Kelsey Matthew, Politte David, Verner Ryan, Zempel John M, Nolan Tracy, Babajani-Feremi Abbas, Prior Fred, Larson-Prior Linda J
Mallinckrodt Institute of Radiology, Washington University School of Medicine, St. Louis, MO 63110, USA.
Annu Int Conf IEEE Eng Med Biol Soc. 2012;2012:4336-40. doi: 10.1109/EMBC.2012.6346926.
Brain electrical activity exhibits scale-free dynamics that follow power law scaling. Previous works have shown that broadband spectral power exhibits state-dependent scaling with a log frequency exponent that systematically varies with neural state. However, the frequency ranges which best characterize biological state are not consistent across brain location or subject. An adaptive piecewise linear fitting solution was developed to extract features for classification of brain state. Performance was evaluated by comparison to an a posteriori based feature search method. This analysis, using the 1/ƒ characteristics of the human ECoG signal, demonstrates utility in advancing the ability to perform automated brain state discrimination.
脑电活动呈现出遵循幂律缩放的无标度动力学。先前的研究表明，宽带频谱功率呈现出与状态相关的缩放，其对数频率指数随神经状态系统地变化。然而，最能表征生物状态的频率范围在脑区位置或个体之间并不一致。开发了一种自适应分段线性拟合解决方案来提取用于脑状态分类的特征。通过与基于后验的特征搜索方法进行比较来评估性能。这项使用人类脑皮层电图（ECoG）信号的1/ƒ特性的分析，证明了在提高自动脑状态判别能力方面的实用性。
