Diagnosis of colorectal cancer by near-infrared optical fiber spectroscopy and random forest

摘要: Near-infrared (NIR) spectroscopy has such advantages as being noninvasive, fast, relatively inexpensive, and no risk of ionizing radiation. Differences in the NIR signals can reflect many physiological changes, which are turn associated with factors vascularization, cellularity, oxygen consumption, or remodeling. spectral differences between colorectal cancer healthy tissues were investigated. A Fourier transform instrument equipped a fiber-optic probe was used to mimic situ clinical measurements. total 186 spectra collected then underwent preprocessing standard normalize variate (SNV) for removing unwanted background variances. All specimen spots collection confirmed staining examination by an experienced pathologist so ensure representative pathology. Principal component analysis (PCA) uncover possible clustering. Several methods including random forest (RF), partial least squares-discriminant (PLSDA), K-nearest neighbor classification regression tree (CART) extract features construct diagnostic models. By comparison, it reveals that, even if obvious difference misclassified ratio (MCR) observed these models, RF is preferable since quicker, more convenient insensitive over-fitting. The results indicate that coupled model serve potential tool discriminating from normal ones.