The changes of directional diffusivities derived from diffusion tensor imaging (DTI), i.e. decreased axial diffusivity (lambda(||)) and increased radial diffusivity (lambda( perpendicular)), have shown significant correlation with axonal and myelin damage, respectively. However, after formalin fixation, reduced sensitivity of lambda(||) in detecting axonal damage in tissue has raised the concern of applying DTI ex vivo. In order to distinguish whether death or the fixation process diminishes the sensitivity of DTI in detecting lesions, in vivo, pre-fixed postmortem, and fixed postmortem DTI were conducted on mouse optic nerves 3 and 14 days after transient retinal ischemia. Our data showed that, from in vivo to pre-fixed postmortem, lambda(||) and lambda( perpendicular) decreased by 50 to 70% in both healthy and injured optic nerves (3 and 14 day injury). From pre-fixed postmortem to fixed postmortem, lambda(||) and lambda( perpendicular) decreased by 40 to 50% in normal and 3-day injured optic nerves, but only by 15 to 25% in 14-day injured optic nerves. Consequently, for the 14-day injured optic nerves, the differences between healthy and injured nerves were not preserved after fixation: the 40% decreased lambda(||) and 200% increased lambda( perpendicular) in injured nerves as compared to the normal nerves were measured in vivo and pre-fixed postmortem, but after the fixation process, 300% increased lambda( perpendicular) and insignificant changes in lambda(||) were found in injured nerves as compared to the normal nerves. This study clarified that fixation process, but not death, could change the sensitivity of DTI in detecting injury.