3D Segmentation for Super-Resolution Imaging of Bacterial Biofilms

The emergence of super-resolution microscopy using lattice light sheet is enabling the exploration of structure and function in living tissues such as bacterial biofilms that have mysterious interconnections and organization. Such a technique overcomes the diffraction limit in traditional confocal microscopy, which also provides non-phototoxic three-dimensional images at resolutions ten times smaller than that provided by traditional light microscopy. Unfortunately, the standard tools used in the image analysis community to perform segmentation and other higher-level analyses cannot be applied naïvely to these data.
In this thesis, Bact-3D is presented for segmenting super-resolution images of multi-leveled, living bacteria cultured in vitro. It incorporates a novel initialization approach that exploits the geometry of the bacterial cells as well as iterative local level set evolution that is tailored to the biological application. In experiments where segmentation is used as precursor to cell detection, the Bact-3D matches or improves upon the Dice score and mean-squared error of three existing methods, while yielding a substantial improvement in cell detection accuracy. In addition to providing improvements in performance over the state-of-the-art, this report also characterizes the tradeoff between imaging resolution and segmentation quality. Evaluation of the algorithm complexity and operating times of Bact-3D as well as alternative solutions are also discussed.
Improvements based on Bact-3D method are proposed to increase the applicability in extended datasets, where bacteria are randomly distributed and densely overlapped. The improved methodology ensures the locality of 3D level set based active surface model, which is achieved by proposing intuitive distance velocity field (DVF), of the level set evolution. Preliminary experimental results that compare the performance of improved approach with two existing methods exhibit competitive advantage in segmenting neighboring cells. Comprehensive application of combining geometric and parametric segmentation techniques on these challenging datasets is also proposed for future work.