Improving Performance in fNIRS Single Trial Analysis: Multidisciplinary Opportunities and Perspective

Advancements in wearable technologies and signal analysis are bringing functional Near-Infrared Spectroscopy (fNIRS) to the forefront of mobile non-invasive brain-computer interface research. As it gains main-stream attention, Diffuse Optical Tomography (DOT), a high-density fNIRS variant, shows great promise by enhancing spatial resolution and brain-imaging contrast while maintaining the ease of use and usability of optical brain imaging techniques. However, to fully unlock the potential of mobile fNIRS and DOT, persisting challenges in extracting meaningful task-evoked hemodynamic signals amidst systemic physiological noise must be overcome, particularly for single-trial analyses. We briefly review the recent advances in wearable fNIRS/DOT instrumentation and highlight multidisciplinary opportunities to improve single trial decoding performance by combining advances in wearable DOT instrumentation with model-driven best practices from the fNIRS neuroscience community and data-driven innovations in multimodal machine learning. Finally, we introduce Cedalion, our recently launched open-source Python toolbox for state-of-the-art fNIRS/DOT analysis and multimodal machine learning.