A Drosophila heart optical coherence microscopy dataset for automatic video segmentation.

Area of Science:

• Cardiovascular Research
• Developmental Biology
• Biomedical Imaging

Background:

• The fruit fly (Drosophila melanogaster) heart is a valuable model for studying cardiac function.
• Optical coherence microscopy (OCM) provides in vivo imaging of the beating fruit fly heart.
• Manual analysis of OCM videos is inefficient and lacks reproducibility.

Purpose of the Study:

• To introduce FlyNet 2.0+, an automated algorithm for segmenting Drosophila hearts.
• To leverage time-series data for consistent and high-quality cardiac segmentation.
• To facilitate advanced deep learning approaches for cardiac function characterization.

Main Methods:

• Development of FlyNet 2.0+, a convolutional neural network with long short-term memory (LSTM).
• Utilizing time-series information from OCM videos for segmentation.
• Creation of a large dataset of 213 Drosophila heart videos (604,000 images) with ground truth masks.

Main Results:

• FlyNet 2.0+ provides robust and accurate automated segmentation of Drosophila hearts.
• The algorithm effectively handles diverse cardiac dynamics, including arrhythmias and heart stops.
• A comprehensive dataset supporting deep learning advancements in cardiac research.

Conclusions:

• FlyNet 2.0+ offers an efficient and reproducible method for analyzing Drosophila heart function.
• The presented dataset will accelerate the development of novel deep learning tools for cardiac research.
• This work enhances the utility of Drosophila as a model organism for cardiovascular studies.