Dual-modal analysis of speech and fNIRS for dementia identification

People with dementia of the Alzheimer’s type (DAT) may have difficulty expressing different emotions via speech prosody, and how they manipulate affective prosody and its neural basis remains unclear. Ten individuals with DAT and another 10 neurologically healthy controls (NHCs) participated in this study. They were instructed to talk about a happy life event, a sad life event, and steps involved in making a peanut butter and jelly sandwich. While completing the speech tasks, a functional near-infrared spectroscopy (fNIRS) attached to the participant's forehead recorded their neural activities in the prefrontal cortex areas. The speech samples were acoustically analyzed for frequency, amplitude, and rate measures and the fNIRS data were analyzed for oxygen and deoxygen levels. We found several acoustic features that significantly differed between the DAT and neurologically healthy controls (NHCs) across different conditions (i.e., Happy, Sad, and Neutral) as well as a general pattern of reduced hemodynamic responses in the DAT group, with the most pronounced and consistent differences emerging in the right hemisphere, particularly under the Happy and Sad conditions. Many of the group differences were not only statistically significant but also had moderate to large effect sizes, highlighting the potential of this dual-modal analysis as behavioral biomarker testing for cognitive decline.