RLS adaptive filtering for physiological interference reduction in NIRS brain activity measurement: a Monte Carlo study.

The non-invasive measurement of cerebral functional haemodynamics using near-infrared spectroscopy (NIRS) instruments is often affected by physiological interference. The suppression of this interference is crucial for reliable recovery of brain activity measurements because it can significantly affect the signal quality. In this study, we present a recursive least-squares (RLS) algorithm for adaptive filtering to reduce the magnitude of the physiological interference component. To evaluate it, we implemented Monte Carlo simulations based on a five-layer slab model of a human adult head with a multidistance source-detector arrangement, of a short pair and a long pair, for NIRS measurement. We derived measurements by adopting different interoptode distances, which is relevant to the process of optimizing the NIRS probe configuration. Both RLS and least mean squares (LMS) algorithms were used to attempt the removal of physiological interference. The results suggest that the RLS algorithm is more capable of minimizing the effect of physiological interference due to its advantages of faster convergence and smaller mean squared error (MSE). The influence of superficial layer thickness on the performance of the RLS algorithm was also investigated. We found that the near-detector position is an important variable in minimizing the MSE and a short source-detector separation less than 9 mm is robust to superficial layer thickness variation.