Micro-FTIR hyperspectral imaging classification for oral cavity histopathology analysis

Abstract

Hyperspectral imaging (HSI) has emerged as a promising tool for integrating spatial and biochemical information in computational pathology analysis. While most studies on oral cavity cancer have employed reflectance-based HSI or Raman spectroscopy in the visible and near-infrared ranges, the potential of mid-infrared micro–Fourier transform infrared (micro-FTIR) spectroscopy remains largely unexplored. This study investigates the feasibility of using micro-FTIR hyperspectral data for the classification of oral cavity tissues. Mid-IR spectra provide detailed biochemical information, including protein, lipid, and nucleic acid signatures, which may be clinically relevant for early diagnosis and characterization of tumor margins. Tissue samples were imaged using micro-FTIR spectroscopy, and voxel-level spectra were preprocessed and classified using a fully-connected neural network. The proposed model achieved an accuracy of 88.41%, sensitivity of 87.64%, and area under the receiver operating characteristic curve (AUC) of 96.51%, demonstrating that micro-FTIR–based HSI can successfully differentiate between healthy and malignant oral cavity tissues. These findings provide the first systematic evidence supporting the clinical potential of mid-infrared spectroscopy in oral oncology.