Many of the advantages of hyperspectral microscopy data are not readily accessible to the researcher. The human brain has the capacity to process no more then three colors and artificial visualization of further color channels comes at the expense of dynamic range, or time and space resolution. To exploit the full potential of enhanced machine vision, it is therefore necessary to computationally analyze data and visualize the result in a way interpretable by the user. We developed a method of blind separation of individual fluorophores in complex hyperspectral images. Our method does not require prior knowledge of the fluorophore spectra, but it is able to incorporate such knowledge if it is available. Our separation methods make it possible to separate all fluorescence signals, even if they differ from the expected ideal spectra due to biological processes or instrument calibration issues.
Improved spectral resolution in confocal microscopy is important in both materials research and biological studies. When studying dynamic fluorescent samples, it is ideal to image at high spectral resolution while retaining good viability and signal sensitivity, as well as high temporal fidelity. In particular there is great need to image at high speeds with spectral resolution to image dynamic events such as cell division, trafficking and cancer invasion and progression. We have developed a high-speed spectral confocal microscope that utilizes a low-loss Amici prism for spectral separation in conjunction with a multi-point confocal design for high-speed image capture. Our system achieves 15 channels of spectral resolution at a 4 frames per second imaging rate and with minimal loss of overall signal.