Figure 2.

Proposed framework for EEG-to-SEEG translation. (A) Raw EEG and SEEG signals are filtered and segmented using a synchronized sliding window. Then, segments from two sources are matched into pairs with the proposed strategy. Short-Time Fourier Transform (STFT) is performed to obtain magnitude and IF spectra. The aligned dataset is constructed with the processed pairs. (B) EEG-to-SEEG generative adversarial network (E2SGAN) is trained on the aligned dataset to synthesize SEEG from simultaneous EEG. Correlative Spectral Attention (CSA) and Weighted Patch Prediction (WPP) are devised to give a further boost to the discriminator.