Sparse sampling for fast hyperspectral coherent anti-Stokes Raman scattering imaging

We demonstrate a method to increase the acquisition speed in coherent anti-Stokes Raman scattering (CARS) hyperspectral imaging while retaining the relevant spectral information. The method first determines the important spectral components of a sample from a hyper-spectral image over a small number of spatial points but a large number of spectral points covering the accessible spectral range and sampling the instrument spectral resolution at the Nyquist limit. From these components we determine a small set of frequencies needed to retrieve the weights of the components with minimum error for a given measurement noise. Hyperspectral images with a large number of spatial points, for example covering a large spatial region, are then measured at this small set of frequencies, and a reconstruction algorithm is applied to generate the full spectral range and resolution. The resulting spectra are suited to retrieve from the CARS intensity the CARS susceptibility which is linear in the concentration, and apply unsupervised quantitative analysis methods such as FSC3. We demonstrate the method on CARS hyperspectral images of human osteosarcoma U2OS cell, with a reduction in the acquisition time by a factor of 25. This method is suited also for other coherent vibrational microscopy techniques such as stimulated Raman scattering, and in general for hyperspectral imaging techniques with sequential spectral acquisition.