Micromolar chemical imaging by high-energy low-photodamage Coherent Anti-stokes Raman Scattering (HELP-CARS)

Coherent anti-Stokes Raman scattering (CARS) microscopy offers label-free chemical imaging capabilities, but its performance is constrained by small Raman scattering cross-section, strong non-resonant background (NRB), and limited signal-to-noise ratio (SNR). Here, we introduce a high-energy, low-photodamage CARS (HELP-CARS) platform designed to overcome these physical limitations. By employing a 1-MHz non-collinear optical parametric amplifier (NOPA) with extensive pulse chirping, HELP-CARS increases the coherent Raman excitation efficiency by ~300 times and improves the signal-to-nonresonant background ratio by 11 times, while inducing negligible damage during live cell imaging. Furthermore, to remove non-independent noise and physically entangled non-resonant background, we incorporate self-supervised deep-learning denoising and background removal based on the Kramers-Kronig relationship, yielding sensitivity improvement by an additional order of magnitude. Together, these advances provide a micromolar imaging sensitivity (160 uM for Dimethyl sulfoxide-d6) corresponding to 1000 molecules in the focal volume. Such high sensitivity enables high-fidelity chemical imaging in both fingerprint and silent windows. Hyperspectral HELP-CARS imaging of deuterated fatty acids allowed first observation of chemical separation with single lipid droplet. Together, HELP-CARS offers a powerful and generalizable approach for ultrasensitive and quantitative vibrational imaging of biological systems.

💡 Research Summary

Coherent anti‑Stokes Raman scattering (CARS) microscopy provides label‑free chemical contrast but suffers from intrinsically weak Raman cross‑sections, a strong non‑resonant background (NRB), and limited signal‑to‑noise ratio (SNR). In this work the authors introduce HELP‑CARS (High‑Energy Low‑Photodamage CARS), a platform that simultaneously tackles all three limitations. The core of the system is a 1 MHz non‑collinear optical parametric amplifier (NOPA) that delivers synchronized pump (650–900 nm) and Stokes (fixed at 1045 nm) beams. By stretching the femtosecond pump to ~30 ps and the Stokes to ~5 ps, the temporal overlap is optimized, suppressing NRB by a factor of 11 relative to a conventional 80 MHz OPO‑based CARS system. Because the NOPA operates at a much lower repetition rate, the peak power is increased ~80‑fold for the same average power, which translates into a ~300‑fold boost in coherent Raman excitation efficiency.

The authors quantify the performance gains both theoretically and experimentally. Calculations predict a 172‑fold increase in photon yield for NOPA‑CARS over OPO‑SRS; measurements show a ~300‑fold increase, confirming the model. However, the lower repetition rate also introduces pulse‑to‑pulse intensity fluctuations, producing non‑independent, non‑identically distributed (non‑i.i.d.) noise that limits the practical SNR. To overcome this, the team develops SPEND (Self‑Permutation Noise2Noise Denoiser), a self‑supervised deep‑learning framework that learns and removes structured spatial and spectral noise without ground‑truth data. Combined with Kramers‑Kronig (KK) background retrieval, SPEND improves the effective SNR by an additional order of magnitude.

With hardware and algorithmic enhancements together, HELP‑CARS achieves a limit of detection (LoD) of 160 µM for dimethyl‑sulfoxide‑d₆, corresponding to roughly 1 000 molecules in the focal volume—an improvement of ~30× over OPO‑SRS and ~7× fewer molecules than OPO‑CARS. The system also benefits from the inherently higher spatial and axial resolution of CARS, halving the excitation volume compared with SRS and further reducing the number of molecules required for detection.

Biological relevance is demonstrated through live‑cell imaging. Using pump and Stokes powers of 70 mW and 50 mW respectively, cells imaged at the C‑H stretch (2850 cm⁻¹) for six minutes show no measurable photodamage: average intensity, cell motility, and viability remain unchanged. Moreover, hyperspectral imaging of deuterated fatty acids reveals chemical segregation within a single lipid droplet, providing the first direct observation of intra‑droplet compositional heterogeneity using CARS.

In summary, HELP‑CARS integrates a high‑energy low‑repetition‑rate laser source, extensive pulse chirping, and advanced self‑supervised deep‑learning denoising to deliver micromolar‑level, quantitative vibrational imaging with negligible photodamage. The approach offers a generalizable pathway to ultrasensitive, high‑resolution, label‑free chemical imaging across both fingerprint and silent Raman windows, positioning HELP‑CARS as a powerful tool for a wide range of biomedical and materials‑science applications.