Fluorescence spectral imaging for characterization of tissue based on multivariate statistical analysis.

Area of Science:

• Biomedical Optics
• Medical Imaging
• Multivariate Statistical Analysis

Background:

• Autofluorescence spectroscopy provides valuable biochemical information about tissues.
• Classifying spectral data for diagnostic purposes requires robust analytical methods.
• Principal Component Analysis (PCA) is effective for reducing spectral data dimensionality.

Purpose of the Study:

• To develop and investigate a novel spectral imaging method for classifying light-induced autofluorescence spectra.
• To utilize PCA for generating diagnostic images from autofluorescence signals.
• To assess the method's accuracy in differentiating cancerous from normal nasopharyngeal tissue.

Main Methods:

• A spectral imaging system employing optical spectral filters linked to diagnostically relevant principal components was designed.
• Autofluorescence spectral data from in vivo nasopharyngeal tissue were collected.
• PCA was applied to process autofluorescence signals, generating principal component score images and diagnostic algorithms.

Main Results:

• The proposed method successfully differentiated carcinoma lesions from normal tissue with 95% sensitivity and 93% specificity in simulations.
• Optimal principal filters and PCA-based algorithms were investigated to enhance diagnostic accuracy.
• The method demonstrated robustness against noise in autofluorescence signals.

Conclusions:

• The novel PCA-based spectral imaging method offers a promising approach for non-invasive tissue diagnosis.
• The system's high sensitivity and specificity suggest potential for early cancer detection.
• Further optimization of filters and algorithms can improve diagnostic performance.