Development and Validation of a Portable, Durable, Rugged Functional Near-Infrared Spectroscopy (fNIRS) Device

Event Abstract Back to Event Development and Validation of a Portable, Durable, Rugged Functional Near-Infrared Spectroscopy (fNIRS) Device Bethany Bracken1*, Elena Festa2, Hsin-Mei Sun2, Calvin Leather1, Gary Strangman3, Noa Palmon1, Filipe Silva4, Manuel Pacheco4 and Blaise Frederick5 1 Charles River Analytics (United States), United States 2 Brown University, United States 3 Massachusetts General Hospital, Harvard Medical School, United States 4 Biosignals Plux, Portugal 5 McLean Hospital, United States BACKGROUND Assessing cognitive workload using functional near-infrared spectroscopy (fNIRS) in labs is well established. Increased workload corresponds with increase in prefrontal blood oxygenation (HbO2) correlated with increased task engagement. Once the task becomes too difficult, HbO2 decreases as does task engagement and performance (Ayaz, et al., 2012). However, fNIRS sensors useful for assessing cognitive workload during normal activities in real-world environments are only recently emerging. Standard sensors are large (e.g., full-head), expensive (~$10K) and require heavy equipment (e.g., batteries, laptops). Under an Army-funded effort (MEDIC), we designed a portable, cost-effective fNIRS sensor. Under a NASA-funded effort (CAPT PICARD), we validated our sensor against the NINScan developed at Massachusetts General Hospital. We used a gold-standard task known to affect cognitive workload (n-back; (Kirchner, 1958)) and a more complex multi-attribute task battery (MATB) (Santiago-Espada et al., 2011). NINScan supports 32 channels, with 8 channels per hemisphere in this test. Because our MEDIC fNIRS sensor only includes one source-detector pair, we further validated our findings by collecting data at two locations: the dorsolateral prefrontal cortex (dlPFC) known to exhibit changes in HbO with increasing cognitive workload, and the medial PFC, which does not exhibit changes in HbO due to cognitive workload. METHOD We recruited 23 healthy adult (21.3 ± 3.0 years; 10 males) students at Brown University. Three withdrew prior to completion. To minimize learning effects, participants completed one practice session. Within the following two sessions, participants wore one of the sensors (NINScan or MEDIC in counterbalanced order) while participants performed the task battery twice. We collected two minutes of baseline rest (eyes-closed) at the start and end of each session, and used the NASA-TLX questionnaire to assess subjective workload. RESULTS N-back performance decreased with increasing difficulty level (increased response time (RT) and decreased accuracy). For the MEDIC sensor, we ran a mixed model with time dummy variables (one for each 10 second period to remove drift effect), a fixed effect of performance, and a by-subjects random effect. We found a significant effect of performance on dlPFC HbO2 (p