Lung cancer diagnosis through extracellular vesicle analysis using label-free surface-enhanced Raman spectroscopy...

Area of Science:

• Biomedical Engineering
• Analytical Chemistry
• Computational Biology

Background:

• Label-free surface-enhanced Raman spectroscopy (SERS) shows promise for extracellular vesicle (EV)-based cancer diagnosis.
• Challenges remain in SERS signal repeatability, stability, and accurate early prediction of multiple cell types from limited samples.

Purpose of the Study:

• To develop a highly accurate classification approach for distinguishing EVs from lung cancer and normal cells using SERS and machine learning.
• To validate the method with mixed cell-derived and plasma-derived EVs from mouse models and human patients.

Main Methods:

• Integrated label-free SERS analysis of EVs with machine learning (SVM, CNN).
• Employed a capillary-based liquid-phase sampling method to preserve the native state of EVs.
• Optimized SERS substrate properties and used Bayesian optimization for SVM hyperparameter tuning.

Main Results:

• Achieved a 3.7% classification error rate and 98.7% accuracy for distinguishing cell-derived EVs (SVM).
• Demonstrated high accuracy (97.5% for SVM, 95.8% for CNN) for plasma-derived EVs from lung cancer mice.
• Showed significant accuracy (91.5% for SVM, 95.4% for CNN) in discriminating plasma-derived EVs from lung cancer patients and healthy individuals.

Conclusions:

• The machine learning-assisted, liquid-phase SERS method offers minimal sample volume, high stability, and excellent accuracy.
• This approach shows potential as a rapid and reliable tool for early lung cancer detection and monitoring via blood analysis.