Mild traumatic brain injury (mTBI) is a heterogeneous disease with a wide variety of clinical symptomatology associated with physiological alteration of brain function after the trauma. Recovery time is also widely variable and while many mTBI subjects are symptom free in 1-2 weeks, part of the mTBI population can suffer from neurological symptoms that significantly affect quality of life for several months. There is still limited understanding of the relationship between physiological and structural change in brain structures and rate of recovery from symptoms or outcomes.¹ In this work, we aim to identify structural signatures in sub-cortical gray matter and their relationship to alteration of brain physiology on the path to recovery in a mTBI population at 4 encounters (3 days to 3 months post-injury). We conduct two separate analyses: 1) correlation analysis between brain volumes and clinical symptoms at each encounter; and 2) longitudinal analysis of brain volumes across multiple encounters.T1-weighted images with 1mm isotropic resolution (MPRAGE²) were acquired using a GE 3T Signa MR750 scaner and a 32 channel brain coil (Nova Medical) on 78 uncomplicated mTBI subjects at four encounters: 3 days (32 subjects), 7 days (61 subjects), 1 month (56 subjects), and 3 months (42 subjects) post injury, and 23 controls (scanned twice, one week apart). The images were preprocessed (skull stripping using ROBEX³, bias/inhomogeneity correction using N4⁴, registration of labeled atlas images using ANTs⁵) and subcortical gray matter segmented using a deformable atlas method.^6,7 The segmented brain structures (illustrated in Figure 1) were further corrected for the effects of i) sex, ii) age, iii) education, and iv) handedness, using multiple linear regression model.⁸ Subjects were also evaluated by an experienced neurologist (TS) at each encounter and the symptom burden (22 symptoms on 0-6 scale⁹, total score 0-124) recorded. Based on the data, two different analyses were conducted: 1) Correlation analysis: the Pearson correlation coefficient was calculated between volume of brain structures and clinical symptoms at each encounter; 2) Longitudinal analysis: volume changes at multiple time points were studied with the one tail paired t-test.For the correlation analysis, no significant correlations were found between brain volumes and symptom scores at encounters 1-3 (E1-E3), while significant (p<0.05) negative correlations were found at encounter 4 (E4) between clinical scores and whole brain, supratentorial, and left/right thalamus volumes. Figure 2 demonstrated the significance levels for all the brain structures and Figure 3 illustrates the Pearson correlations. For the longitudinal analysis, we find significant reduction in volume of mTBI patients in the left thalamus (E2>E4, p<0.01) and left hippocampus (E2>E4, p<0.05) in the patient population. Supratentorial brain volume increased significantly between early and late encounters (E1<E4, p<0.05; E2<E4, p<0.01). Figure 4 shows the effect sizes of the paired differences between encounters, corresponding to the significant structures. There are no significant longitudinal changes between encounters 1 and 2 of controls.A longitudinal MRI study based on volumetric analysis was conducted to investigate the structural brain changes in a mild TBI population. The results of this study showed significant volume changes between the acute and sub-acute time points, and, at 3 months post-injury, volumes of the thalamus and whole brain were significantly anti-correlated with clinical symptom scores.