Maia Pedro D, Raj Ashish, Kutz J Nathan
Weill Cornell Medicine, Department of Radiology, New York, NY, USA.
Weill Cornell Medicine, Brain and Mind Research Institute, New York, NY, USA.
J Comput Neurosci. 2019 Aug;47(1):1-16. doi: 10.1007/s10827-019-00714-8. Epub 2019 Jun 4.
We introduce a computational model for the cellular level effects of firing rate filtering due to the major forms of neuronal injury, including demyelination and axonal swellings. Based upon experimental and computational observations, we posit simple phenomenological input/output rules describing spike train distortions and demonstrate that slow-gamma frequencies in the 38-41 Hz range emerge as the most robust to injury. Our signal-processing model allows us to derive firing rate filters at the cellular level for impaired neural activity with minimal assumptions. Specifically, we model eight experimentally observed spike train transformations by discrete-time filters, including those associated with increasing refractoriness and intermittent blockage. Continuous counterparts for the filters are also obtained by approximating neuronal firing rates from spike trains convolved with causal and Gaussian kernels. The proposed signal processing framework, which is robust to model parameter calibration, is an abstraction of the major cellular-level pathologies associated with neurodegenerative diseases and traumatic brain injuries that affect spike train propagation and impair neuronal network functionality. Our filters are well aligned with the spectrum of dynamic memory fields including working memory, visual consciousness, and other higher cognitive functions that operate in a frequency band that is - at a single cell level - optimally guarded against common types of pathological effects. In contrast, higher-frequency neural encoding, such as is observed with short-term memory, are susceptible to neurodegeneration and injury.
我们引入了一种计算模型，用于研究由于主要形式的神经元损伤（包括脱髓鞘和轴突肿胀）导致的放电率滤波在细胞水平上的影响。基于实验和计算观察结果，我们提出了描述尖峰序列失真的简单唯象输入/输出规则，并证明38 - 41赫兹范围内的慢伽马频率对损伤最为稳健。我们的信号处理模型使我们能够在最小假设的情况下，推导细胞水平上受损神经活动的放电率滤波器。具体而言，我们通过离散时间滤波器对八个实验观察到的尖峰序列变换进行建模，包括那些与不应期增加和间歇性阻断相关的变换。通过将尖峰序列与因果高斯核卷积来近似神经元放电率，也可得到滤波器的连续对应形式。所提出的信号处理框架对模型参数校准具有鲁棒性，它是与神经退行性疾病和创伤性脑损伤相关的主要细胞水平病理的一种抽象，这些病理会影响尖峰序列传播并损害神经元网络功能。我们的滤波器与动态记忆场的频谱很好地对齐，包括工作记忆、视觉意识和其他在单个细胞水平上在一个对常见类型病理效应具有最佳防护作用的频带中运行的高级认知功能。相比之下，高频神经编码，如在短期记忆中观察到的，容易受到神经退行性变和损伤的影响。
