Data-efficient extraction of optical properties from 3D Monte Carlo TPSFs using Bi-LSTM transfer learning
For researchers in time-resolved spectroscopy needing real-time optical property extraction, this method offers a data-efficient solution that bridges the gap between fast analytical models and accurate but slow Monte Carlo simulations.
This work tackles the computationally expensive extraction of optical properties from 3D Monte Carlo time-resolved spectroscopy data. Using a Bi-LSTM with physics-informed transfer learning, they achieve near-instantaneous inference with competitive error, eliminating systematic bias of analytical models.
Time-Resolved Spectroscopy (TRS) is a powerful modality for non-invasive characterization of turbid media. However, extracting optical properties, absorption $μ_a$ and reduced scattering $μ_s'$, from 3D stochastic measurements remains computationally expensive for real-time applications. In this paper, we propose a data-efficient, physics-informed transfer learning strategy using a Bidirectional Long Short-Term Memory (Bi-LSTM) network. By leveraging a fast deterministic solver to establish a physical prior before fine-tuning on a restricted set of 3D Monte Carlo simulations, our model successfully bridges the analytical-to-stochastic domain gap. The proposed method eliminates the systematic bias of analytical models while maintaining a competitive error with near-instantaneous inference time.