Evaluation of Partial Volume Correction Techniques for Sodium MRI of the Achilles Tendon

To evaluate partial volume correction (PVC) techniques for sodium MRI of the Achilles tendon in situ and in vivo.

Five PVC methods were evaluated including a volume ratio of the proton and sodium segmentations (PSSR), a modified least trimmed square (3D-mLTS) linear regression, a geometric transfer matrix (GTM) approach, a single target correction (STC), and a novel estimated single target correction (eSTC). Their performance was tested using simulated data and 3 T MR data of two volunteers' Achilles tendons acquired at different resolutions: 1.5, 2.0, 3.0, and 4.5 mm³. Since there was no in vivo ground truth, the highest-resolution apparent tissue sodium contents (aTSC) were used.

In the simulation, all PVC methods reduced the difference between the actual and calculated concentrations and were 11.69 ± 6.17 mM without PVC, 4.90 ± 5.40 mM with the PSSR, 4.86 ± 5.19 mM with the mLTS, 1.72 ± 4.13 mM with the GTM, 0.36 ± 1.77 mM with STC and 0.26 ± 1.63 mM with the eSTC. In vivo, the difference in aTSCs between the lower and the highest resolution decreased with all PVCs ranging from 3.6 to 38.8 mM without PVC, 2.8-20.4 mM with PSSR, 4.5-25.9 mM with mLTS, 0.9-7.8 mM with GTM, 0.1-23.8 mM with STC, and 0.7-7.7 mM with eSTC.

PVC generally improved the accuracy of aTSC calculations. The newly introduced eSTC produced the most accurate results for the Achilles tendon.