The Cognitive Arc: Speech Biomarkers as Potential Indicators of Cognitive Decline in Older Adults
Ria Tahiliani1, Meenakshi Shah2, Ana Salazar3, Kushal Thakur1, Urvashi Gurbani4, Sadhana Hingorani5
1G.M.E.R.S Hospital and University, Gotri, 2Medicine, G.M.E.R.S Hospital and University, Gotri, 3University of Monterrey, 4Biomedical Sciences, Navrachana University, 5SMIMER Hospital and Medical College
Objective:
To evaluate the potential of digital speech analysis in detecting and characterizing cognitive changes in older adults by examining key acoustic and temporal speech features.
Background:
Early cognitive decline is often subtle and might not be picked by conventional assessments. Paper-based tests can be challenging in high-volume clinics or for individuals with limited literacy and can often be influenced by practice effects and patient variability thus limiting timely detection and intervention for cognitive decline. Speech-based assessments offer a non-invasive way to detect these changes early on and more reliably.
Design/Methods:
The study included older adults with varied cognitive abilities. Speech samples were obtained as participants read a standardized text aloud. Acoustic and temporal features (specifically pitch, intensity, harmonicity, shimmer, speech-to-pause ratio, pause time, DTW, and vocal centralization) were extracted and quantified using tools like Praat and Python. Group differences were examined using t-tests, and associations with MoCA scores were explored through Spearman correlations. The most informative speech markers were further analyzed for cognitive differences.
Results:
The study included 67 participants, 40 of whom demonstrated cognitive decline. Among the speech features analyzed, DTW similarity showed the strongest positive correlation with MoCA scores (rho = 0.578, p < 0.001). Voice centralization (rho = −0.467, p < 0.001), Pause Time (rho = −0.449, p < 0.001), and Pitch Slope (rho = −0.374, p < 0.002) were negatively associated with cognitive performance, while Speech-to-Pause Ratio demonstrated a modest positive correlation (rho = 0.365, p < 0.002).
Conclusions:
Speech markers like DTW similarity, voice centralization, time of pause, slope of pitch, and speech-to-pause ratio exhibited strong correlations with cognitive outcomes. Together, these results suggest that speech-based assessments offer a non-invasive method for early detection and monitoring of cognitive changes. Thus, enabling timely interventions and reducing the burden of undiagnosed cognitive decline in older adults.
Disclaimer: Abstracts were not reviewed by Neurology® and do not reflect the views of Neurology® editors or staff.