RamPINN: Recovering Raman Spectra From Coherent Anti-Stokes Spectra Using Embedded Physics

2025-10-07

AI Summary

In Coherent Anti-Stokes Raman Scattering (CARS) spectroscopy, the weak resonant Raman signal is overwhelmed by a strong non-resonant background, leading to ill-posed spectral recovery and absence of ground-truth labels. Method: We propose a physics-informed dual-decoder neural network: a shared encoder jointly models resonant and non-resonant components, while two dedicated decoders reconstruct them separately; a differentiable Hilbert transform loss enforces Kramers–Kronig causality, and a smoothness prior on the non-resonant component enables fully self-supervised training. Contribution/Results: The method requires no ground-truth Raman spectra, supports zero-shot transfer, and achieves state-of-the-art performance on both synthetic and experimental CARS data—demonstrating superior generalization, robustness, and physical consistency with underlying optical dispersion relations.

Technology Category

• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization
• Computer Vision: Low Level & Physics-based Vision

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Social Networks and Social Media: Computational social science

Abstract

Transferring the recent advancements in deep learning into scientific disciplines is hindered by the lack of the required large-scale datasets for training. We argue that in these knowledge-rich domains, the established body of scientific theory provides reliable inductive biases in the form of governing physical laws. We address the ill-posed inverse problem of recovering Raman spectra from noisy Coherent Anti-Stokes Raman Scattering (CARS) measurements, as the true Raman signal here is suppressed by a dominating non-resonant background. We propose RamPINN, a model that learns to recover Raman spectra from given CARS spectra. Our core methodological contribution is a physics-informed neural network that utilizes a dual-decoder architecture to disentangle resonant and non-resonant signals. This is done by enforcing the Kramers-Kronig causality relations via a differentiable Hilbert transform loss on the resonant and a smoothness prior on the non-resonant part of the signal. Trained entirely on synthetic data, RamPINN demonstrates strong zero-shot generalization to real-world experimental data, explicitly closing this gap and significantly outperforming existing baselines. Furthermore, we show that training with these physics-based losses alone, without access to any ground-truth Raman spectra, still yields competitive results. This work highlights a broader concept: formal scientific rules can act as a potent inductive bias, enabling robust, self-supervised learning in data-limited scientific domains.

Problem

• Recovering Raman spectra from noisy CARS measurements
• Disentangling resonant and non-resonant signals using physics
• Enabling robust learning without ground-truth Raman spectra data

Innovation

• Physics-informed neural network with dual-decoder architecture
• Enforces Kramers-Kronig relations via differentiable Hilbert transform
• Uses synthetic training data with zero-shot generalization capability