EchoSlicer

This repository contains the official code for extracting standard echocardiography views from 3D+1 volumes.

What is EchoSlicer?

EchoSlicer automatically extracts standard echocardiography views (A2C, A3C, A4C, A5C, PLAX, PSAX) from 3D echocardiography volumes using deep learning-based landmark detection and view classifier guided search.

Installation

1. Clone the repository

git clone https://github.com/echonet/3d-echo
cd 3d-echo

2. Download pre-trained models

View Classifier:

wget https://github.com/echonet/EchoPrime/releases/download/v1.0.0/model_data.zip
unzip model_data.zip
mv model_data/weights/view_classifier.pt view_classifier/
rm -r model_data
rm model_data.zip

LV Segmentation Model:

wget https://github.com/echonet/dynamic/releases/download/v1.0.0/deeplabv3_resnet50_random.pt
mv deeplabv3_resnet50_random.pt lv_segmenter/

3. Install dependencies

Using uv:

uv sync && source .venv/bin/activate

Demo

python demo.py

The extracted views will appear in the sample_output directory. To visualize the output in the interactive streamlit app

   streamlit run app.py

A demo should be running, as shown in this video

Output

The output contains:

• 8 mp4 files for each extracted view: A2C, A3C, A4C, A5C, PLAX, PSAX (apex), PSAX (papillary muscles), PSAX (mitral valve),
• 8 csv files containing metadata for each view, including:
  • Plane orientation parameters (angles and position)
  • View dimensions in centimeters (height and width)
  • View classifier confidence score

Dataset

We also release an open-source dataset of 29 three-dimensional echocardiogram videos. The data was collected by the authors themselves, who visited the echocardiography lab and recorded their own heart images. The dataset includes 15 volumes from Person A, 5 from Person B, 5 from Person C, and 4 from Person D. We hope this resource will encourage further research on AI in echocardiography.

The dataset can be downloaded from the github release

wget https://github.com/echonet/3d-echo/releases/download/v1.0/dataset.zip

Citation

If you use this code or the dataset in your research, please cite the following paper:

Automated Interpretable 2D Video Extraction from 3D Echocardiography Milos Vukadinovic, Hirotaka Ieki, Yuki Sahashi, David Ouyang, and Bryan He.
arXiv preprint arXiv:2511.15946, 2025.

@article{vukadinovic2025automated,
  title={Automated Interpretable 2D Video Extraction from 3D Echocardiography},
  author={Vukadinovic, Milos and Ieki, Hirotaka and Sahashi, Yuki and Ouyang, David and He, Bryan},
  journal={arXiv preprint arXiv:2511.15946},
  year={2025},
  url={[https://arxiv.org/abs/2511.15946](https://arxiv.org/abs/2511.15946)}
}