Premium
Sci‐Fri AM: MRI and Diagnostic Imaging ‐ 03: The influence of sampling percentage in deformable registration on kinetic model analysis results in DCE‐MRI of the breast
Author(s) -
Mouawad Matthew,
Biernaski Heather,
Brackstone Muriel,
Klassen Martyn,
Lock Michael,
Prato Frank S.,
Thompson R. Terry,
Gaede Stewart,
Gelman Neil
Publication year - 2016
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.4961834
Subject(s) - image registration , goodness of fit , computation , nuclear medicine , mathematics , statistics , magnetic resonance imaging , confidence interval , sampling (signal processing) , artificial intelligence , computer science , medicine , computer vision , algorithm , image (mathematics) , radiology , filter (signal processing)
Purpose: Dynamic contrast enhanced (DCE) MRI is applied extensively for diagnosis and treatment monitoring of breast cancer. However, patient motion can introduce artificial variation in the signal enhancement curves. Non‐rigid registration can improve the curves but computation time can be long. Reducing the percentage of the image sampled (PS) can reduce time at the theoretical cost of registration accuracy. This work investigates the influence of PS on kinetic model analysis results and goodness‐of‐fit. Methods: DCE images were acquired using a 3T Siemens Biograph mMR. Deformable registration was performed on one patient dataset with 3Dslicer using PS values of 5, 20, and 100%. For three regions of interest within the tumor, tissue contrast agent concentration values versus time were generated and analyzed using the TOFTS pharmacokinetic model. Model parameters, their 95% confidence intervals and the coefficients of variation (CV), which served as a measure of goodness of fit, were recorded. Results: Computation time was approximately 16, 8, and 4 minutes/image for 100, 20, and 5 PS. The CV decreased following registration and there was a trend of decreasing CV with increasing PS. However, no differences in parameter values obtained with 100% PS and parameters values obtained with lower PS were observed. Substantial differences were found between parameter values obtained with versus without registration Conclusions: Increasing PS led to improved goodness‐of‐fit for the kinetic model analysis, at the expense of substantially increased computation time. However, this improved fit did not appear to influence parameter values for this patient.