Label-Free Differentiation of Cancer and Non-Cancer Cells Based on Machine-Learning-Algorithm-Assisted Fast Raman...

Area of Science:

• Biomedical Optics
• Spectroscopy
• Machine Learning in Biology

Background:

• Accurate identification of cancer cells is crucial for diagnosis and research.
• Existing methods for cell identification can be invasive, time-consuming, or require labels.
• Raman spectroscopy offers label-free molecular fingerprinting capabilities.

Purpose of the Study:

• To develop a rapid, label-free, and non-invasive method for distinguishing murine cancer cells (B16F10 melanoma) from non-cancer cells (C2C12 muscle cells).
• To leverage machine learning-assisted Raman spectroscopic imaging for enhanced cell identification.
• To demonstrate the retrieval of essential biomolecular information from hyperspectral Raman data for cell differentiation.

Main Methods:

• Utilized rapid Raman spectroscopic imaging to acquire hyperspectral data from cell samples.
• Applied a machine learning-based hyperspectral data processing approach for analysis.
• Focused on extracting biomolecular information (nucleic acids, proteins, lipids) from the spectral data.

Main Results:

• Successfully distinguished B16F10 melanoma cancer cells from C2C12 muscle cells using the developed approach.
• Demonstrated the capability of machine learning to analyze cell structure from Raman data without compromising spectral information.
• Showcased efficient retrieval of biomolecular information from low-quality hyperspectral Raman datasets.

Conclusions:

• The proposed machine learning-assisted Raman spectroscopic imaging is a viable rapid, label-free, and non-invasive technique for cancer cell identification.
• Biomolecular information extracted from Raman spectra can be effectively used for differentiating cell lines.
• This method holds potential for advancing cell analysis in biological and medical research.