AI 中文总结
研究在固定通道预算下跟踪解离神经元网络非平稳自发活动的自适应电极选择方法，采用折扣泊松 - 伽马模型与汤普森采样，经离线回放和在线记录评估，能提高记录效率，为自适应神经活动感知提供基础。
AI 中文摘要
目的：开发并评估一种自适应电极选择方法，用于在固定通道预算下的长期高密度微电极阵列（HD-MEA）记录中跟踪非平稳自发活动。方法：将电极分配表述为顺序子集选择问题，并使用带汤普森采样的折扣泊松 - 伽马模型。该方法根据观察到的尖峰计数更新特定电极的活动估计，并随时间重新分配固定通道预算。通过对九个34小时的HD-MEA记录进行离线回放（从529个密集布线的候选电极中选择100个电极）以及在使用1024个布线电极的代表性在线记录中进行评估。主要结果：在离线记录中，前100个活跃电极集变化很大，在34小时时更替率达到47.8%。贝叶斯方法在测试策略中捕获了神谕选择器可用尖峰的最大比例，在最终时间点比静态选择超出17.2个百分点。在在线记录中，自适应选择捕获了第一个同步爆发并支持活动中心轨迹分析。意义：不确定性感知探索和时间折扣可以在固定读出约束下提高HD-MEA记录效率，为自适应感知不断演变的神经活动提供基础。
英文摘要
Objective. To develop and evaluate an adaptive electrode-selection method for tracking non-stationary spontaneous activity during long-term high-density microelectrode array (HD-MEA) recordings under a fixed channel budget. Approach. We formulated electrode allocation as a sequential subset-selection problem and used a discounted Poisson-Gamma model with Thompson sampling. The method updated electrode-specific activity estimates from observed spike counts and reallocated a fixed channel budget over time. We evaluated it by offline replay of nine 34 h HD-MEA recordings, selecting 100 electrodes from 529 densely routed candidates, and in a representative online recording using 1,024 routed electrodes. Main results. Across offline recordings, the top 100 active-electrode set changed substantially, reaching 47.8% turnover at 34 h. The Bayesian method captured the largest fraction of the spikes available to an oracle selector among the tested strategies and exceeded static selection by 17.2 percentage points at the final time point. In the online recording, adaptive selection captured the first synchronized burst and supported center-of-activity trajectory analysis. Significance. Uncertainty-aware exploration and temporal discounting can improve HD-MEA recording efficiency under fixed readout constraints, providing a basis for adaptive sensing of evolving neural activity.
