Assessing the consistency and sensitivity of the neural correlates of narrative stimuli using functional near-infrared spectroscopy

• Published in: Imaging neuroscience (Cambridge, Mass.) Vol. 2
• Main Authors: Kolisnyk, Matthew, Novi, Sergio, Abdalmalak, Androu, Ardakani, Reza Moulavi, Kazazian, Karnig, Laforge, Geoffrey, Debicki, Derek B., Owen, Adrian M.
• Format: Journal Article
• Language: English
• Published: 255 Main Street, 9th Floor, Cambridge, Massachusetts 02142, USA MIT Press 24.10.2024
• Subjects: functional near-infrared spectroscopy; higher-order cognition; inter-participant reproducibility; inter-subject correlation; machine learning; movie listening; movie watching; functional near-infrared spectroscopy; higher-order cognition; movie listening; inter-subject correlation; movie watching; machine learning; inter-participant reproducibility
• ISSN: 2837-6056; 2837-6056
• DOI: 10.1162/imag_a_00331

Investigating how the brain responds to rich and complex narratives, such as engaging movies, has helped researchers study higher-order cognition in “real-world” scenarios. These neural correlates are particularly useful in populations where behavioral evidence of cognition alone is inadequate, such as children and certain patient populations. While this research has been primarily conducted in fMRI and EEG, whether functional near-infrared spectroscopy (fNIRS) can reliably detect these neural correlates at an individual level, which is required for effective use in these populations, has yet to be established. This study replicated widespread inter-subject correlations (ISCs) in the frontal, parietal, and temporal cortices in fNIRS in healthy participants when they watched part of the TV episode and listened to an audio clip from the movie Conversely, these ISCs were primarily restricted to temporal cortices when participants viewed scrambled versions of those clips. To assess whether these results were reliable at the single-participant level, two follow-up analyses were conducted. First, the consistency analysis compared each participant’s ISCs against group results that excluded that individual. This approach found that 24 out of 26 participants in and 20/26 participants in were statistically similar to the group. Second, the sensitivity analysis measured whether machine-learning algorithms could decode between intact conditions and their scrambled counterparts. This approach yielded balanced accuracy scores of 81% in and 79% in . Overall, the neural correlates of narrative stimuli, as assessed by fNIRS, are reproducible across participants, supporting its broad application to clinical and developmental populations.