Uncertainty principle-guided intra-volume motion correction in full-field swept-source OCT for phase-stable retinal imaging.

Full-field swept-source optical coherence tomography (FF-SS-OCT) enables high-throughput volumetric mapping of cellular dynamics in the living human eye. However, intra-volume axial motion caused by eye movements and system drifts during relatively slow wavelength sweeps induces dynamic, wavenumber-dependent phase errors that severely limit phase stability. Here, we show that image quality and phase stability in FF-SS-OCT are disproportionately compromised by high-frequency components (100-200 Hz) of axial motion, despite their relatively small displacement amplitudes. Accordingly, we propose a multi-scale, multi-reference motion correction method. This method estimates and compensates for intra-volume phase errors across multiple spectral (and thus temporal) scales guided by the Heisenberg-Gabor limit. In addition, a multi-reference scheme effectively mitigates speckle noise inherent to single-band correlations. We validate the proposed method through phantom experiments and in vivo human retinal imaging, where it improves retinal image reconstruction and phase stability. Notably, in controlled phantom experiments emulating the power spectrum of human retinal movement, this method significantly suppresses motion artifacts and achieves phase sensitivity approaching the fundamental limit set by the signal-to-noise ratio. Such phase stabilization capability is critical for label-free optical imaging of neural activity in optoretinography and in vivo dynamic OCT using FF-SS-OCT.