Syllable-based speech characteristics as potential biomarker for differential diagnosis of Parkinson's disease, multiple system atrophy, and cerebellar ataxia.

Speech disorders differ between Parkinson's disease (PD) and multiple system atrophy (MSA), but studies focusing on group differences based on syllables or including cerebellar ataxia (CA) are lacking until now. This cross-sectional study aimed to analyze syllable-based speech characteristics in patients with PD, MSA, and CA, as well as healthy controls, to determine their diagnostic utility. Speech samples were collected from 68 PD, 52 MSA, 23 CA, and 70 healthy controls. Participants performed four speech tasks: producing high- and low-pitched sounds for five Korean vowels, repeating 14 Korean consonants with the vowel /a/, raising and lowering pitch of the vowel /a/, and continuously repeating /pa-ta-ka/ for 5 s. Acoustic analysis and artificial intelligence-based exploratory analysis were conducted to identify the syllable combinations that best distinguished between disease groups. Among the four speech tasks, the sequential motion rate task (/pa-ta-ka/ repetition) demonstrated the highest classification accuracy in distinguishing PD, MSA, and CA from the other groups, with accuracies of 68.90%, 77.42%, and 73.39%, respectively. For single syllable sequence, the /ka-ka-ka/ sequence achieved the highest accuracy, distinguishing CA from other groups with an accuracy of 78.92%. Among combined syllable sequence, the /aaa-hahaha/ sequence exhibited accuracies of 78.63% and 83.33% in differentiating PD and CA, respectively, while the /dadada-aaa/ sequence showed an accuracy of 80.24% in distinguishing MSA from other groups. These findings suggest that syllable-based speech characteristics, along with acoustic parameters, can discriminate among parkinsonian disorders and CA, highlighting their potential as a promising diagnostic tool.