Vibrational spectroscopy of biological tissues

Abstract

The vibrational modes of molecules in biological tissues can be assessed by either Raman spectroscopy, through inelastic scattering, or infrared spectroscopy, through direct measurement of transmittance or reflectance. When combined with mathematical methods, vibrational spectroscopic techniques have shown promising results for evaluation of biochemical changes in biological samples, and such combination can be used to develop new tools for medical diagnosis. Here, we provide an overview of the infrared spectral imaging techniques we use to characterize biological tissues and describe how we employ these techniques to diagnose cancer and to evaluate inflammatory processes. In the last decade, we have studied thyroid and colon cancer tissues as well as inflammatory processes attributed to an early stage of cancer. All the samples were obtained from human biopsy embedded in paraffin and cut according to the usual procedures in pathology. The sample slides were deposited over a Calcium Fluoride window that is transparent in the infrared spectral region. An FTIR spectrometer with 4-cm-1 resolution coupled to a microscope with 6x6 microns of effective pixel size was employed. Pre-processing algorithms were necessary to remove unwanted absorption bands such as water vapor, carbon dioxide, and paraffin absorption bands. After that, the data of hyperspectral images were processed to classify and to predict tissue regions by using machine learning techniques. More recently deep learning algorithms have been employed to pre-diagnose colon and thyroid cancer. Aiming to identify tissue changes, deep neural networks can be trained under a supervised process by using the spectral values in different frequencies. The proposed study can be extended to other tissues and applied to a wide range of samples. A good dataset of samples to train the algorithms is key to achieving higher accuracy.