Files in this item

FilesDescriptionFormat

application/pdf

application/pdfECE499-Sp2019-he-Yuchen.pdf (1MB)Restricted to U of Illinois
(no description provided)PDF

Description

Title:Automated sperm morphology identification using machine learning
Author(s):He, Yuchen
Contributor(s):Popescu, Gabriel
Subject(s):sperm fertility analysis
quantitative phase imaging
semantic segmentation
Abstract:Morphology is a branch of biology that studies the form of organisms and their specific structural features. Sperm morphology is one of the factors that contribute to male fertility. Thus, the analysis of sperm morphological features is an important part of reproduction research. Last year, the Quantitative Light Imaging (QLI) Laboratory at University of Illinois started a project using quantitative phase imaging (QPI) methods to study the correlation between sperm morphology and fertility outcomes in bulls. Unlike fluorescence-based methods, QPI methods like spatial light interference microscopy (SLIM) avoid reducing cell viability. Once the images were acquired, researchers would manually segment each sperm into three parts and collected the morphological parameters like mass and size. This thesis proposes an end-to-end system which, given an image acquired using SLIM, can identify the morphology parameters automatically. The images will be processed and fed into a semantic segmentation model based on U-Net architecture. This thesis shows that the model can achieve satisfactory accuracy and it greatly facilitates the original project by providing analyzed data efficiently. The model can also be integrated into the imaging systems at the QLI lab for real-time application. Finally, this work explores the possibility of predicting fertility outcome, such as cleavage percentage and pregnancy percentage, based on the segmentation map generated by the model.
Issue Date:2019-05
Genre:Other
Type:Text
Language:English
URI:http://hdl.handle.net/2142/104016
Date Available in IDEALS:2019-06-14


This item appears in the following Collection(s)

Item Statistics