Self-optimized spectral distance for low-light high-throughput Raman hyperspectral imaging.

Area of Science:

• Spectroscopy
• Imaging Science
• Computational Science

Background:

• Raman hyperspectral imaging offers detailed sample analysis by combining vibrational spectroscopy and spatial imaging.
• Weak Raman scattering signals often require long acquisition times or high-power lasers, limiting practical applications.
• Existing methods may depend on extensive training datasets or high energy inputs.

Purpose of the Study:

• To develop a computational method for efficient Raman imaging under challenging conditions.
• To enable high-throughput Raman imaging by overcoming limitations of acquisition time and laser power.
• To reconstruct high-quality Raman images from low-quality measurements without large training datasets.

Main Methods:

• An unsupervised learning-based method, self-optimized spectral distance (SSD), was developed.
• SSD reconstructs Raman images directly from low-quality, 'noisy' measurements.
• The method eliminates the need for large-scale training datasets, long imaging times, and high-energy lasers.

Main Results:

• The SSD method successfully reconstructs Raman images from measurements acquired with short integration times or low-power lasers.
• High imaging quality was achieved across diverse applications, including cellular structure analysis, microparticle detection, and pharmaceutical ingredient identification.
• Acquisition time and excitation power were reduced by at least one order of magnitude.

Conclusions:

• The developed SSD method significantly enhances the efficiency and applicability of Raman hyperspectral imaging.
• This approach facilitates high-throughput Raman imaging, making the technique more accessible and versatile.
• SSD enables advanced imaging under challenging conditions, reducing resource requirements and broadening potential applications.