Aims/Objectives/Background Mild traumatic brain injury (mTBI) accounts for one million emergency department attendances in the UK every year. Whilst 30–50% of patients suffer from persistent symptoms, unselected follow up would overwhelm the health care system. Magnetic resonance imaging (MRI), may help to stratify patients for clinical follow up and interventional trials. We therefore aimed to identify:

1. Neuroanatomical features of concussion on MRI and

2. the optimal timing for magnetic resonance imaging (<72h or 2–3 weeks after injury).

This is the largest study to date using serial scanning acutely in patients with mTBI.

Methods/Design Data originated from two prospective cohorts: the Collaborative European NeuroTrauma Effectiveness Research in Traumatic Brain Injury (CENTER-TBI) study (2014–2017) and a local cohort (2012–2013). Eligible patients presented to hospital within 24h of a mTBI (Glasgow Coma Score 13–15), satisfied local criteria for computed tomography scanning and received two MRIs: one within 72h (MR1) and one 2–3 weeks after injury (MR2). In addition, 104 controls were enrolled. Volumes and diffusion parameters for brain regions of interest were extracted via automated pipelines. Symptoms were measured using the Rivermead Post-Concussion Questionnaire acutely and the extended Glasgow Outcome Score at three months.

Results/Conclusions The study included 81 patients (73 from CENTER-TBI, 8 local) with a median age of 44 years (range 14–85) and 57 (70%) men. Within patients, cerebral white matter volume decreased (MR1/MR2 0.98, p=0.001) and ventricular volume increased (MR1/MR2 1.06, p<0.001). Compared to controls, white matter volume was normal on MR1 (patient/control 1.00, p=0.277) but reduced on MR2 (patient/control 0.97, p<0.001). Diffusion changes followed one of three trajectories: progressive injury, minimal change, or pseudonormalisation. Concussion symptoms worsened, improved and were variable in the three groups respectively (delta [IQR] + 5.00 [+2.00-+5.00], -4.5 [-9.25-+1.75], 0.00 [-6.25 to +9.00], p=0.018). MR1 predicted three-month outcome better than MR2 (AUC [95% CI]: 0.93 [0.83–1.00] vs 0.72 [0.51–0.92]).