Burst-suppression (BS) is a highly structured electroencephalogram (EEG) pattern characterized by alternating periods of high-voltage activity (bursts) and low or flatline activity (suppression). In non-anesthetized adults, BS indicates severe brain pathology, e.g. due to oxygen and glucose deprivation [1]. There are two types of BS: one with identical bursts that are highly periodic and uniform in shape, often signaling a poor prognosis, and one without identical bursts [2]. The neuronal mechanisms behind BS and these variations remain largely unknown. Some studies suggest the involvement of the thalamus and other deep brain structures [3], [4], which is contradicted by the observation of BS in the isolated cortex [5]. Theories also point to the roles of ketone bodies [6] and impaired brain metabolism [7], but these do not account for all characteristics of BS. Computational studies have identified that BS dynamics are low-dimensional, involving a fast dynamic responsible for bursting and a slow dynamic responsible for suppression [8], [9]. However, the precise neural origins of this fast-slow system and why it becomes prominent in BS patients are unclear. Interestingly, BS closely resembles the network bursting behavior seen in random in vitro neuronal networks, both spectrally and spatially. This raises the question: how can the entire brain behave almost identically to a small network of randomly connected neurons, and what can be learned from this similarity? We propose that the fast-slow system driving BS is identical to that in in vitro neuronal networks, involving recurrent excitation and neuronal adaptation. In a healthy brain, the synchronizations caused by these properties are suppressed due to inhibition and the brain’s complex structure. Through in vitro neuronal network experiments and computational modeling, we demonstrate that the absence of structure and inhibition leads to highly organized pathological signals with identical bursts, whereas the presence of structure or inhibition results in more physiological signals. We continuously compare our in vitro measurements and simulations to EEGs of patients in post-anoxic coma displaying BS patterns. We hypothesize that hypoxia-induced massive loss of synapses, especially inhibitory synapses, underlies the emergence of BS in these patients.