Graduation Semester and Year

Spring 2024

Language

English

Document Type

Thesis

Degree Name

Master of Science in Computer Science

Department

Computer Science and Engineering

First Advisor

Dr. Jacob Luber

Abstract

This study focuses on the performance analysis of a diffusion model utilized for generating realistic three-dimensional (3D) reconstructions from two-dimensional (2D) microscopy data. The research aims to assess and enhance the model's efficacy in predicting various features such as volume, surface area, curvature, and surface roughness of the reconstructed shapes. Drawing inspiration from previous work on Transformer architectures, modifications are proposed to optimize the training stability of the diffusion model. Specifically, two methods are implemented and evaluated to improve the model's performance in predicting 3D shapes from 2D microscopy images. This study focuses on the performance analysis of a diffusion model utilized for generating realistic three-dimensional (3D) reconstructions from two-dimensional (2D) microscopy data. The research aims to assess and enhance the model's efficacy in predicting various features such as volume, surface area, curvature, and surface roughness of the reconstructed shapes. Relative errors for the predicted features are calculated as part of experimental evaluations, which shed light on the precision and accuracy of the model. The suggested approaches are evaluated against benchmark datasets and biomedical imaging-related tasks. In comparison to the baseline design, the study's findings show improvements in one of the methods in the model's performance metrics. With implications for a range of biomedical imaging applications, these findings advance the capabilities of diffusion models in producing high-fidelity 3D reconstructions from 2D microscopy data. The study also emphasizes how crucial performance improvements are for biomedical imaging, especially when it comes to 3D reconstruction. Understanding biological processes, disease mechanisms, and cellular structures from 2D microscopy images requires accurate and efficient reconstruction.

Keywords

Diffusion model

License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.