Sparse sampling for fast hyperspectral coherent anti-Stokes Raman scattering imaging

Area of Science:

• Spectroscopy
• Microscopy
• Biomedical Imaging

Background:

• Coherent anti-Stokes Raman scattering (CARS) hyperspectral imaging offers rich chemical information but suffers from slow acquisition speeds.
• Acquisition speed is a major bottleneck for in-situ and high-throughput applications of CARS microscopy.
• Existing methods often compromise spectral resolution or information content to achieve faster imaging.

Purpose of the Study:

• To develop and demonstrate a novel method for accelerating CARS hyperspectral imaging acquisition.
• To maintain essential spectral information and quantitative analysis capabilities despite reduced acquisition time.
• To enable faster imaging of biological samples and other applications of vibrational microscopy.

Main Methods:

• Determining key spectral components from a limited spatial region with high spectral resolution.
• Identifying a minimal set of frequencies required for accurate component weight retrieval.
• Acquiring hyperspectral images at the selected frequencies and reconstructing the full spectrum using a dedicated algorithm.
• Applying the method to CARS hyperspectral imaging of human osteosarcoma U2OS cells.

Main Results:

• Achieved a 25-fold reduction in acquisition time for CARS hyperspectral imaging.
• Successfully reconstructed full spectral range and resolution from reduced frequency measurements.
• Demonstrated the ability to retrieve CARS susceptibility, which is linear with concentration.
• Enabled unsupervised quantitative analysis of hyperspectral data.

Conclusions:

• The developed method significantly enhances CARS hyperspectral imaging speed without sacrificing spectral information.
• This approach is applicable to various coherent vibrational microscopy techniques and general hyperspectral imaging.
• The acceleration facilitates faster quantitative analysis and broader application of CARS microscopy in biological and material sciences.