Robust Detection, Visualization, Recognition, and Analysis of Cytoskeletal Structures in Fibrillar Scaffolds from 3-Dimensional Confocal Image

Living cells need to undergo subtle shape adaptations in response to the topography of their three-dimensional substrates. These shape changes are mainly determined by reorganizations of their internal cytoskeleton, with a major contribution from filamentous (F) actin. Bundles of F-actin play a major role in determining cell shape and their interaction with substrates, either as "stress fibers,” or as newly discovered "Concave Actin Bundles" (CABs), which mainly occur while endothelial cells wrap micro-fibers in culture. To better understand the morphology and functions of these CABs, it is necessary to detect, visualize, recognize, and analyse many of them in complex cellular ensembles. This dissertation presents novel algorithms to respond to this necessity.;First, we present an algorithm to detect and quantify actin-based structures in 3D cellular ensembles. In this work, we propose visual-analytic tools to delineate specific structures involving F-actin in cells. CABs often occur in hybrid cell-seeded fibrillar scaffolds and seem to envelope the fibers, as a possible mechanism of stable attachment. There is much uncertainty that accompanies the detection and identification of fibers. The tools in this research relied on well-known algorithms of image analysis. We first delineated fibers by employing an adaptive min-cut-max-flow algorithm. Then, from the extremities of the segmented fibers, a template matching, and a fiber tracking algorithm were applied to characterize the fibers more precisely within the image. CABs that surround the scaffold fibers transversally are located by observing their radial distribution around the nearby templates in focus. Then, we visually examine candidate templates that possibly contain CABs and further determine whether candidate CABs are indeed legitimate. It can be unequivocally stated that in the absence of the proposed visual analytic tools, the detection of CABs represents an intractable task.;Second, recognizing and validating CABs in complex cellular ensembles is a demanding and time-consuming task. Thus, we developed another novel algorithm to automatically recognize CABs without further human intervention. We employed a multilayer perceptron artificial neural network ("the recognizer"), which was trained to identify CABs on a voxel level on the 3D planar actin distribution image created via probability density estimation. The recognizer demonstrated high overall recognition rate and reliability in both randomized training, and in the subsequent testing experiments, for each candidate CAB, either in the group of CABs or in the group of non-CABs (assessed by human visual recognition). The remarkable recognition abilities of 3D concave cytoskeletal bundles by our neural networks-based recognizer indicated that this approach could be a good replacement of the validation by visual detection that is both tedious and inherently prone to errors.;Finally, we answered an "existential question" related to CABs: Where do CABs exist? In order to answer this question, we hypnotized that CABs exist in an area where the compactness of fibers, the compactness of actins, and density of cells are high. We proved this hypothesis by incorporating a two-point correlation function (TPCF). Based on the TPCF, we first developed a compactness measure that measures compactness of fibers, actin, and cells. Then, by using Bayes' theorem, we showed that the hypothesis was correct. In the future, we will explore additional questions: for instance, why do CABs exist in a certain area?;As a newly found structure of F-actin organization, CABs have multiple possible applications in tissue engineering. Since this dissertation is the first research on the computation of CABs, it laid a foundation for future research in the area of CABs. Future research based on the present study may attempt sophisticated and advanced applications in tissue engineering.