Objective — the aim of this study was to estimate the prognostic value of magnetic resonance imaging (MRI) classification of traumatic brain lesion localization and levels in patients with a brain injury of various severity in a few days to three weeks after the injury. Material and methods. The cohort of 278 patients with traumatic brain injury (TBI) of various severity aged 8—74 y.o. (average —31.4±13.8, median — 29 (21.3; 37.0) was included in the analysis. The severity of TBI at admission varied from 3 to 15 Glasgow coma scores (GCS) (average — 8±4, median — 7 (5; 12). The main indications and conditions for MRI were: inconsistency between computed tomography (CT) data and neurological status, the necessity to clarify the location and type of brain damage, the absence of metal implants, the stabilization of the patient’s vital functions, etc. MRI was performed during the first three weeks after the injury using T1, T2, T2-FLAIR, DWI, T2*GRE, SWAN sequences. The damage to the brain was classified according to 8 grades depending on the lesion levels (cortical-subcortical level, corpus callosum, basal ganglia and/or thalamus, and/or internal, and/or external capsules, uni- or bilateral brain stem injury at a different level). Outcomes were assessed by the Glasgow outcome scale (GOS) 6 months after injury. Results. The significant correlations were found for the entire cohort between MRI grading and TBI severity (by GCS) and outcome (by GOS) of the injury (R=–0.66; p<0.0001; R=–0.69; p<0.0001, respectively). A high accuracy (77%), sensitivity (77%) and specificity (76%) of the proposed MRI classification in predicting injury outcomes (AUC=0.85) were confirmed using the logistic regression and ROC analysis. The assessment of MRI-classification prognostic value in subgroups of patients examined during the first, second, and third weeks after injury showed significant correlations between the GCS and the GOS as well as between MRI-grading and GCS, and GOS in all three subgroups. In the subgroup of patients examined during the first 14 days after the injury, the correlation coefficients were higher compared with those obtained in a subgroup examined 15—21 days after the injury. The highest correlations between MRI grading, TBI severity, and the outcome were found in the subgroup of patients who underwent MRI in the first three days after the injury (n=58). Conclusion. The proposed MRI classification of traumatic brain lesion levels and localization based on the use of different MR sequences reliably correlated with the clinical estimate of TBI severity by GCS and the outcomes by GOS in patients examined during the first three weeks after injury. The strongest correlation was observed for patients examined during the first three days after the injury.