Raman spectral post-processing for oral tissue discrimination - a step for an automatized diagnostic system

Area of Science:

• Biomedical Optics
• Medical Diagnostics
• Spectroscopy

Background:

• Oral lesion diagnosis typically relies on invasive histopathology, which is time-consuming.
• Raman spectroscopy presents a real-time, minimally invasive alternative for disease diagnosis.
• Data post-processing can enhance the diagnostic capabilities of Raman spectroscopy.

Purpose of the Study:

• To evaluate preprocessing techniques and multivariate analysis for classifying oral tissue spectra.
• To compare the performance of different classifiers for distinguishing normal and pathological oral tissues.

Main Methods:

• Optical fiber Raman-based spectroscopy (OFRS) was used to acquire 80 spectra from normal and abnormal oral tissues.
• Spectra underwent area or maximum intensity normalization and Principal Component Analysis (PCA) preprocessing.
• Classifiers including K-nearest neighbors (KNN), J48, Radial Basis Function (RBF), Random Forest (RF), and Multilayer Perceptron (MLP) were applied using WEKA software.

Main Results:

• Maximum intensity normalization combined with the Multilayer Perceptron (MLP) classifier yielded the highest classification accuracy.
• The study identified optimal data processing strategies for Raman spectroscopic analysis of oral tissues.

Conclusions:

• Maximum intensity normalization and MLP offer a promising approach for automated oral lesion diagnosis using Raman spectroscopy.
• Further validation with larger datasets could lead to clinical software for rapid spectroscopic data interpretation and diagnosis.