Machine Learning-Assisted Sampling of Surfance-Enhanced Raman Scattering (SERS) Substrates Improve Data Collection...

Area of Science:

• Analytical Chemistry
• Spectroscopy
• Biotechnology

Background:

• Surface-enhanced Raman scattering (SERS) offers sensitive, label-free analysis for chemical and biological samples.
• Current SERS automation focuses on data processing, not data acquisition, which remains manual and requires expert judgment.
• This bottleneck limits SERS application in point-of-care settings lacking trained spectroscopists.

Purpose of the Study:

• To develop and evaluate a machine learning-assisted method for improving SERS data acquisition.
• To address the need for automated spectral quality assessment during SERS measurements.
• To facilitate the development of automated SERS diagnostic platforms for clinical use.

Main Methods:

• Introduced a machine learning-assisted approach integrated into the SERS data acquisition stage.
• Evaluated six common algorithms for their performance in judging spectral quality.
• Developed an open-source Python package for rapid expert user annotation to train machine learning models.

Main Results:

• Demonstrated the feasibility of using machine learning to assist in SERS spectral quality judgment during data collection.
• Identified optimal algorithms for spectral quality assessment in the SERS acquisition context.
• Created a tool to enable efficient training of machine learning models for SERS data acquisition.

Conclusions:

• Machine learning can significantly enhance the data acquisition process in SERS experiments.
• This approach can overcome limitations in manual spectral quality assessment, crucial for automation.
• The developed methods and tools are foundational for future point-of-care SERS diagnostic platforms.