Spectral pattern complexity analysis and the quantification of voice normality in healthy and radiotherapy patient groups

Abstract

Vocal fold functionality may alter in response to direct radiotherapy or indirectly by perturbation of the hypothalamic–pituitary axis. Perceptual assessment of voice quality is difficult to summarise in a single, reliable figure of normality and normality itself is undefined. In this study spectral analysis of vocal fold vibration, based on impedance variations measured across the larynx using an electro-glottogram, is used to build a single parameter description of standard vowel phonation in the normal male population. Patient data and perceptual assessment are then compared to this standard. The spectral pattern of the vowel/i/ electro-glottogram time series is analysed using approximate entropy after dynamic fundamental-harmonic frequency normalisation. The approximate entropy provides a single estimate of the spectral pattern complexity. A cohort of 89 normal males formed two statistically distinct groups, G1, with strong spectral pattern and high complexity 0.338 (±0.036), and G2 with a weak spectral pattern and low complexity 0.175 (±0.049). Membership ratio G1:G2 was 2:1. A cohort of 30 male larynx cancer cases were analysed approximately 3–6 months after irradiation, and three male prophylactic cranial irradiation cases some years after treatment. Two-thirds of patients had G2 or lower levels of complexity. The lower G2 complexity level appears to be the subjective, as well as the objective, threshold for voice normality.