Maximum likelihood estimation of blood velocity using Doppler optical coherence tomography

Abstract

A recent trend in optical coherence tomography (OCT) hardware has been the move towards higher A-scan rates. However, the estimation of axial blood flow velocities is affected by the presence and type of noise, as well as the estimation method. Higher acquisition rates alone do not enable the accurate quantification of axial blood velocity. Moreover, decorrelation is an unavoidable feature of OCT signals when there is motion relative to the OCT beam. For in-vivo OCT measurements of blood flow, decorrelation noise affects Doppler frequency estimation by broadening the signal spectrum. Here we derive a maximum likelihood estimator (MLE) for Doppler frequency estimation that takes into account spectral broadening due to decorrelation. We compare this estimator with existing techniques. Both theory and experiment show that this estimator is effective, and outperforms the Kasai and additive white Gaussian noise (AWGN) ML estimators. We find that maximum likelihood estimation can be useful for estimating Doppler shifts for slow axial flow and near transverse flow. Due to the inherent linear relationship between decorrelation and Doppler shift of scatterers moving relative to an OCT beam, decorrelation itself may be a measure of flow speed.