Improved accuracy and precision in 3D T2 mapping of knee cartilage with dictionary fitting and patch-based denoising

Abstract

We implemented and compared three different reconstructions for 3D T2 mapping of the knee: I) a standard image reconstruction followed by an analytical fit, II) a standard image reconstruction followed by a dictionary fit, and III) a denoised image reconstruction followed by a dictionary fit. We optimized and compared these techniques in phantoms, five healthy volunteers, and five patients with mild osteoarthritis. The third reconstruction resulted in the highest accuracy and precision while retaining the spatial resolution, and allowed the load-bearing cartilage in the mild-OA patients to be differentiated from that in the healthy volunteers.