Clinical Neurophysiology, MIRA-Institute for Biomedical Technology and Technical Medicine, University of Twente, The Netherlands.
J Neurosci Methods. 2011 Aug 30;200(1):86-93. doi: 10.1016/j.jneumeth.2011.06.008. Epub 2011 Jun 23.
Automated interpretation of clinical EEG recordings will reduce subjectivity and visual bias from analysis and can reduce the time required for interpretation. As a first step in the design of a fully automated system, a method is presented to characterize the main properties of the posterior dominant rhythm (PDR), in particular its frequency, symmetry and reactivity. The presented method searches for dominant peaks in the EEG spectra during eyes-closed states with a three-component curve-fitting technique. From the fitted curve, the frequency and amplitude are estimated. The symmetry and the reactivity are found using the spectral power at the PDR frequencies. In addition, a certainty value is introduced as a measure of confidence for each estimate. The method was evaluated on a test set of 1215 clinical EEG recordings and compared to the PDR frequencies obtained from the visual analysis, as reported in the diagnostic reports. The calculated PDR frequencies were within 1.2Hz of the visual estimates in 92.5% of the cases. Even higher accuracies were reached when estimates with low certainty values were discarded. The presented method quantifies essential features of the PDR with a matched accuracy to visual inspection, making it a feasible contribution to the design of a fully automated interpretation system.
自动解释临床脑电图记录将减少分析中的主观性和视觉偏差，并可以减少解释所需的时间。作为设计全自动系统的第一步，本文提出了一种方法来描述后优势节律（PDR）的主要特性，特别是其频率、对称性和反应性。所提出的方法使用三分量曲线拟合技术在闭眼状态下搜索脑电图谱中的主导峰。从拟合曲线中，估计频率和幅度。使用 PDR 频率处的光谱功率找到对称性和反应性。此外，引入了置信值作为每个估计的置信度的度量。该方法在 1215 个临床脑电图记录的测试集中进行了评估，并与诊断报告中报告的视觉分析获得的 PDR 频率进行了比较。在 92.5%的情况下，计算出的 PDR 频率与视觉估计值相差 1.2Hz 以内。当丢弃具有低置信值的估计值时，甚至可以达到更高的准确性。所提出的方法以与视觉检查相匹配的准确性量化了 PDR 的基本特征，为设计全自动解释系统提供了可行的贡献。
