Motor Unit Template Estimation Using Integral Shape Averaging

Motor unit action potential (MUAP) template estimation is essential for quantitative electromyography (QEMG) and the diagnosis of neuromuscular disorders. However, estimation accuracy is often degraded by background noise, motor unit interference, and temporal jitter. Existing approaches-such as ensemble, median, trimmed, weighted, and closest-averaging techniques-have shown limited robustness under such conditions. This study introduces Integral Shape Averaging (ISA) as a computationally efficient method designed to improve MUAP template accuracy. Simulated MUAP trains with physiologically realistic noise levels, jitter values (50-300 µs), and maximum voluntary contraction (MVC) intensities (5% and 7.5%) were generated to compare ISA against established methods. Performance was evaluated using signal-to-noise ratio (SNR), root mean square error (RMSE), and cross-correlation (CORR) with gold-standard templates. Across all tested conditions, ISA consistently achieved up to 65% lower RMSE, 8-15 dB higher SNR, and 0.10-0.25 higher CORR than mean, trimmed, and shape averaging algorithms, with its advantages being most pronounced under high-noise and high-jitter conditions. Additionally, ISA's vectorized integration-averaging-differentiation framework eliminates iterative alignment, achieving linear computational complexity to the number of signals and samples. This efficiency, combined with superior robustness, makes ISA a strong candidate for real-time clinical applications, including intraoperative EMG monitoring, bedside diagnostics, and automated QEMG systems.