Segmentation of stimulated raman microscopy images using a 1D convolutional neural network

Stimulated Raman Scattering (SRS) microscopy is a powerful nonlinear optical imaging technique deriving contrast from Raman active molecular vibrations. We demonstrate, using a supervised convolutional neural network (RM-Net), the creation of chemical maps from hyperspectral Stimulated Raman Scattering images. Using a limited number (800) of training spectra, the trained RM-Net model was successfully applied to new hyperspectral images without compromising accuracy.
2022 Photonics North (PN), May 24-26, 2022, Niagara Falls, Ontario, Canada