Background There is a need for accurate quantitative measures of speech in Huntington’s disease that can be used for home and in-clinic assessment. Mobile devices provide an easy-to-use, low-cost alternative to laboratory standard microphone-amplifier equipment but require validation prior to clinical use
Aims To establish if mobile devices are suitable for monitoring speech in HD, and to examine how acoustic features from speech vary between HD and control participants. Acoustic features from speech tasks recorded using mobile devices from a HD and control group were then compared.
Methods Voice was recorded using a mobile device in a control group (Xiaomi Mi 8 or Samsung Tab S6 Lite, N=11, 8 females, Ireland), and a HD group (Wales: Samsung Tab A, N=8, 2 females; Spain: Huawei Mate 10 Lite, N=19, 12 females). In the control group, voice was recorded simultaneously with a gold standard microphone-amplifier and mobile devices. Acoustic data were recorded during sustained vowel phonation (/a:/) and syllable repetition (/pa/). Features estimated are detailed in the supplementary material. Statistical analysis was performed using linear mixed models.
Results In the control group, Bland-Altman analysis revealed good agreement between devices for all features. Linear mixed models showed no significant difference between devices during both tasks (p>0.05). When control and HD data were compared, no significant difference was observed in signal quality. Preliminary data suggests, similar to previous studies, an effect of gender1 and differences in acoustic features between HD and control groups.2
Conclusion Preliminary findings indicate the potential clinical use of mobile devices for objectively measuring speech in HD. Data collection is ongoing to allow comparison across clinical sites, with age and gender matched control participants.
Supplementary Material: Features estimated were: signal-to-noise ratio (SNR), mean, standard deviation and perturbation of fundamental frequency (f0, stdf0), PPQ5, harmonics-to-noise ratio (HNR), jitter (j%), and shimmer (s%). Figure 1 presents mean and standard deviation for participants groups, data collection sites and Bland-Altman plots for fundamental frequency and jitter features across recording devices. Bland-Altman plots revealed low bias estimated from different devices. Analysis of preliminary data suggests a tendency of decreased f0 and HNR, and increased j%, s%, stdf0 in HD group, consistent with what has been reported previously in a French cohort of 24 controls and 61 HD participants.2
Fitch James L, Anthony Holbrook. Modal vocal fundamental frequency of young adults. Archives of Otolaryngology 1970;92.4:379–382.
Riad Rachid, et al. Vocal markers from sustained phonation in huntington’s disease. arXiv preprint arXiv:2006.05365 2020.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.