Estimating paediatric normative values for nerve studies using clustering techniques

Abstract

Objective:

To estimate normative values from mixed clinical paediatric electroneurography data using an unsupervised clustering approach.

Methods:

Electroneurography studies from paediatric patients (2009–2024) were analysed for common motor and sensory nerves. Motor parameters included distal motor latency, CMAP amplitude, duration, area, and conduction velocity; sensory parameters included SNAP amplitude and conduction velocity. Data were grouped into age windows, and within each, t-distributed stochastic neighbour embedding (t-SNE) was applied to identify the normative distribution. The mean, 5th, and 95th centiles were derived and modelled using exponential fits.

Results:

Normative values were estimated for ages 0–18 years. Motor amplitudes increased with age, and conduction velocities rose rapidly until 3–4 years before plateauing. Distal motor latency showed a brief early dip followed by an increase. Sensory amplitudes peaked between 1 and 8 years, while sensory conduction velocities increased sharply in the first year, then gradually declined.

Conclusion:

Unsupervised clustering can derive normative paediatric electroneurography values from heterogeneous clinical data, yielding trends consistent with published references.

Significance:

This data-driven approach is practical, generalisable, and enables identification of likely healthy individuals using multivariate electrophysiological parameters.