Both the formation and retrieval of episodic memories are closely associated with the activation of the hippocampal complex and medial temporal lobe (MTL) areas. These areas however are not sufficient by themselves to support most episodic memory functions. Instead, MTL areas are embedded as a central hub into a more distributed cortical network that supports both the storing, consolidation and subsequent reactivation of memories. Recently emerging evidence however indicates that cortical areas like posterior parietal regions and prefrontal areas contribute more to the processes than previously assumed. This begs the question of how distributed the episodic memory network actually is. To identify memory-specific activity patterns across the cortex, including MTL areas, we conducted a recognition experiment with drug-resistant epilepsy patients whose neural activity was recorded with intracranial stereotactic electroencephalography (sEEG). Each patient was exposed to up to 3000 natural scenes across several days and had to indicate whether the stimulus was new or repetitions, both within and across recording session days. The trial data was split into seven frequency bands, and the most relevant regional power values per condition (new/intra-session repetition/inter-session repetition) were identified. This was accomplished by training patient-individual Support Vector Machines (SVM) and Random Forests (RF) on the band power data to predict patient responses. Both model performances exceeded chance levels (SVM: 59.16%; RF: 63.7%). The most predictive features per category ranged across several frequency bands and were distributed across frontal, medial, and posterior parietal areas, supporting a more distributed network model.