Ngo Cuong Q, Truong Bao C Q, Jones Timothy W, Nguyen Hung T
Annu Int Conf IEEE Eng Med Biol Soc. 2018 Jul;2018:3862-3865. doi: 10.1109/EMBC.2018.8513069.
Nocturnal hypoglycemia is dangerous that threatens patients because of its unclear symptoms during sleep. This paper is a study of hypoglycemia from 8 patients with type 1 diabetes (T1D) at night. O1 and O2 EEG data of the occipital lobe associated with glycemic episodes were analyzed. Frequency features were computed from Power Spectral Density using Welch's method. Centroid alpha frequency reduced significantly ($\mathrm{P}\lt 0.0001$) while centroid theta increased considerably ($\mathrm{P}\lt 0.01$). Spectral entropy of the unified theta-alpha band rose significantly ($\mathrm{P}\lt 0.005$). These occipital features acted as the input of a Bayesian regularized neural network for detecting hypoglycemic episodes. The classification results were 73% and 60% of sensitivity and specificity, respectively.
夜间低血糖很危险，因其在睡眠期间症状不明确而威胁着患者。本文是对8例1型糖尿病（T1D）患者夜间低血糖情况的研究。分析了与血糖事件相关的枕叶O1和O2脑电图数据。使用 Welch 方法从功率谱密度计算频率特征。质心α频率显著降低（P < 0.0001），而质心θ频率显著增加（P < 0.01）。统一θ-α频段的频谱熵显著上升（P < 0.005）。这些枕叶特征作为贝叶斯正则化神经网络检测低血糖事件的输入。分类结果的灵敏度和特异性分别为73%和60%。
