Inverse Algorithms for Depth-Resolved Spectroscopic OCT: A Comparative Study

We present a systematic comparison of inverse algorithms for extracting depth-resolved spectroscopic information from optical coherence tomography data. Six methods are evaluated—short-time Fourier transform, continuous wavelet transform, dual-window technique, maximum likelihood estimation, compressed sensing, and a novel physics-constrained neural network—on both simulated and experimental datasets. The physics-constrained neural network achieves superior spectral fidelity (RMSE ¡ 0.02) while maintaining computational efficiency suitable for real-time processing.