This project was the first of three aims completed during my PhD.
Background: Breast MRI demonstrates superior sensitivity for breast cancer detection; however, high costs limit its use as a screening tool. Abbreviated MRI was introduced as a screening alternative (see Kuhl et al.) with similar sensitivity levels, but the absence of delayed post-contrast phases may impede quantitative characteristics of tumor perfusion known to have important diagnostic value.
Objective: In this work, we investigated errors induced in perfusion parameters (volume transfer constant, K^trans, and extravascular/extracellular volume fraction, v_e) by fitting the Standard Kety-Tofts perfusion model and its Patlak variant to retrospectively abbreviated research DCE-MRI series with 15 s temporal resolution collected from breast cancer patients from two data sources.
Methods: Using MATLAB, we built a pipeline that performed a voxel-wise fit of the Kety-Tofts perfusion model and the Patlak model variation within segmented tumors to retrospectively abbreviated time courses (ATCs) derived from full time course (FTC) DCE-MRI series. All images were preprocessed with B1 correction. We used the concordance correlation coefficient (CCC) to evaluate the agreement between perfusion model parameters from the FTC and ATC model fits, where CCC=1 indicates perfect agreement.
Results: We found that it was feasible to abbreviate the time courses by as much as 50% while obtaining a CCC of 0.90 or above. An even greater abbreviation of 85% was feasible while yielding a CCC of 0.80 or greater for most patient datasets. Thus, perfusion modelling may be feasible in abbreviated MRI without much loss in perfusion model estimation accuracy.
Objective: In this work, we investigated errors induced in perfusion parameters (volume transfer constant, K^trans, and extravascular/extracellular volume fraction, v_e) by fitting the Standard Kety-Tofts perfusion model and its Patlak variant to retrospectively abbreviated research DCE-MRI series with 15 s temporal resolution collected from breast cancer patients from two data sources.
Methods: Using MATLAB, we built a pipeline that performed a voxel-wise fit of the Kety-Tofts perfusion model and the Patlak model variation within segmented tumors to retrospectively abbreviated time courses (ATCs) derived from full time course (FTC) DCE-MRI series. All images were preprocessed with B1 correction. We used the concordance correlation coefficient (CCC) to evaluate the agreement between perfusion model parameters from the FTC and ATC model fits, where CCC=1 indicates perfect agreement.
Results: We found that it was feasible to abbreviate the time courses by as much as 50% while obtaining a CCC of 0.90 or above. An even greater abbreviation of 85% was feasible while yielding a CCC of 0.80 or greater for most patient datasets. Thus, perfusion modelling may be feasible in abbreviated MRI without much loss in perfusion model estimation accuracy.