In patients with iron overload, liver iron concentration can be measured by R2- or R2*-MRI. While gradient-recalled-echo (GRE) based R2*-MRI holds practical advantages over spin-echo based R2-MRI, the R2* technique is more challenging in patients with severe iron overload with rapid signal decay. We introduce a technique known as spin density projection (SDP), which uses liver-to-subcutaneous fat spin density ratios as an internal reference to improve robustness of R2* fitting. We performed a computer simulation and retrospective analysis of GRE imaging data in 45 patients with suspected iron overload and compared R2* fitting without and with SPD. We found that SPD increased robustness by reducing standard deviation of R2* estimation.