Abdi-Sargezeh Bahman, Oswal Ashwini, Sanei Saeid
Department of Computer Science, Nottingham Trent University, Nottingham, UK.
Nuffield Department of Clinical Neurosciences, University of Oxford, Oxford, UK.
IEEE Stat Signal Processing Workshop. 2023 Jul 2;27:710-714. doi: 10.1109/SSP53291.2023.10207965.
Both scalp and intracranial electroencephalograms (EEGs) are of great importance for diagnosing brain disorders. However, the scalp EEG (sEEG) is attenuated by the skull and contaminated with artifacts. At the same time, intracranial EEG (iEEG) is almost free of artifacts and can capture all brain activities without any attenuation due to being close to the brain sources. In this study, the aim is to enhance the performance of sEEG by mapping the sEEG to the iEEG. To do so, we here develop a deep neural network using a generative adversarial network to estimate the sEEG from the iEEG. The proposed method is applied to sEEG and iEEG recorded simultaneously from epileptics to detect interictal epileptiform discharges (IEDs). The proposed method detects IEDs with 76% accuracy outperforming the state-of-the-art methods. Furthermore, it is at least twelve times less complex than the compared methods.
头皮脑电图（EEG）和颅内脑电图对于诊断脑部疾病都非常重要。然而，头皮脑电图（sEEG）会被颅骨衰减并受到伪迹的污染。同时，颅内脑电图（iEEG）几乎没有伪迹，并且由于靠近脑源，可以捕捉所有脑活动而没有任何衰减。在本研究中，目的是通过将sEEG映射到iEEG来提高sEEG的性能。为此，我们在此开发了一种深度神经网络，使用生成对抗网络从iEEG估计sEEG。所提出的方法应用于从癫痫患者同时记录的sEEG和iEEG，以检测发作间期癫痫样放电（IEDs）。所提出的方法以76%的准确率检测到IEDs，优于现有方法。此外，它的复杂度至少比比较方法低12倍。
