Conventional approaches to quantify whole brain T(2)(*) maps use nonlinear regression with intensive computational requirements that therefore likely limit quantitative T(2)(*) mapping for real-time applications. To overcome these limitations an alternative method, NumART(2)(*) (NUMerical Algorithm for Real-time T(2)(*) mapping) that directly calculates T(2)(*) by a linear combination of images obtained at three or more different echo times was developed. NumART(2)(*), linear least-squares, and nonlinear regression techniques were applied to multiecho planar images of the human brain and to simulated data. Although NumART(2)(*) may overestimate T(2)(*), it yields comparable values to regression techniques in cortical and subcortical areas, with only moderate deviations for echo spacings between 18 and 40 ms. NumART(2)(*), like linear regression, requires 2% of the computational time needed for nonlinear regression and compares favorably with linear regression due to its higher precision. The use of NumART(2)(*) for continuous on-line T(2)(*) mapping in real time fMRI studies is shown.
Copyright 2002 Wiley-Liss, Inc.