Histopathology provides the gold standard assessment of colonoscopic biopsies. Infrared spectroscopy can potentially map biochemical changes across a tissue section identifying disease. The purpose of this study was to determine if infrared spectroscopy could classify different colorectal pathologies and to investigate biochemical composition. Colonoscopic tissue biopsies were snap frozen at colonoscopy. 10 micron thick sections were mounted on CaF2 slides. 3- D spectral datasets (2 spatial dimensions and one spectral) were measured from thawed specimens using a Perkin Elmer infrared imaging system in transmission mode. Contiguous tissue sections stained with H&E were reviewed by a specialist gastrointestinal pathologist for comparison. Tissue spectra from epithelial tissues were classified using principal components fed linear discriminant analysis with leave one out cross validation. Reference spectra from purchased biochemicals (Sigma-Aldrich) were measured. Ordinary least squares analysis estimated the relative biochemical signal contribution from epithelial regions. Spectra from tissue epithelia measured from normal tissue, hyperplastic polyps, adenomatous polyps, cancer and ulcerative colitis samples were classified with accuracies in excess of 90%. Ordinary least squares analysis demonstrated a higher DNA to cytoplasm ratio in cancer compared to normal tissue. FTIR spectra from epithelia can be used to classify colorectal pathologies with high accuracy. Ordinary least squares analysis shows promise for extraction of useful biochemical information. These techniques could aid the histopathologist and ultimately lead to automated histopathological processing.
展开▼
