R2* bias correction in brain multi-echo GRE data

Motivation: R2* mapping often suffers from artifacts from flows and field inhomogeneity. Goal(s): We propose a R2* bias correction method, targeting voxels with rapid signal decays. Approach: When the R-squared value of the fitted exponential function does not exceed a threshold, later echoes in the voxel are excluded, and the fitting is repeated. After iteration, the voxels that still do not meet the threshold are masked out. Results: The bias correction model allows us to improve R2* mapping and provides a mask for inaccurate region. Impact: Our method enhances accuracy of R2* mapping by utilizing an iterative approach which excludes the later echoes with unreliable signals. By identifying and masking out regions where correction is ineffective, the approach improves the precision of R2* values.