Spike Train Matching and Waveform Tracking Disagree: A Graph-Based Motor Unit Agreement Framework

Background Motor units (MU) are the fundamental building blocks of movement, and reliably identifying the same MU across recordings is essential for neuroscience, rehabilitation, and neural interface research. Two mainstream methods for this purpose, spike train matching and waveform tracking, are widely used and often assumed to be interchangeable, yet this assumption has never been thoroughly tested. Methods We first decomposed MU activities from two independent datasets, utilizing the high-density surface electromyography (HDsEMG) decomposition technique. Then, to build up the agreement for comparing these two tracking methods using the MUs' activities, we decomposed them within the same experimental trial. The comparison was performed in two settings: first, the agreement between the two methods was evaluated using the real decomposed data and the simulated data. Then, the agreement between the two methods was evaluated using the real decomposed data. Results Analyzing over 14,000 motor units across two independent datasets, this study demonstrates that the two methods are not interchangeable, that their agreement depends critically on threshold selection and dataset characteristics, and that no single threshold works across datasets. Conclusions These findings demonstrate that the two methods are not interchangeable, that a high waveform similarity score alone does not guarantee MU identity, and that applying a fixed universal threshold is inappropriate. Dataset-specific calibration through a pilot study is necessary before either method is applied, with direct implications for prosthetics, rehabilitation, and neuromuscular disease research. These findings challenge decades of standard practice and establish that dataset-specific calibration is necessary