Development of algorithms for the restoration of focus in widefield fluorescence microscopy images of cells cultured on flat and textured substrates

Widefield fluorescence microscopy is widely used for the visualization of biological structures, but image quality can be strongly affected by out-of-focus blur, particularly when cells are cultured on three-dimensional substrates. The aim of this thesis was to develop a computational reconstruction pipeline to improve the visualization of biological structures in widefield fluorescence microscopy images acquired from cells cultured on both flat glass coverslips and three-dimensional GreenBone scaffolds. The pipeline combined preprocessing, Richardson–Lucy deconvolution and post-processing steps. Within the deconvolution step, three different point spread function (PSF) models were investigated: an Airy-based PSF, a Gaussian PSF and a software-generated PSF. The performance and accuracy of the reconstruction pipeline were then evaluated through visual analysis and quantitative metrics, and the pipeline was also compared with DeconvolutionLab2. The results showed that the proposed approach improved the visualization of both nuclear and cytoskeletal structures. However, no single PSF provided the best performance under all experimental conditions. The optimal PSF depended on the substrate, biological structure and level of image degradation. Overall, this work demonstrates the potential of computational reconstruction to improve the visualization of biological structures in widefield fluorescence microscopy images acquired on both flat and textured substrates, but the reconstruction parameters need to be adapted to the imaging conditions.