Image-driven OCT calibration from a single B-scan

Optical coherence tomography (OCT) requires accurate calibration of wavenumber sampling and dispersion compensation to achieve optimal axial resolution. Conventional calibration methods rely on dedicated mirror measurements acquired at multiple optical path differences, requiring dedicated acquisitions, careful alignment, and additional acquisition time. We present a physics-informed optimisation framework that recovers OCT calibration directly from a single routine B-scan. Within the complex leader-follower reconstruction framework, the method jointly estimates wavenumber remapping and dispersion compensation functions by optimising image quality in reconstructed OCT data. Experiments on a swept-source OCT system demonstrate that the recovered calibration functions produce image quality comparable to conventional mirror-based calibration while requiring only a single acquired image and 30-60 seconds of optimisation. Samples containing depth-distributed structure recover calibration functions closely matching the underlying system calibration, whereas structurally limited samples produce image-specific effective calibration within the observed depth range. These results demonstrate the feasibility of rapid software-defined OCT calibration directly from imaging data, reducing dependence on dedicated calibration measurements, and simplifying OCT system operation.