Neurodegenerative diseases, including dementia with Lewy bodies, may affect speech already in prodromal stages; however, these changes can be difficult to detect and are therefore poorly understood. Speech analysis represents a sensitive, non-invasive method for capturing such manifestations, motivating the need for a systematic description of acoustic and linguistic parameters in this population. The aim of this master's thesis is to identify subtypes of speech and language impairment in patients with mild cognitive impairment with Lewy bodies (MCI-LB) based on extracted acoustic and linguistic parameters, and to place the findings in the context of available literature.
Within the thesis, parameters describing respiration, phonation, articulation, prosody, and the linguistic domain of speech, including lexical, semantic, and syntactic parametrization, were designed and implemented. The parameters were implemented in Python in two separate modules, acoustic and linguistic, and extracted for the entire database of speech recordings. The data were corrected for covariates and normalized against healthy controls, with weights for individual parameters determined based on AUC values. Subsequently, cluster analysis was performed in the acoustic, linguistic, and multimodal domains.
The acoustic analysis revealed three subtypes: a subtype with broad motor-speech impairment, a subtype with relatively preserved motor speech, and a subtype with selective prosodic and respiratory impairment. The linguistic analysis identified a linguistically preserved subtype and a lexically impaired subtype. The multimodal analysis further revealed a subtype with complex respiratory-lexical impairment, a subtype with predominantly cognitive-linguistic impairment, and a subtype with predominantly motor speech impairment. The limited overlap between acoustic and linguistic clusters suggests that motor-speech and linguistic-cognitive impairments represent largely independent domains in MCI-LB, developing independently of one another in individual patients.