Drosophila melanogaster heart tube segmentation in optical coherence tomography through an attention LSTM U-Net model

Xiangping Ouyang, Abigail Matt, Fei Wang, Elena Gracheva, Chao Zhou

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

The Drosophila Melanogaster is a powerful tool for cardiac research due to its ability for disease modeling. OCM provides cross-sectional images of its beating heart tube, which can be segmented to quantify heart parameters. Here, we expanded upon an optimized LSTM U-Net model introduced in 2023, by Fishman et al., to improve segmentation performance when artifacts are present. We incorporated attention gates via skip connections between each level of the LSTM U-Net model. This model increases the prediction intersection over union (IOU) from 0.86 to 0.89 for images with reflection artifacts and from 0.81 to 0.89 for those depicting frequent heart movement.

Original languageEnglish
Title of host publicationDiagnostic and Therapeutic Applications of Light in Cardiology 2024
EditorsLaura Marcu, Gijs van Soest, Christos Bourantas
PublisherSPIE
ISBN (Electronic)9781510668973
DOIs
StatePublished - 2024
EventDiagnostic and Therapeutic Applications of Light in Cardiology 2024 - San Francisco, United States
Duration: Jan 27 2024Jan 28 2024

Publication series

NameProgress in Biomedical Optics and Imaging - Proceedings of SPIE
Volume12819
ISSN (Print)1605-7422

Conference

ConferenceDiagnostic and Therapeutic Applications of Light in Cardiology 2024
Country/TerritoryUnited States
CitySan Francisco
Period01/27/2401/28/24

Keywords

  • attention model
  • Convolutional Neural Network
  • Drosophila
  • LSTM
  • Optical Coherence Tomography

Fingerprint

Dive into the research topics of 'Drosophila melanogaster heart tube segmentation in optical coherence tomography through an attention LSTM U-Net model'. Together they form a unique fingerprint.

Cite this