BACKGROUND: Although most patients with post-traumatic olfactory dysfunction (PTOD) undergo MRI, there is no consensus about its diagnostic or prognostic value.The aims were: 1) to classify the extent of post-traumatic neurodegeneration; 2) todetermine its relationship with chemosensory dysfunction (smell, taste, trigeminal); and 3) to establish whether MRI can predict olfactory improvement. METHODOLOGY: We conducted a retrospective cohort study based on a series of 56 patients with PTOD. All patients underwent validated psychophysical tests of their smell, taste, and trigeminal functions, otorhinolaryngologic evaluation, and MRI. An experiencedradiologist blinded to patient data evaluated 40 chemosensory-relevant brain regions according to a four-point scale (0=no lesion to 3=large lesion). Follow up data after 4 years (on average) were available in 46 patients. RESULTS:The cluster analysis showed 4 brain lesion patterns that differed in lesion localization and severity. They are associatedwith diagnostic categories: anosmia, hyposmia and normosmia. Two clusters were highly specific for anosmia (100% specificity)and could accurately predict this condition (100% positive predictive value). No clusters were associated with trigeminal or tastedysfunction. Regarding improvement, 72.7% of patients in the cluster with mild lesions experienced subjective and measurable olfactory improvement whereas this was only the case in 21.7-37.5% of patients with larger lesions. The odds of subjective smellimprovement were 5.9 times higher in patients within the milder cluster compared to larger ones. CONCLUSIONS: The analysis of brain lesions in PTOD allows corroboration of smell test results and prediction of subjective and measurable improvement.