Maccabee P J, Hassan N F, Cracco R Q, Schiff J A
Electroencephalogr Clin Neurophysiol. 1986 May;65(3):177-87. doi: 10.1016/0168-5597(86)90052-3.
Medium nerve somatosensory evoked potentials (SSEPs) and intraoperative spinal evoked potentials were analyzed using different analog and zero phase shift digital high pass filter and by power spectrum. Additionally, high pass analog and digital filtering was performed on various sine, triangular and rectangular waves manufactured by a wave form generator. Recordings were also transformed to the 1st and 2nd time derivatives. The great abundance of spectral energy for scalp recorded median nerve SSEPs was below 125 c/sec but lower energy fast frequency components consistently extended to 500 c/sec. Power spectrum of the Erb's point compound nerve action potential revealed a wide band of spectral energy commencing at about 50-100 c/sec, peaking at about 250-270 c/sec and extending to nearly 1000 c/sec. This suggests that synchronous axonal activity generates predominantly faster frequencies above 100 c/sec. High pass analog filter confers phase non-linearity which results in various distortions including latency shift and a morphological change which may be visually similar to the 1st or 2nd time derivatives. High pass zero phase shift digital filter removes selected low frequencies without accompanying phase distortion. This accentuates fast peaks seen at open bandpass as well as transition points between baseline and component ascent or descent. Zero phase shift digital filter may also generate peaks that are not visualized at open pass but which reflect the sum of frequencies which were not removed by filtering. These peaks do not necessarily correspond to discrete singular neuroanatomical structures. Although peaks observed in high pass analog and digital filter appear similar and comparable, their underlying activity may be of different origin. This is because high pass analog filter projects a considerable amount of overlap from earlier onto later waves. For clinical correlation it is important that restricted bandpass analog or digitally filtered recordings be compared with open pass data. Only those peaks visualized in both open and restricted bandpass can be considered authentic. Examples of spinal and scalp SSEPs indicate that selective filtering may, under certain circumstances, distinguish axonal or lemniscal from synaptic generators.
采用不同的模拟和零相位数字高通滤波器并通过功率谱分析正中神经体感诱发电位（SSEPs）和术中脊髓诱发电位。此外，对波形发生器产生的各种正弦波、三角波和矩形波进行了高通模拟和数字滤波。记录还被转换为一阶和二阶时间导数。头皮记录的正中神经SSEPs的大量频谱能量低于125赫兹，但较低能量的高频成分一直延伸到500赫兹。Erb点复合神经动作电位的功率谱显示，频谱能量宽带从约50 - 100赫兹开始，在约250 - 270赫兹处达到峰值，并延伸至近1000赫兹。这表明同步轴突活动主要产生高于100赫兹的较快频率。高通模拟滤波器会产生相位非线性，这会导致各种失真，包括潜伏期偏移和形态变化，其在视觉上可能类似于一阶或二阶时间导数。高通零相位数字滤波器去除选定的低频而不会伴随相位失真。这会突出在开放带通以及基线与成分上升或下降之间的过渡点处看到的快速峰值。零相位数字滤波器还可能产生在开放通带中不可见但反映未通过滤波去除的频率总和的峰值。这些峰值不一定对应于离散的单个神经解剖结构。尽管在高通模拟和数字滤波器中观察到的峰值看起来相似且具有可比性，但它们潜在的活动可能起源不同。这是因为高通模拟滤波器会使较早的波与较晚的波有相当多的重叠。为了进行临床相关性分析，重要的是将受限带通模拟或数字滤波记录与开放通带数据进行比较。只有在开放和受限带通中都可见的那些峰值才能被视为真实的。脊髓和头皮SSEPs的例子表明，在某些情况下，选择性滤波可以区分轴突或薄束核与突触发生器。
