Figure 2
From: Deep learning data augmentation for Raman spectroscopy cancer tissue classification
Top row (a–c) Genuine Raman spectra measured from a dataset with three categories: BCC (basal cell carcinoma), NORMAL, and SCC (squamous cell carcinoma), which are representative of the range of spectra measured in this work. Each sample has 1608 dimensions, where each dimension correspond to the wavelength number of the Raman shift (ranging from 600.237 to 1699.39 \({\hbox {cm}}^{-1}\)). The value of each dimension represents the Raman intensity. “Treated” means that the sample has been treated using a high-powered IR laser to ablate the tissue surface, or no laser treatment otherwise (“Untreated”). Each colored curve represents one RS sample. Bottom Row (d–f) Synthetic Raman spectra samples generated using GAN for each cancer category (BCC, NORMAL, and SCC). Each category has two generated RS samples.
