Category: Neurophysiology (Non-PD)
Objective: The aim of this study is to achieve a direct and objective discrimination between voice samples recoded from patients with essential tremor and dysphonia both manifesting voice tremor.
Background: Patients with essential tremor and adductor-type spasmodic dysphonia may manifest a prominent voice tremor. The diagnosis of voice tremor is currently based on perceptual and qualitative analysis. We have recently demonstrated in two independent studies that advanced voice analysis with machine learning objectively discriminates normal voices from those recorded from patients with essential tremor and dysphonia [1–3].
Method: We investigated 33 patients with adductor-type spasmodic dysphonia (7 males, 65.6±11.7y), 36 patients with essential tremor and voice tremor (9 males, 72.4±8.6y), and 74 age-matched controls (20 males, 71.0±12.4y). We recorded voice samples during sustained vowel emission using a high-definition audio recorder. The classification of voice samples was achieved by means of a dedicated machine learning algorithm.
Results: Receiver Operating Characteristic curves showed that machine learning objectively discriminated between controls and essential tremor (Accuracy: 96.1%; AUC: 0.95), controls and dysphonia (Accuracy: 98.5%; AUC: 0.98) and finally essential tremor and dysphonia (Accuracy: 97.5%; AUC: 0.97).
Conclusion: Advanced voice analysis using machine learning objectively recognize voice tremor in patients with essential tremor and dysphonia discriminating the two forms of voice tremor with high accuracy. Our findings suggest that voice tremor differs in patients with essential tremor and spasmodic dysphonia. This finding points to different pathophysiological mechanisms underlying voice tremor in the two conditions.
References: 1. Asci, F.; Costantini, G.; Di Leo, P.; Zampogna, A.; Ruoppolo, G.; Berardelli, A.; Saggio, G.; Suppa, A. Machine-Learning Analysis of Voice Samples Recorded through Smartphones: The Combined Effect of Ageing and Gender. Sensors 2020, 20, 5022, doi:10.3390/s20185022. 2. Suppa, A.; Asci, F.; Saggio, G.; Marsili, L.; Casali, D.; Zarezadeh, Z.; Ruoppolo, G.; Berardelli, A.; Costantini, G. Voice Analysis in Adductor Spasmodic Dysphonia: Objective Diagnosis and Response to Botulinum Toxin. PARKINSONISM & RELATED DISORDERS 2020, 73, 23–30, doi:10.1016/j.parkreldis.2020.03.012. 3. Suppa, A.; Asci, F.; Saggio, G.; Di Leo, P.; Zarezadeh, Z.; Ferrazzano, G.; Ruoppolo, G.; Berardelli, A.; Costantini, G. Voice Analysis with Machine Learning: One Step Closer to an Objective Diagnosis of Essential Tremor. Mov Disord 2021, doi:10.1002/mds.28508.
To cite this abstract in AMA style:
F. Asci, P. Di Leo, G. Ruoppolo, G. Saggio, G. Costantini, A. Berardelli, A. Suppa. Unravelling Voice Tremor In Movement Disorders: A Machine Learning Study [abstract]. Mov Disord. 2021; 36 (suppl 1). https://www.mdsabstracts.org/abstract/unravelling-voice-tremor-in-movement-disorders-a-machine-learning-study/. Accessed June 15, 2025.« Back to MDS Virtual Congress 2021
MDS Abstracts - https://www.mdsabstracts.org/abstract/unravelling-voice-tremor-in-movement-disorders-a-machine-learning-study/