HIGH-DENSITY SPATIOTEMPORAL SINGLE-UNIT ACTIVITY PROFILING AND CELL-TYPE IDENTIFICATION IN THE HUMAN NEOCORTEX

High-density silicon probes have transformed the study of neuronal populations in animal models, yet their use in the human cortex remains limited. Using clinically adapted Neuropixels probes, we analyzed intraoperative extracellular recordings to characterize single-unit waveform properties, classify putative neuronal types, and identify spatiotemporal propagation at cellular resolution. After motion correction, spike sorting, and manual curation, we computed cluster quality and template-based waveform metrics in SpikeInterface, including mean waveform, peak-to-valley time, half-width, and peak amplitude. Following empirical thresholding of peak-to-valley distributions, unsupervised algorithms such as feature-based and waveform-based k-means were applied. WaveMAP produced the most coherent separation, thus we used its four waveform groups - two narrow, one broad, and one triphasic - for downstream analysis. Narrow waveforms were interpreted as putative interneurons, broad units as pyramidal cells, and triphasic waveforms as mixed or axonal signals, yielding an unsupervised, data-driven cell-type separation in the absence of histological ground truth. To quantify signal propagation, we computed spread and vertical propagation velocity as multi-channel propagation metrics. A subset of units showed extracellular signatures consistent with somatodendritic action potential backpropagation (bAP), expressed as systematic propagation across adjacent contact sites. bAP presence was independent of spike amplitude, firing rate, or spike count, but correlated with WaveMAP-defined classes, matching canonical rodent bAP patterns. In contrast, we observed a high-amplitude but spatially confined “small-footprint” narrow units. While their physiological origin remains unresolved, recent studies suggest they may instead reflect axonal signals from Ranvier nodes.