Automated analysis of lipid vesicle images using deep learning

Presenter Information

Cassandra Michalakis
Rina Tabuchi

Presentation Type

Poster

Faculty Advisor

Eli Lee

Access Type

Event

Start Date

26-4-2023 1:44 PM

End Date

26-4-2023 2:45 PM

Description

The importance of membrane reconstruction with giant unilamellar vesicles and optical imaging in biochemistry & biophysics calls for a more efficient mode of analysis, as opposed to the current, manual process done by humans. A promising avenue to secure this analysis is with deep-learning. Its uses are expanding in numerous technological applications, even with everyday items such as face-recognition in cell phones. Applying a deep-learning algorithm in a biological framework consists of processing hundreds of vesicle images to classify the vesicle state, and quantify the fluorescence intensity. The automated vesicle-analysis program we created from previous research is currently satisfying the basic demands of vesicle analysis (specifically to detect, classify, and quantify vesicle images). This program operates with the assistance of the deep-learning algorithm. The following objectives are to expand the program's capability for more compounded detections (ie. vesicles with complex morphologies), while also instituting computer-assisted analysis in future classrooms (via educational biochemistry lab modules). Artificially-created GUV’s (made using fluorescence microscopy) will be created, so we can use images of the GUVs to train, test, and develop deep-learning software.

This document is currently not available here.

Share

COinS
 
Apr 26th, 1:44 PM Apr 26th, 2:45 PM

Automated analysis of lipid vesicle images using deep learning

The importance of membrane reconstruction with giant unilamellar vesicles and optical imaging in biochemistry & biophysics calls for a more efficient mode of analysis, as opposed to the current, manual process done by humans. A promising avenue to secure this analysis is with deep-learning. Its uses are expanding in numerous technological applications, even with everyday items such as face-recognition in cell phones. Applying a deep-learning algorithm in a biological framework consists of processing hundreds of vesicle images to classify the vesicle state, and quantify the fluorescence intensity. The automated vesicle-analysis program we created from previous research is currently satisfying the basic demands of vesicle analysis (specifically to detect, classify, and quantify vesicle images). This program operates with the assistance of the deep-learning algorithm. The following objectives are to expand the program's capability for more compounded detections (ie. vesicles with complex morphologies), while also instituting computer-assisted analysis in future classrooms (via educational biochemistry lab modules). Artificially-created GUV’s (made using fluorescence microscopy) will be created, so we can use images of the GUVs to train, test, and develop deep-learning software.