Mid-Infrared Imaging Characterization to Differentiate Lung Cancer Subtypes

Area of Science:

• Biomedical Optics
• Spectroscopy
• Medical Diagnostics

Background:

• Lung cancer is a leading global malignancy with distinct subtypes (squamous cell carcinoma, adenocarcinoma, small cell carcinoma) requiring accurate differential diagnosis for effective treatment.
• Vibrational spectroscopy, particularly mid-infrared (MIR) based techniques, offers a non-destructive approach for investigating biological tissues and has seen significant advancements.
• Label-free microscopic techniques are increasingly important in life sciences for analyzing cellular and tissue characteristics without exogenous labeling.

Purpose of the Study:

• To investigate the potential of label-free mid-infrared spectra for differentiating major lung cancer histological subtypes.
• To apply supervised multivariate analysis methods to spectral data for accurate classification of lung cancer subtypes.
• To evaluate the efficacy of different analytical approaches (patient-based vs. pixel-based) and algorithms in spectral analysis.

Main Methods:

• Formalin-fixed paraffin-embedded (FFPE) tissue samples of squamous cell carcinoma (SQ), adenocarcinoma (LUAD), and small cell carcinoma (SCLC) were analyzed.
• Mid-infrared (MIR) spectra were acquired from 2 μm thick sections using a transflection optical setup on an infrared microscope.
• Supervised multivariate analyses, including Linear Discriminant Analysis (LDA) and Support Vector Machine (SVM) methods, were applied to the mid-infrared fingerprint region (1800-648 cm⁻¹).

Main Results:

• Patient-based analysis using LDA and SVM models demonstrated varying degrees of separation between lung cancer subtypes.
• The linear C-support vector classification (C-SVC) SVM model achieved 100% accuracy for differentiating the three subtypes when using a 50% cut-off value.
• Pixel-based analysis also identified the linear C-SVC SVM as the most efficient, yielding high sensitivity for SQ (81.65%), LUAD (82.89%), and SCLC (88.89%).

Conclusions:

• Mid-infrared imaging combined with supervised multivariate analysis is a promising, label-free method for differentiating FFPE lung cancer subtypes.
• The choice of spectral cut-off, kernel function, and algorithm significantly influences classification accuracy.
• This approach holds potential for developing advanced spectroscopic diagnostic tools to revolutionize cancer identification and differential diagnostics.