. This technique may allow for selection of temporal footprint in DCE MRI at an individual level.