EchoXFlow

A Beamspace Echocardiography Dataset for Cardiac Motion, Flow, and Function

---

Overview

EchoXFlow provides echocardiography recordings stored as Croissant metadata and Zarr recording stores.

This repo supports parsing, visualisation and dataloading of time-resolved 1D, 2D, and 3D B-mode and Doppler data.

---

Highlights

• Read EchoXFlow data from Croissant metadata and Zarr stores
• Discover recordings directly from metadata
• Load typed streams for imaging, Doppler, and ECG
• Render frames and videos for inspection
• Use scaffolded examples for segmentation and prediction tasks

---

Benchmark Tasks

Task 1: Tissue Velocity

Task 2: Blood Flow

Task 3: LV Segmentation

---

Dataset Setup

Download the dataset from:

🤗 https://huggingface.co/datasets/Ahus-AIM/EchoXFlow

Then set the dataset root in src/echoxflow/config/defaults.yml:

data:
  root: "/path/to/EchoXFlow"

Alternatively, set:

export ECHOXFLOW_DATA_ROOT=/path/to/EchoXFlow

---

Reproduce Our Results

Generate the dataset statistics table with scripts/croissant_summary_table.py:

uv run python scripts/croissant_summary_table.py /path/to/EchoXFlow/croissant.json

Run the benchmark matrix with scripts/run_full_benchmark.sh:

scripts/run_full_benchmark.sh --data-root /path/to/EchoXFlow --cv --gpus 0,1,2

---

Installation

uv venv
uv pip install --editable . --requirement requirements-dev.txt

---

Quick Start

from pathlib import Path

from echoxflow import (
    find_recordings,
    load_croissant,
    open_recording,
    render_recording_video,
)

catalog = load_croissant("/path/to/EchoXFlow/croissant.json")

records = find_recordings(
    croissant=catalog,
    array_paths=(
        "2d_brightness_mode",
        "2d_color_doppler_velocity",
        "2d_color_doppler_power",
    ),
    require_all=True,
)

record = records[0]

output = Path("outputs/color_doppler.mp4")
render_recording_video(record, output, view_mode="both")

store = open_recording(record)
power = store.load_stream("2d_color_doppler_power")

print(power.data.shape)

---

Development

uv run pre-commit install
uv run pre-commit run --all-files

---

Citation

@misc{stenhede_2026_echoxflow,
  title = {EchoXFlow: A Beamspace Echocardiography Dataset for Cardiac Motion, Flow, and Function},
  author = {Stenhede, Elias and Sulkowska, Joanna and Orstad, Eivind Bj{\o}rkan and Schirmer, Henrik and Ranjbar, Arian},
  year = {2026},
  eprint = {2605.05447},
  doi = {10.48550/arXiv.2605.05447},
  url = {https://arxiv.org/abs/2605.05447}
}