Infrared spectroscopy of exfoliated cervical cell specimens. Proceed with caution

Area of Science:

• Biomedical spectroscopy
• Cervical cancer diagnostics
• Infrared spectroscopy

Background:

• Cervical cancer screening relies on cytology and histology.
• Exfoliated cervical cell analysis offers a potential non-invasive diagnostic avenue.
• Infrared spectroscopy provides molecular fingerprints of biological samples.

Purpose of the Study:

• To assess the diagnostic potential of infrared spectra from exfoliated cervical cells.
• To evaluate a novel spectral feature extraction method combined with discriminant analysis for classifying cervical cell spectra.
• To determine if spectral data can differentiate normal cells from those with cervical intraepithelial neoplasia (CIN).

Main Methods:

• Infrared spectra from over 800 exfoliated cervical cell specimens were analyzed.
• Clinical diagnoses were based on cytology and histology reports.
• Classification models were developed using a training set and tested on a separate set to distinguish normal from CIN 1, 2, and 3 spectra.

Main Results:

• Classification accuracies for distinguishing normal from CIN varied.
• Training set accuracies ranged from 68% (normal/CIN 1) to 81% (normal/CIN 3).
• Test set accuracies were 60% (normal/CIN 1), 60% (normal/CIN 2), and 67% (normal/CIN 3).

Conclusions:

• Infrared spectra of exfoliated cervical cells contain information about dysplasia.
• Diagnostic information regarding the presence or absence of dysplasia is recoverable from cell spectra.
• While imperfect, the method shows promise for analyzing real-life cervical cell preparations.