In Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy
will defend his dissertation proposal
Analysis of Scanning Electron Microscopy Images with Illumination Inhomogeneity and Touching-Crossing Cells
Clostridioides difficile infection (CDI) is a significant cause of death due to infectious gastroenteritis in the USA and a significant source of morbidity. Treatments for CDI are being developed but currently there are no method to quantify the efficiency of the treatments. Extraction of cell-related information (e.g., length, location, deformation) in scanning electron microscopy (SEM) images is the critical step in quantification and comparison of CDI treatments. However, analysis of SEM images is challenging due to the following challenges: Inhomogeneous illumination causes shadows on the cells and bright areas around the cells that hinders the segmentation the cells from the background. Moreover, in many cases, cells are cluttered creating touching and crossing cells. Separation of touching cells is challenging due to partial overlap of cell walls and minor occlusion of the cell body. Furthermore, in case of crossing cells, a large portion of the cell body of one or multiple cells are occluded, making the detection of cells challenging. The goal of this research is to develop a computational tool that analyses scanning electron microscopy images of elongated cells, addressing challenges of inhomogeneous illumination, touching cells, and crossing cells, to provide specific cell-related information (i.e., length, deformation, location.) The specific objectives are to: (1) Generate a dataset of synthesized SEM cell images with more than 20,000 images; (2) Develop and evaluate a pipeline for segmentation of cell images with inhomogeneous illumination; (3) Modify the pipeline to address the challenges of inhomogeneous illumination and separating touching cells; (4) Modify the pipeline to address the challenges of inhomogeneous illumination, separating touching cells, and separating crossing cells; (5) Develop and evaluate a computational tool that provides specific cell-related information for all the cells present in the SEM image. In this proposal, we detail the progress made so far and outline the remaining work.
Date: Thursday, December 6, 2018
Time: 10:00 AM
Place: HBS 317
Advisors: Dr. Ioannis A. Kakadiaris
Faculty, students, and the general public are invited.