Introduction The intracerebral iron quantification in MRI is of great clinical interest for aging and neurodegeneration studies. Currently, the most widely used techniques are R2∗ relaxometry and QSM. Despite their limitations in terms of reproducibility [1] , they are the most adapted to clinical constraints. The work mainly focused on the most robust technique, R2∗ relaxometry, and its optimization and evaluation for the specific iron quantification task. Methods The physiological aging model was used in two structures of interest (putamen and thalamus) as a modification source. A brain MRI (Intera Achieva 3 Tesla, Philips) was acquired for 39 young subjects (29 ± 6 years old) and 38 elderly subjects (68 ± 6 years old). For each subject, gradient echoes (3D FFE EPI) were acquired with 6 different echo times (TE): 6, 12, 20, 30, 45 et 60 ms. R2∗ maps using the 6 TEs (R2∗s) then the first 3 (R2∗p) and finally the last 3 (R2∗t) were made by mono-exponential non-linear voxel-to-voxel regression. Results For the thalamus, all subjects combined, R2∗p and R2∗t were significantly different from R2∗s. This invalidates the hypothesis of a strictly mono-exponential decay signal. Between young and elderly subjects, a significant difference was found for R2∗s (p = 0.025) and R2∗p (p = 0.001) but none for R2∗t (p = 0.38). The taking into account of the first TEs increases the interindividual heterogeneity while the R2∗t present a small variance between the subjects. The same results were found for the putamen. The use of late TEs increases the magnetic susceptibility artefacts in areas near interfaces (such as sinuses). This artifact common to all subjects standardizes the different groups. More pronounced differences between groups are observed using R2∗p. In this case, the use of short TEs did not allow the signal sensitization to a specific quantity but the eviction of standardization due to the artifact. In areas distant from the artifact, short TEs sensitize R2∗ to fast signal components such as myelin-bound water. Conclusions Although the mono-exponential decay model assumes that R2∗ is an absolute quantity, this is not really the case for a voxel. Therefore, it is not necessary to try to make R2∗ a reproducible quantity. The research for reproducibility can be done at the expense of sensitivity. On the contrary, a relaxometry adapted to the quantity measured or to the studied zone should be considered by playing among other things on the TEs selections.