In vivo Raman spectroscopy coupled with machine learning methods was used for tissue diagnostics of precancerous and cancerous lesions of the colon and rectum.

Colorectal cancer is one of the most common and often fatal cancers in humans, but it has the highest chance of a cure if detected at an early precancerous stage. Carcinogenesis in the colon begins as an uncontrolled growth forming polyps. Some of these polyps can finally be converted to colon cancer. Early diagnosis of adenomatous polyps is the main approach for screening and preventing colorectal cancer, and vibration spectroscopy can be used for this purpose. This work is focused on evaluating FTIR and Raman spectroscopy as a tool in the ex vivo analysis of colorectal polyps, which could be important for the early diagnosis of colorectal carcinoma. Multivariate analyses (PCA and LDA) were used to assist the spectroscopic discrimination of normal colon tissue, as well as benign and malignant colon polyps. The spectra demonstrated evident differences in the characteristic bands of the main tissue constituents, i.e., proteins, nucleic acids, lipids, polysaccharides, etc. Suitable models for discriminating the three mentioned diagnostic groups were proposed based on multivariate analyses of the spectroscopic data. LDA classification was especially successful in the case of a combined set of 55 variables from the FTIR, FT Raman and dispersion Raman spectra. This model can be proposed for ex vivo colorectal cancer diagnostics in combination with the colonoscopic extraction of colon polyps for further testing. This pilot study is a precursor for the further evaluation of the diagnostic potential for the simultaneous in vivo application of colonoscopic Raman probes.