Towards Improving The Evaluation Of Speech Production Deficits In Chronic Stroke

One of the most devastating consequences of stroke is aphasia - a disorder that impairs communication across the domains of expressive and receptive language. In addition to language difficulties, stroke survivors may struggle with disruptions in speech motor planning and/or execution processes (i.e., a motor speech disorder, MSD). The clinical management of MSDs has been challenged by debates regarding their theoretical nature and clinical manifestations. This is especially true for differentiating speech production errors that can be attributed to aphasia (i.e., phonemic paraphasias) from lower-level motor planning/programming impairments (i.e., articulation errors that occur in apraxia of speech; AOS). Therefore, the purposes of this study were 1) to identify objective measures that have the greatest discriminative weight in diagnostic classification of AOS, and 2) using neuroimaging, to localize patterns of brain damage predictive of these behaviors. Method: Stroke survivors (N=58; 21 female; mean age=61.03±10.01; months post-onset=66.07±52.93) were recruited as part of a larger study. Participants completed a thorough battery of speech and language testing and underwent a series of magnetic resonance imaging (MRI) sequences. Objective, acoustic measures were obtained from three connected speech samples. These variables quantified inter-articulatory planning, speech rhythm and prosody, and speech fluency. The number of phonemic and distortion errors per sample was also quantified. All measures were analyzed for group differences, and variables were subject to a linear discriminant analysis (LDA) to determine which served as the best predictor of AOS. MRI data were analyzed with voxel-based lesionsymptom mapping and connectome-symptom mapping to relate patterns of cortical necrosis and white matter compromise to different aspects of disordered speech. Results: Participants with both AOS and aphasia generally demonstrated significantly poorer performance across all production measures when compared to those with aphasia as their only impairment, and compared to those with no detectable speech or language impairment. The LDA model with the greatest classification accuracy correctly predicted 90.7% of cases. Neuroimaging analysis indicated that damage to mostly unique regions of the pre- and post-central gyri, the supramarginal gyrus, and white matter connections between these regions and subcortical structures was related to impaired speech production. Conclusions: Results support and build upon recent studies that have sought to improve the assessment of post-stroke speech production. Findings are discussed with regard to contemporary models of speech production, guided by the overarching goal of refining the clinical evaluation and theoretical explanations of AOS