In vivo diagnosis of cervical precancer using Raman spectroscopy and genetic algorithm techniques

Area of Science:

• Biomedical Optics
• Molecular Spectroscopy
• Gynecologic Oncology

Background:

• Cervical cancer diagnosis relies on colposcopy and biopsies, which can be invasive.
• Identifying molecular signatures of cervical dysplastic transformation is crucial for early detection.
• Novel spectroscopic techniques offer potential for non-invasive tissue analysis.

Purpose of the Study:

• To assess the clinical utility of near-infrared (NIR) Raman spectroscopy for identifying biomolecular changes in cervical tissues.
• To evaluate the effectiveness of genetic algorithm-partial least squares-discriminant analysis (GA-PLS-DA) in differentiating normal from precancerous cervical tissues.
• To explore the potential of Raman spectroscopy combined with advanced algorithms for improved colposcopic examination.

Main Methods:

• Collected 105 in vivo Raman spectra from 57 cervical sites (35 normal, 22 precancer) in 29 patients.
• Utilized GA-PLS-DA with double cross-validation (dCV) for spectral analysis and feature selection.
• Identified seven diagnostically significant Raman bands associated with proteins, nucleic acids, and lipids.

Main Results:

• The GA-PLS-DA model achieved an overall diagnostic accuracy of 82.9% for precancer detection.
• Sensitivity was 72.5% (29/40) and specificity was 89.2% (58/65) for identifying cervical precancerous lesions.
• Significant Raman bands were identified in specific spectral ranges linked to key biomolecules.

Conclusions:

• Raman spectroscopy, coupled with GA-PLS-DA and dCV, demonstrates potential for molecular-level discrimination between normal and precancerous cervical tissues.
• This approach may enhance the diagnostic capabilities during colposcopic examinations.
• Further research is warranted to validate these findings in larger clinical settings.