Feed

Healthy aging is accompanied by widespread changes in cortical structure and function, particularly within networks supporting multisensory integration and cognitive control. However, it remains unclear whether age-related neural alterations can be reliably decoded from individual brain activity patterns, especially under different cognitive states. This study examined whether age-related neural alterations can be decoded from time-domain functional near-infrared spectroscopy (fNIRS) signals during resting and complex audiovisual processing. fNIRS data were collected from cognitively normal younger and older adults across temporal, frontal, and parietal cortices. Nine time-domain features of oxygenated hemoglobin (HbO) were extracted and analyzed using supervised machine learning (SVM). Decoding accuracy was low at rest (peak = 0.651 with kurtosis alone) and remained poor across different classifiers. In contrast, during the audiovisual dual-task, a combination of five features-variance, peak, time-to-peak, slope, and skewness-achieved high classification accuracy (0.810), revealing robust age-related neural signatures. Spatial analysis showed that discriminative optodes were bilaterally distributed but functionally asymmetric: left temporoparietal regions were primarily involved in auditory processing and multisensory integration, whereas right frontoparietal regions supported attentional control and top-down regulation. These findings demonstrate that age-related neural dynamics are best captured under high cognitive load, highlighting time-domain fNIRS features as sensitive markers of cortical reorganization and potential indicators of healthy brain aging.