隨著神經科學對高通道數、高解析度及長時間單一神經元記錄的需求日益增加，傳統平面多電極陣列 (multi-electrode array, MEA)，如 Michigan-type 探針、Utah 陣列與微絲電極，常因探針漂移或神經元漂移導致訊號遺失，導致神經動作電位分類(spike sorting)的困難度增加。雖然目前已有廣泛應用的自動化或半自動化分類工具，主要基於主成分分析與模板匹配技術，但在長時間記錄條件下，其分類準確性仍會隨之下降。
為此，本研究旨在比較 Kilosort 演算法 (主要應用於平面探針)與基於立體探針之三邊定位法 (Trilateration)在不同三維探針結構下的神經動作電位分類效能。本研究結合了簡單立方 (Simple Cubic)、面心立方 (Face-Centered Cubic)、體心立方(Body-Centered Cubic) 與簡單六方 (Simple Hexagonal) 等結構進行探針設計，並利用三邊定位法從記錄數據中計算神經元的相對空間座標，以提供基於空間資訊的神經訊號分類依據。此外，Kilosort 演算法也經擴展以支援三維探針資料，並作為效能比較的基準，用以驗證三邊定位法在分類準確率與抗雜訊能力上的表現。
實驗結果顯示，當探針間距 150 µm 且採用面心立方結構時，結合三邊定位法可達到最佳分類效能，在無雜訊條件下的準確率 92.41%。相較於三維延伸版本的Kilosort，在相同條件下僅 32.06% 的準確率。然而，在高雜訊環境 (SNR = -1 dB) 下，Kilosort 顯現出較佳的穩定性，仍保有 15.49% 的準確率，而三邊定位法則降至 9.68%。本研究提出一種更有效率的神經訊號分析策略，並有潛力成為應用於單一神經元刺激(single-neuron stimulation) 高精度電生理研究的重要工具。

With the growing demand in neuroscience for high-density, high-resolution, and long-term single-neuron recordings, traditional planar multi-electrode arrays (MEAs), such as Michigan-type probes, Utah arrays, and microwire electrodes, often suffer from signal loss due to probe drift or neuronal drift, reducing spike sorting accuracy over time. Although automated or semi-automated tools based on principal component analysis (PCA) or template matching are widely used, their sorting accuracy inevitably declines over prolonged recording sessions.
To address this challenge, we compare the spike sorting performance of the trilateration method, originally developed for wireless localization, with Kilosort, a widely used algorithm originally designed for planar arrays. We propose a series of 3D electrode layouts, including Simple Cubic, Face-Centered Cubic, Body-Centered Cubic, and Simple Hexagonal structures. The trilateration method leverages these spatial arrangements to extract the neuronal coordinates, offering a spatially informed alternative for spike sorting, which we term the trilateration-based spike sorting (TSS) method. Meanwhile, the Kilosort algorithm is extended to support 3D layouts, serving as a benchmark for assessing the accuracy and noise robustness of the TSS approach.
The experimental results show that when the electrode pitch is 150 µm and a FCC configuration is used, the combination with the TSS method achieves the best sorting performance, reaching an accuracy of 92.41% under noise-free conditions. In comparison, the 3D-extended version of Kilosort achieves 32.06% accuracy under the same conditions. However, under high-noise conditions (SNR = –1 dB), Kilosort demonstrates better robustness, maintaining 15.49% accuracy, while the trilateration method drops to 9.68%. This study proposes a more efficient neural signal analysis strategy and provides a promising framework for high-precision electrophysiological experiments, such as single-neuron stimulation.

[1] H. Berger, “Über das Elektrenkephalogramm des Menschen,” Arch. Psychiatr. Nervenkr., vol. 87, pp. 527–570, 1929.
[2] C. J. Milligan and S. Pachernegg, “Utilising automated electrophysiological platforms in epilepsy research,” Methods Mol. Biol., vol. 2188, pp. 133–155, 2021.
[3] R. Fiáth et al., “Fine-scale mapping of cortical laminar activity during sleep slow oscillations using high-density linear silicon probes,” J. Neurosci. Methods, vol. 316, pp. 58–70, 2019.
[4] J. G. Ojemann et al., “Detection of brain tumor margins using electrocorticographic gamma frequency activity,” Neurosurgery, vol. 57, no. 2, pp. 305–314, 2005.
[5] J. J. Jun et al., “Fully integrated silicon probes for high-density recording of neural activity,” Nature, vol. 551, pp. 232–236, 2017.
[6] N. A. Steinmetz et al., “Neuropixels 2.0: A miniaturized high-density probe for stable, long-term brain recordings,” Science, vol. 372, no. 6539, Art. no. eabf4588, 2021.
[7] B. C. Raducanu et al., “Time multiplexed active neural probe with 1356 parallel recording sites,” Sensors, vol. 17, Art. no. 2388, 2017.
[8] E. Musk, “An integrated brain–machine interface platform with thousands of channels,” J. Med. Internet Res., vol. 21, no. 10, Art. no. e16194, 2019.
[9] S. Little et al., “Adaptive deep brain stimulation in advanced Parkinson disease,” Ann. Neurol., vol. 74, no. 3, pp. 449–457, 2013.
[10] A. M. Lozano et al., “Subcallosal cingulate gyrus deep brain stimulation for treatment-resistant depression,” Biol. Psychiatry, vol. 64, no. 6, pp. 461–467, 2008.
[11] S. Ghosh et al., “Review of multimodal data acquisition approaches for brain–computer interfaces,” BioMed, vol. 4, no. 4, pp. 548–587, 2024.
[12] H. G. Rey, C. Pedreira, and R. Quian Quiroga, “Past, present and future of spike sorting techniques,” Brain Research Bulletin, vol. 119, pp. 106–117, 2015.
[13] M. Pachitariu et al., “Kilosort: Realtime spike-sorting for extracellular electrophysiology with hundreds of channels,” bioRxiv, 2016. [Online]. Available: https://doi.org/10.1101/061481.
[14] M. Pachitariu, S. Sridhar, J. Pennington et al., “Spike sorting with Kilosort4,” Nat. Methods, vol. 21, pp. 914–921, 2024.
[15] J. Lee et al., “YASS: Yet another spike sorter,” in *Proc. NeurIPS*, 2017. [Online]. Available: https://www.biorxiv.org/content/early/2017/06/19/151928.
[16] J. E. Chung et al., “A fully automated approach to spike sorting,” Neuron, vol. 95, no. 6, pp. 1381–1394.e6, 2017.
[17] J. F. Magland and J. J. Jun, “IronClust: Spike sorting software,” GitHub, 2021. [Online]. Available: https://github.com/flatironinstitute/ironclust.
[18] G. T. Einevoll et al., “Towards reliable spike-train recordings from thousands of neurons with multielectrodes,” Curr. Opin. Neurobiol., vol. 22, no. 1, pp. 11–17, 2012.
[19] D. N. Hill, S. B. Mehta, and D. Kleinfeld, “Quality metrics to accompany spike sorting of extracellular signals,” J. Neurosci., vol. 31, no. 24, pp. 8699–8705, 2011.
[20] J. W. Pillow et al., “A model-based spike sorting algorithm for removing correlation artifacts in multi-neuron recordings,” PLoS ONE, vol. 8, no. 5, Art. no. e62123, 2013.
[21] C. Ekanadham, D. Tranchina, and E. P. Simoncelli, “A unified framework and method for automatic neural spike identification,” J. Neurosci. Methods, vol. 222, pp. 47–55, 2014.
[22] M. A. Nicolelis et al., “Chronic, multisite, multielectrode recordings in macaque monkeys,” Proc. Natl. Acad. Sci. USA, vol. 100, no. 19, pp. 11041–11046, 2003.
[23] G. Buzsáki, “Large-scale recording of neuronal ensembles,” Nat. Neurosci., vol. 7, no. 5, pp. 446–451, 2004.
[24] E. M. Maynard et al., “The Utah intracortical electrode array: A recording structure for potential brain–computer interfaces,” Electroencephalogr. Clin. Neurophysiol., vol. 102, no. 3, pp. 228–239, 1997.
[25] P. J. Rousche and R. A. Normann, “Chronic intracortical microstimulation using the Utah intracortical electrode array,” IEEE Trans. Rehabil. Eng., vol. 7, no. 1, pp. 56–68, 1998.
[26] K. Najafi, K. D. Wise, and T. Mochizuki, “A high-yield IC-compatible multichannel recording array,” IEEE Trans. Electron Devices, vol. 32, no. 7, pp. 1206–1211, 1985.
[27] G. Hong et al., “Implantable, multifunctional, flexible neural interfaces for brain research and clinical translation,” Nat. Rev. Neurosci., vol. 19, no. 5, pp. 273–286, 2018.
[28] J. P. Seymour and D. R. Kipke, “Neural probe design for reduced tissue encapsulation in CNS,” Biomaterials, vol. 28, no. 25, pp. 3594–3607, 2007.
[29] G. Guitchounts and D. Cox, “64-Channel carbon fiber electrode arrays for chronic electrophysiology,” Sci. Rep., vol. 10, Art. no. 383, 2020.
[30] D. Jones, *lattpy* [Computer software], GitHub, 2022. [Online]. Available: https://github.com/dylanljones/lattpy.
[31] A.-C. Yang et al., “Real-time spike sorting with 3D neural probe and triangulation localization,” bioRxiv, 2025. [Online]. Available: https://doi.org/10.1101/2025.03.30.645752
[32] R. T. Narayanan, D. Udvary, and M. Oberlaender, “Cell type-specific structural organization of the six layers in rat barrel cortex,” Front. Neuroanat., vol. 11, Art. no. 91, 2017.
[33] E. Hagen, S. Næss, T. V. Ness, and G. T. Einevoll, “Multimodal modeling of neural network activity: Computing LFP, ECoG, EEG, and MEG signals with LFPy 2.0,” Front. Neuroinform., vol. 12, Art. no. 92, 2018.
[34] M. L. Hines, A. P. Davison, and E. Muller, “NEURON and Python,” Front. Neuroinform., vol. 3, Art. no. 1, 2009.
[35] H. Markram et al., “Reconstruction and simulation of neocortical microcircuitry,” Cell, vol. 163, no. 2, pp. 456–492, 2015.