Three-dimensional fluorescence imaging is an essential tool in biology, used for everything from long-term imaging in developmental biology to short-term, high-resolution imaging of single cells and molecules. In recent years, there has been an influx of new imaging techniques that push the limits on both resolution and the ability to perform extended time-course imaging. Many new techniques, like structured illumination microscopy (SIM), require multiple images of the sample or increased excitation intensity to create a high-resolution image. This increased exposure can lead to an increase of sample degradation through photobleaching and phototoxicity. In this work, we focus on reducing these damaging effects by developing and implementing new imaging models for light-sheet microscopes that improve collection efficiency by allowing for additional views of the sample to be acquired in such a way that there is no increase in sample exposure or imaging time. As part of a collaboration with researchers at the National Institutes of Health, we worked with two new microscope designs. In the three-view diSPIM, we were able to improve quality for thin samples and show that for every noise level, adding the third objective increased image resolution. In our work on reflective imaging, our implementation allowed for imaging with a reflective coverslip, which improved both collection efficiency and imaging speed. In addition to adding more views of the sample, we worked on creating a more accurate imaging model of the diSPIM system to determine if image quality improved for data that has previously been acquired. In simulations, we found that a more precise model improved image quality, but using real data, we did not see such significant improvements. This suggests that there might be other factors that have a more significant contribution to the artifacts seen in the final reconstructions. We also work on determining the theoretical resolution limits for structured illumination microscopes. This resolution limit allows us to choose the necessary parameters for acquisition that produce an image quality that is adequate for analysis. Finally, we look at different reconstruction methods for SIM and use both simulations and real data to determine if these reconstruction methods approach or converge to the theoretical resolution limit. In addition to testing the theoretical vs. realized resolutions limits, we used these reconstruction methods to test the viability of using fewer acquired images during reconstruction. Previous reconstruction methods are unable to account for redundancies in the data that occur when isotropic resolution is needed. With the methods that we implemented, we found that using just four images during reconstruction, we could get comparable image quality to reconstructions that used all nine images. This reduction in data would allow for faster imaging and less exposure to the sample. Both of which are necessary for imaging fast actions in live cells.



Downloads Statistics