Abstract
Functional near-infrared spectroscopy (fNIRS) is non-invasive, lightweight, and portable, so it can be an ideal tool in the case of patients with less mobility. This review assesses the use of fNIRS-based brain-computer interfaces (BCIs) in the context of gait rehabilitation, with regard to their potential to decode motor intentions, and be used to drive lower limb exoskeletons. fNIRS-BCIs have demonstrated a high rate of classification accuracy when it comes to identifying the presence of a walking intention as opposed to resting states with a median range of 85-98 %. With machine learning algorithms like the k-nearest neighbor (kNN), gradient boosting decision trees (GBDT), and deep learning models, when combining it with other machine learning techniques, it increased its detection accuracy by as much as 15 % in comparison to traditional methods. In addition, there is even 10-20% increase in the detection of intent in unimodal systems with the use of hybrid Electroencephalography-functional near infrared spectroscopy (EEG-fNIRS) systems. Clinical trials have demonstrated the improvement in the gait speed and step symmetry (p < 0.05) between stroke and spinal cord injury patients who used fNIRS controlled exoskeletons. These outcomes indicate the possibilities of disposing Traumatic Brain Injury (TBI) related motor deficit using fNIRS-BCIs to facilitate mobility, neuroplasticity, and real-time adjustment rehabilitation interventions.