Background The diagnosis of atypical inflammatory demyelinating lesions can be difficult. Brain biopsy is often required to exclude neoplasms. Moreover, the relationship between these lesions and multiple sclerosis and NMOSD is not clear. Objectives Our objectives were to describe radiological and pathological characteristics of patients with acute inflammatory demyelinating lesions. Methods We retrospectively identified patients with brain biopsy performed for diagnostic uncertainty revealing a demyelinating lesion. A complete clinical, biological, radiological and pathological analysis was performed. Results Twenty patients (15 with a single lesion) were included. MRI disclosed a wide range of lesions including infiltrative lesions (40%), ring-like lesion (15%) Baló-like lesion (15%) and acute haemorrhagic leukoencephalitis (20%). In spite of a marked heterogeneity, some findings were common: a peripheral B1000 hyperintense rim (70%), a slight oedema with mild mass effect (75%) and an open-rim peripheral enhancement (75%). Histopathology revealed that all cases featured macrophages distributed throughout, extensive demyelination, axonal preservation and absence of haemorrhagic changes. In the majority of cases, macrophages were the predominant inflammatory infiltrate and astrocytes were reactive and dystrophic. Aquaporin-4 staining was systematically preserved. After a mean follow-up of 5 years (1–12), 16/20 patients had a diagnosis of monophasic acute atypical inflammatory demyelinating lesion. One patient was diagnosed with MS and 3 with AQP4 negative NMOSD. Discussion Although imaging findings in patients with atypical inflammatory demyelinating lesions are heterogeneous, some common features such as peripheral DWI hyperintense rim with open-rim enhancement and absence of oedema argue in favour of a demyelinating lesion and should preclude a brain biopsy. In this context, AQP4 staining is systematically preserved and argues against an AQP4-positive NMOSD. Moreover, long-term follow-up is characterized by low recurrence rate.