Binnie C D, Batchelor B G, Bowring P A, Darby C E, Herbert L, Lloyd D S, Smith D M, Smith G F, Smith M
Electroencephalogr Clin Neurophysiol. 1978 May;44(5):575-85. doi: 10.1016/0013-4694(78)90125-6.
A multivariate pattern recognition technique has been developed, to distinguish the EEGs of patients with cerebral pathology from those of normal controls and to localize any abnormalities detected. Two methods of feature extraction have been used, power spectral density and slope descriptor analysis, together with various types of feature compression. These techniques have been evaluated on EEGs from 63 patients with proven pathology. Spectral analysis proved more reliable than slope descriptor analysis and predicted the site of cerebral pathology more accurately than did visual assessment of the EEGs. This apparent improvement over the diagnostic reliability of visual analysis in considered to justify further development and evaluation of this technique.
已经开发出一种多元模式识别技术，用于区分患有脑部病变的患者与正常对照者的脑电图，并定位检测到的任何异常。使用了两种特征提取方法，功率谱密度和斜率描述符分析，以及各种类型的特征压缩。这些技术已在63例经证实患有病变的患者的脑电图上进行了评估。频谱分析证明比斜率描述符分析更可靠，并且比脑电图的视觉评估更准确地预测了脑部病变的部位。这种在视觉分析诊断可靠性方面的明显提高被认为足以证明对该技术进行进一步开发和评估的合理性。
