Background:The ability to predict posttraumatic stress disorder (PTSD) is a critical issue in the management of patients with mild traumatic brain injury (mTBI),as early medical and rehabilitative interventions may reduce the risks of long-term cognitive changes.The aim of the present study was to investigate how diffusion tensor imaging (DTI) metrics changed in the transition from acute to chronic phases in patients with mTBI and whether the alteration relates to the development of PTSD.Methods:Forty-three patients with mTBI and 22 healthy volunteers were investigated.The patients were divided into two groups:successful recovery (SR,n =22) and poor recovery (PR,n =21),based on neurocognitive evaluation at 1 or 6 months after injury.All patients underwent magnetic resonance imaging investigation at acute (within 3 days),subacute (10-20 days),and chronic (1-6 months) phases after injury.Group differences of fractional anisotropy (FA) and mean diffusivity (MD) were analyzed using tract-based spatial statistics (TBSS).The accuracy of DTI metrics for classifying PTSD was estimated using Bayesian discrimination analysis.Results:TBSS showed white matter (WM) abnormalities in various brain regions.In the acute phase,FA values were higher for PR and SR patients than controls (all P ＜ 0.05).In subacute phase,PR patients have higher mean MD than SR and controls (all P ＜ 0.05).In the chronic phase,lower FA and higher MD were observed in PR compared with both SR and control groups (all P ＜ 0.05).PR and SR groups could be discriminated with a sensitivity of 73％,specificity of 78％,and accuracy of 75.56％,in terms of MD value in subacute phase.Conclusions:Patients with mTBI have multiple abnormalities in various WM regions.DTI metrics change over time and provide a potential indicator at subacute stage for PTSD following mTBI.