Improved estimation of hemodynamic response in fNIRS using protocol constraint and wavelet transform decomposition based adaptive algorithm

Background Near infrared spectroscopy allows monitoring of oxy and deoxyhemoglobin concentration changes associated with hemodynamic response function (HRF). HRF is mainly affected by physiological interferences which occur in the superficial layers of the head. This makes HRF extracting a very challenging task. Recent studies have used an additional near channel which is sensitive to the systemic interferences of the superficial layers. This additional information can be used to remove the systemic interference from the HRF. New Method This paper presents a novel wavelet-based constrained adaptive procedure to define the proportion of the physiological interferences in the brain hemodynamic response. The proposed method decomposes the near channel signal into several wavelet transform (WT) scales and adaptively estimates proper weights of each scale to extract their share in the HRF. The estimation of the weights are done by applying data acquisition protocol as a coefficient on recursive least square (RLS), normalized least mean square (NLMS) and Kalman filter methods. Results Performance of the proposed algorithm is evaluated in terms of the mean square error (MSE) and Pearson’s correlation coefficient (R 2 ) criteria between the estimated and the simulated HRF. Comparison with Existing Methods Results: showed that using the proposed method is significantly superior to all past adaptive filters such as EMD/EEMD based RLS/NLMS on estimating HRF signals. Conclusions we recommend the use of WT based constraint Kalman filter in dual channel fNIRS studies with a defined protocol paradigm and using WT based Kalman filter in studies without any pre-defined protocol.