Multispectral fluorescence microscopic imaging based on compressive sensing

Compressive sensing (CS) theory is used in fluorescence microscopy imaging and a new microscopic imaging system is designed and implemented. A liquid crystal light valve is employed to calculate the linear projection of an image onto pseudorandom patterns. Fluorescence is collected on a point detector. Images of the samples are acquired combined with the reconstruction theory of CS. The number of samples is smaller than that imposed by the Nyquist-Shannon theorem. The system hardware is simple as scanning is unnecessary during the imaging process. Compared with the traditional spectral imaging modalities, such as using optical filter and raster scanning, this system only needs a spectrometer to acquire signal and then multispectral images are reconstructed from measurements corresponding to a set of sub-bands. As the fluorescence microscopy imaging suffers fluorescence decay during imaging process, in this experiment, data preprocessing such as intensity normalization is applied and the results indicate that the influence of fluorescence decay on reconstruction is eliminated effectively with this processing method.