Research

Research focus areas for the Speech Innovation Group in Neurology, Artificial Intelligence and Linguistics include:

• Digital biomarkers and speech-physiology mapping. Our lab is working to develop objective, technology-enabled biomarkers that quantify neurological and motor speech impairment. By using acoustic analysis, we can map digital signals to physical speech physiology. For example, goodness of pronunciation is known to reflect articulatory imprecision, but it is not yet specifically sensitive to lingual or labial weakness that might occur secondary to neurological disease. Our research aims to create a high-resolution "digital fingerprint" of a patient's condition. This digital fingerprint provides a precise way to track disease progression over time that offers greater specifics of what aspects of speech are progressing. This digital fingerprint is more sensitive to change than traditional clinical observation.
• Differential diagnosis. A core pillar of our work focuses on the differential diagnosis of motor speech disorder subtypes as they occur in frontotemporal dementia spectrum disorders. Because syndromes such as primary progressive aphasia, progressive apraxia of speech, progressive supranuclear palsy and corticobasal syndrome often have overlapping symptoms, early and accurate diagnosis is a significant challenge. We use acoustic and linguistics-based features to identify the subtle nuances in speech and language that distinguish these subtypes, ensuring patients receive the correct diagnosis and the correct care plan as early as possible.
• Innovative treatment and clinical trials. Our lab investigates the efficacy of various therapeutic approaches. These therapies range from behavioral tools such as delayed auditory feedback to neuromodulation techniques such as transcranial direct current stimulation and pharmacological interventions for a range of neurological conditions, including multisystem atrophy and brain tumors. We aim to contribute to treatment response tracking to facilitate precise care recommendations.
• Multimodal integration. To gain a holistic view of brain health, we link speech and language data to other physiological measures, primarily neuroimaging. By correlating acoustic and linguistic changes with structural and functional scans such as MRI, positron emission tomography, electromyography and electroencephalography, we are uncovering the neural networks that underlie breakdowns in speech production. This multimodal approach ensures that our digital tools are grounded in the biological reality of the disease.