Stokes shift spectroscopy and machine learning for label-free human prostate cancer detection

Area of Science:

• Biomedical Optics
• Molecular Spectroscopy
• Cancer Diagnostics

Background:

• Prostate cancer diagnosis relies on invasive methods.
• Label-free spectroscopic techniques offer potential for non-invasive biomarker detection.
• Endogenous fluorophores like tryptophan, collagen, and NADH are crucial in cellular function and can be altered in cancer.

Purpose of the Study:

• To evaluate the efficacy of Stokes shift spectra (S3) for label-free prostate cancer detection.
• To identify key endogenous biomolecules (tryptophan, collagen, NADH) as potential cancer biomarkers using S3 spectroscopy.
• To apply machine learning algorithms for analyzing S3 spectral data and classifying tissue types.

Main Methods:

• Collected label-free S3 spectra from human cancerous and normal prostate tissues.
• Utilized machine learning algorithms: Principal Component Analysis (PCA), Non-negative Matrix Factorization (NMF), and Support Vector Machines (SVMs).
• Analyzed spectral components and evaluated SVM classification performance using sensitivity, specificity, and accuracy.

Main Results:

• Significant differences in component weights were observed between cancerous and normal prostate tissues.
• The S3 spectral analysis successfully identified distinct spectral signatures associated with cancer.
• SVM classification demonstrated high performance in differentiating between normal and cancerous tissues.

Conclusions:

• S3 spectroscopy is an effective label-free method for detecting prostate cancer.
• The approach accurately identifies changes in endogenous fluorophore concentrations indicative of cancer development.
• This technique holds promise for improved, non-invasive cancer diagnostics.