Semantic Segmentation of LNPs from (Cryo-TEM) images using Self-supervised learning techniques and automatic LNPs size distribution quantification
Presenter Type
UNO Graduate Student (Doctoral)
Major/Field of Study
Computer Science
Other
PhD in IT
Author ORCID Identifier
0000-0002-3898-0089
Advisor Information
Department Chair, Professor, Charles W. and Margre H. Durham Distinguished Professor, Computer Science, UNO
Location
MBSC306 - G (Doctoral)
Presentation Type
Oral Presentation
Start Date
24-3-2023 10:30 AM
End Date
24-3-2023 11:45 AM
Abstract
It is important to accurately measure the size distribution of lipid nanoparticles (LNPs) which play a crucial role in drug encapsulation efficiency, biodistribution, and cellular uptake in drug delivery systems. Cryogenic Transmission Electron Microscopy (Cryo-TEM) is a technique used to visualize the morphological structures of LNPs, but measuring accurate LNP size distribution is challenging. This project aims to quantification of LNPs size distribution automatically. The activities to be performed can be divided into two main steps. In the first step, self-supervised techniques are used for semantic segmentation tasks. A pretext task is designed for learning the representation of unlabeled Cryo-TEM images, which is then fine-tuned using a few labeled images. The resulting learning representation is used to perform semantic segmentation tasks for identifying LNPs in Cryo-TEM images. In the second step, image processing techniques are used to identify the positions and radii of each LNP from the segmented images generated in the previous step. Finally, a histogram graph is generated to automatically quantify the LNPs in the images.
Scheduling
10:45 a.m.-Noon, 2:30 -3:45 p.m.
Semantic Segmentation of LNPs from (Cryo-TEM) images using Self-supervised learning techniques and automatic LNPs size distribution quantification
MBSC306 - G (Doctoral)
It is important to accurately measure the size distribution of lipid nanoparticles (LNPs) which play a crucial role in drug encapsulation efficiency, biodistribution, and cellular uptake in drug delivery systems. Cryogenic Transmission Electron Microscopy (Cryo-TEM) is a technique used to visualize the morphological structures of LNPs, but measuring accurate LNP size distribution is challenging. This project aims to quantification of LNPs size distribution automatically. The activities to be performed can be divided into two main steps. In the first step, self-supervised techniques are used for semantic segmentation tasks. A pretext task is designed for learning the representation of unlabeled Cryo-TEM images, which is then fine-tuned using a few labeled images. The resulting learning representation is used to perform semantic segmentation tasks for identifying LNPs in Cryo-TEM images. In the second step, image processing techniques are used to identify the positions and radii of each LNP from the segmented images generated in the previous step. Finally, a histogram graph is generated to automatically quantify the LNPs in the images.