Bispectral de-noising of the compound action potential for estimation of the nerve conduction velocity distribution

The distribution of the conduction velocities (DCV) of a peripheral nerve is a powerful diagnostic tool for the assessment of neuromuscular disorders. Its efficient calculation depends on the signal-to-noise ratio (SNR) of the acquired electroneurograms (ENGs), thus, time averaging is solely used. An alternative way of improving the SNR is based on averaging in the bispectrum domain and it is proposed in this work. The compound action potential (CAP) is a linear summation of the single fiber action potentials (SFAPs) propagating along the nerve fibers and can be expressed, in the discrete time, as the circular convolution of a delay sequence (DS) and the sampled SFAP. In the proposed method, averaging of low SNR CAP measurements is done in third order spectrum domain so no time alignment is required. Averaged bispectra are introduced in modified Hirose's method, to estimate the delay sequence for a conduction distance l1. The lost linear phase is recovered by using the delay phase cepstrum. Finally, the DCV can be calculated from the estimated DS, according to the formulation of the forward problem. Comparison between time and bispectrum averaging is performed using simulated data, proving the more efficient performance of the proposed method, especially in the case of noisy ENGs.