Phase-Restoring Subpixel Image Registration: Enhancing Motion Detection Performance in Fourier-domain Optical Coherence Tomography

Phase-sensitive Fourier-domain optical coherence tomography (FD-OCT) enables in-vivo, label- free imaging of cellular movements with detection sensitivity down to the nanometer scale, and it is widely employed in emerging functional imaging modalities, such as optoretinography (ORG), Doppler OCT, and optical coherence elastography. However, when imaging tissue dynamics in vivo, tissue movement or bulk motion introduces decorrelation noise that compromises motion detection performance, particularly in terms of sensitivity and accuracy. Here, we demonstrate that the motion-related decorrelation noise in FD-OCT can be accurately corrected by restoring the initial sampling points using our proposed Phase-Restoring Subpixel Image Registration (PRESIR) method. Derived from a general FD-OCT model, the PRESIR method enables translational shifting of complex-valued OCT images over arbitrary displacements with subpixel precision, while accurately restoring phase components. Unlike conventional approaches that shift OCT images either in the spatial domain at the pixel level or in the spatial frequency domain for subpixel correction, our method reconstructs OCT images by correcting axial displacement in the spectral domain (k domain) and lateral displacement in the spatial frequency domain. We validated the PRESIR method through simulations, phantom experiments, and in-vivo optoretinography in both rodents and human subjects. Our approach significantly reduced decorrelation noise during the imaging of moving samples, achieving phase sensitivity close to the fundamental limit determined by the signal-to-noise ratio (SNR).