Mollazadeh M, Aggarwal V, Thakor N V, Law A J, Davidson A, Schieber M H
Department of Biomedical Engineering, Johns Hopkins University, Baltimore, MD, USA.
Department of Biomedical Eng., University of Rochester, Rochester, NY, USA.
Int IEEE EMBS Conf Neural Eng. 2009 Apr-May;2009:506-509. doi: 10.1109/NER.2009.5109344. Epub 2009 Jun 23.
Local field potentials (LFP) represent the dendritic activity of a population of cells near the recording electrode. However, how LFP activity is related to single unit activity, and if it provides any additional information has not been well studied. Previously we have shown that temporal spectral modulation of LFP activity can be used to decode dexterous movements of the hand. Here, we analyze simultaneous spike and LFP recordings from M1 cortex in a rhesus monkey performing fine hand movements. Using multitaper spectral analysis, we found that both LFP and spiking activity show an increase in power in the <12 Hz and 70-200 Hz (high gamma) ranges, but, were significantly coherent only during the pre-movement time at low frequencies (<12 Hz). Furthermore, using either LFP or spiking activity, we were able to decode amongst three different hand grasps with high accuracy (99% using 97 spikes and 70% using 8 LFP channels). However, while spikes were better in decoding movement types, LFPs performed much better (94% success) than spikes (77%) when differentiating between rest and movement. We also found that combining spike and LFP activity can improve decoding performance when fewer spikes are considered, as may be the case when single unit recordings degrade over time (71% using 40 spikes and 76% using 8 LFPs, vs 88% using 40 spikes + 8 LFPs). Thus, the relative stability of LFP activity can help augment single-unit activity for the chronic operation of a multimodal BMI.
局部场电位（LFP）代表记录电极附近一群细胞的树突活动。然而，LFP活动与单个单元活动之间的关系，以及它是否提供任何额外信息，尚未得到充分研究。此前我们已经表明，LFP活动的时间频谱调制可用于解码手部的灵巧运动。在此，我们分析了恒河猴在进行精细手部运动时，来自M1皮质的同步尖峰和LFP记录。使用多窗谱分析，我们发现LFP和尖峰活动在<12 Hz和70 - 200 Hz（高伽马）范围内的功率均增加，但仅在低频（<12 Hz）的运动前时间显著相干。此外，使用LFP或尖峰活动，我们能够高精度地解码三种不同的手部抓握（使用97个尖峰时准确率为99%，使用8个LFP通道时准确率为70%）。然而，虽然尖峰在解码运动类型方面表现更好，但在区分休息和运动时，LFP的表现（成功率94%）比尖峰（77%）好得多。我们还发现，当考虑较少的尖峰时，结合尖峰和LFP活动可以提高解码性能，例如在单个单元记录随时间退化的情况下（使用40个尖峰时为71%，使用8个LFP时为76%，而使用40个尖峰 + 8个LFP时为88%）。因此，LFP活动的相对稳定性有助于增强多模态脑机接口长期运行时的单个单元活动。
