The need to use animal models to develop imaging markers that could be linked to electrophysiological abnormalities in epilepsy and able to predict epileptogenicity in human studies is widely acknowledged. This study aimed to investigate the value of early magnetic resonance imaging (MRI) in predicting the long-term increased seizure susceptibility in the clinically relevant model of post-traumatic epilepsy (PTE). Moderate traumatic brain injury (TBI) was induced by lateral fluid-percussion in two groups of adult rats (34 injured, 16 controls). In Experiment 1, MRI follow-up was performed using a 4.7 T magnet at 3 h, 3 days, 9 days, 23 days, 2 months, 3 months and 6 months after TBI. T2 and 1/3 of the trace of the diffusion tensor (D(av)) were quantified from a single slice using a fast spin-echo sequence. In Experiment 2, MRI was performed at 7 and 11 months post-injury. In both groups, seizure susceptibility was tested by injecting a single dose of pentylenetetrazol at 12 months post-injury. Electrographic and behavioural responses were monitored for 1 h. Total number of spikes, total number of epileptiform discharges (EDs) and latency to first spike were measured. Finally, the severity of mossy fibre sprouting was evaluated. In both experiments, EEG parameters such as total number of spikes or EDs proved to be reliable indicators of increased seizure susceptibility in injured animals when compared to controls (P < 0.05). In the hippocampus ipsilateral to TBI, D(av) correlated with these EEG parameters at both early (3 h), and chronic (23 days, 2, 3, 6, 7 and 11 months) time points after TBI, as well as with the density of mossy fibre sprouting. These results for the first time demonstrate that quantitative diffusion MRI can serve as a tool to facilitate prediction of increased seizure susceptibility in a clinically relevant model of human PTE.