Introduction: It remains controversial whether the periodic discharges (PDs) pattern is an ictal or interictal phenomenon. The aims of the study are to apply time-frequency and power spectrum analysis to study the PDs pattern and prediction of seizures.Methods: We retrospectively searched continuous electroencephalography (cEEG) recordings to identify patients exhibiting the PDs pattern. Artifact-free cEEG segments demonstrating the PDs pattern with stable baselines were chosen for time-frequency and power spectrum analysis.Results: In total, 72 patients (1.3%) exhibited the PDs pattern, with a mean age 36.0 ± 20.7 years (range, 8–76 years). The median spectral power of PDs with a length of 60 s was 70.94 μV2 and that of PDs with a length of 20 s was 195.80 μV2. During follow-up, patients with spectral power of PDs of length 60 and 20 s lower than 28.65 and 36.09 μV2, respectively, exhibited no seizure. For predicting seizures, when the spectral power for PDs of 60 and 20 s equaled to 17.26 and 21.40 μV2, respectively, the diagnostic sensitivity was 100% and specificity was 86%. The locations of maximal spectral power of PDs, crude seizure onset zone (SOZ) judged from scalp EEG, and the most prominent regions of hyper- or hypo-metabolism on FDG-PET were congruent.Conclusions: Spectral power might be a candidate seizure marker of the PDs pattern. High spectral power predicted a high risk of seizures, and low spectral power was associated with a low risk of seizures.