This item is under embargo and not available online per the author's request. For access information, please visit http://digital.wustl.edu/publish/etd-search.html.

Abstract

The Drosophila Melanogaster is a powerful tool for cardiac research and model human disease. Previously, its heart function has been studied using Optical Coherence Microscopy (OCM), which generates videos of a beating heart cross section over time. OCM generates large volumes of data, and therefore to efficiently analyze images it is necessary to automate image segmentation. Previously, we built FlyNet using a fully convolutional neural network with added the Long-short-term memory and optimized the GPU utilization. However, the performance of these models diminishes in the presence of artifacts, such as image reflection and heart movement, resulting in manual intervention. Therefore, we proposed an attention U-Net segmentation model, FlyNet 3.0, designed for the precise segmentation of fruit fly heart OCM videos. The incorporated attention mechanism focuses weight on features extracted from the heart region by punishing unrelated areas in the input images. As a result, the model increases the prediction accuracy to 87% for images with reflections and 85% for those depicting frequent heart movements. We also extend the model to measure the dynamic heart wall thickness and validate the algorithms with histology measurement of hypertrophic cardiomyopathy.

Document Type

Article

Author's School

McKelvey School of Engineering

Author's Department

Electrical and Systems Engineering

Class Name

Electrical and Systems Engineering Undergraduate Research

Date of Submission

12-8-2023

Available for download on Friday, December 18, 2026

Share

COinS