Does the magnetization transfer effect bias chemical exchange saturation transfer effects? Quantifying chemical exchange saturation transfer in the presence of magnetization transfer.
Smith AK., Ray KJ., Larkin JR., Craig M., Smith SA., Chappell MA.
PURPOSE: Chemical exchange saturation transfer (CEST) is an MRI technique sensitive to the presence of low-concentration solute protons exchanging with water. However, magnetization transfer (MT) effects also arise when large semisolid molecules interact with water, which biases CEST parameter estimates if quantitative models do not account for macromolecular effects. This study establishes under what conditions this bias is significant and demonstrates how using an appropriate model provides more accurate quantitative CEST measurements. METHODS: CEST and MT data were acquired in phantoms containing bovine serum albumin and agarose. Several quantitative CEST and MT models were used with the phantom data to demonstrate how underfitting can influence estimates of the CEST effect. CEST and MT data were acquired in healthy volunteers, and a two-pool model was fit in vivo and in vitro, whereas removing increasing amounts of CEST data to show biases in the CEST analysis also corrupts MT parameter estimates. RESULTS: When all significant CEST/MT effects were included, the derived parameter estimates for each CEST/MT pool significantly correlated (P < .05) with bovine serum albumin/agarose concentration; minimal or negative correlations were found with underfitted data. Additionally, a bootstrap analysis demonstrated that significant biases occur in MT parameter estimates (P < .001) when unmodeled CEST data are included in the analysis. CONCLUSIONS: These results indicate that current practices of simultaneously fitting both CEST and MT effects in model-based analyses can lead to significant bias in all parameter estimates unless a sufficiently detailed model is utilized. Therefore, care must be taken when quantifying CEST and MT effects in vivo by properly modeling data to minimize these biases.