Stacks of Gold: Utilizing GANs to Enhance 3D Microscopy Imaging Data
Name
Dmytro Fedorenko
Abstract
Confocal microscopy, a pivotal tool in biomedical research, offers detailed visualizations of living cells, providing insights into their spatial morphology, interactions, and life cycle progression. However, capturing and analyzing these images involve significant trade-offs. Transmitted-light (TL) microscopy, while non-invasive and relatively straightforward, yields low-contrast images of suboptimal quality, which are hard to analyze. Conversely, fluorescence (FL) microscopy delivers superior image quality but is expensive, time-consuming, and potentially harmful to cells. This thesis explores the potential of Generative Adversarial Networks (GANs) to address these challenges. We focus on extracting detailed 3D information from TL, specifically bright-field (BF) images, and enhancing the quality of 3D FL microscopy images through deconvolution, denoising, and deblurring. We present several successful GAN applications across diverse datasets, revealing the potential for in silico extraction of accurate 3D information from BF images, which was previously considered unattainable, and high-quality signal recovery from optically distorted 3D FL images. One case study demonstrates the downstream application of our in silico enhanced FL images to improve 3D reconstruction from BF. These findings could expedite the biomedical imaging workflow by reducing time expenditure and enabling novel imaging experiments, such as the non-invasive study of volumetric cell morphology.
Graduation Thesis language
English
Graduation Thesis type
Master - Computer Science
Supervisor(s)
Dmytro Fishman
Defence year
2024