Beyond the Clinic: Leveraging Speech Acoustics and Phonetics for Cognitive Monitoring in PKU

Phenylketonuria (PKU) is one of the most common genetic metabolic disorders, characterized by the body’s inability to metabolize phenylalanine (Phe) properly. Without proper management, elevated blood Phe levels can lead to variable cognitive functioning in several domains, including executive dysfunction, processing speed, and verbal fluency, as well as an increase in psychological symptoms, such as anxiety and depression. Additionally, uncontrolled PKU associated with variable blood level Phe and Tyrosine (Tyr) may result in motor dysfunction, seizures, and white matter abnormalities, further impacting quality of life. Given these risks, regular monitoring is essential to prevent neurological decline and ensure optimal cognitive and behavioral outcomes. Blood level Phe monitoring is typically done through finger-prick blood samples, but this can be invasive and burdensome. As a less intrusive alternative, Ecological Momentary Assessments (EMA) have been explored in PKU to provide a more detailed and dynamic representation of cognitive and psychological fluctuations in daily life. When combined with speech analysis, this approach could offer a more accessible biomarker for PKU monitoring. In this study, we analyzed 191 speech samples, including spontaneous speech responses and Semantic fluency tasks collected from 20 individuals with PKU across six sessions. Our findings reveal that acoustic and phonetic features extracted from these responses can infer performance on cognitive testing, or cognitive assessment metrics (up to r = 0.52), and blood level Tyrosine (r = 0.44). Furthermore, incorporating demographic information enhances predictive performance, reaching a correlation of r = 0.58 for the TestMyBrain (TMB) Multiple Object Tracking test. These results underscore the potential of speech-based biomarkers as a noninvasive tool for assessing everyday cognitive function in PKU, offering a promising, scalable, accessible method for continuous disease monitoring.