YongchengYAO/FeTA24-Biometrics

About

FeTA24-Biometrics is based on FeTA 2024 (FeTA24) data from the University Children’s Hospital Zurich.

Dataset summary:

• 65 T2-weighted brain MR scans, 7-class segmentation masks, and 5 biometric measurements for each case

What's new?

• We transformed the original images and segmentaion masks into the biometric measurement space using the transformation matrix in FeTA24 dataset.
• We converted the landmark masks into JSON files.
  • Step 1: Get the corrdinates of 2 landmarks for each measurement.
  • Step 2: Label landmarks according to the definitions in landmarks_map.
  • Step 3: Save the coordinates of all 10 landmarks to a JSON file.
• We provided visualization of biometric measurements in corresponding 2D slices and landmarks.
• Cases without biometric measurement data are excluded from FeTA24.

Changelog 🔥

• [2 Mar, 2025] Update data structure in JSON files within Landmarks.zip. Landmarks coordinates remain unchanged.
• [2 Mar, 2025] Initial release

Data Usage Agreement

By using the dataset, you agree to the terms as follow.

• You must agree to the terms of use for FeTA24 dataset
• You must refer to the source of this dataset in any publication: https://huggingface.co/datasets/YongchengYAO/FeTA24-Biometrics

Download from Huggingface

#!/bin/bash
pip install --upgrade huggingface-hub[cli]
huggingface-cli login --token $HF_TOKEN

# python
from huggingface_hub import snapshot_download
snapshot_download(repo_id="YongchengYAO/FeTA24-Biometrics", repo_type='dataset', local_dir="/your/local/folder")

Data Info

Landmarks:

landmarks_map = {
    "P1": "most anterior point of corpus callosum",
    "P2": "most posterior point of corpus callosum",
    "P3": "most superior point of vermis",
    "P4": "most inferior point of vermis",
    "P5": "right parietal eminence ",
    "P6": "left parietal eminence",
    "P7": "right skull parietal eminence",
    "P8": "left skull parietal eminence",
    "P9": "most right point of cerebellar hemisphere",
    "P10": "most left point of cerebellar hemisphere",
}

Biometric measurement:

biometrics_map = {
    "1": "Corpus_Callosum_Length",
    "2": "Vermis_Height",
    "3": "Brain_Biparietal_Diameter",
    "4": "Skull_Biparietal_Diameter",
    "5": "Transverse_Cerebellar_Diameter",
}

Slice dimensions for measurements/landmarks:

• {"P1": 0} means the landmark point P1 should be located from dim 0 (i.e., sagittal slices)
• 0: sagittal slice
• 1: coronal slice
• 2: axial slice

LANDMARKS_SLICE_DIM = {
    "P1": 0,
    "P2": 0,
    "P3": 0,
    "P4": 0,
    "P5": 2,
    "P6": 2,
    "P7": 2,
    "P8": 2,
    "P9": 1,
    "P10": 1,
}

License

This dataset is released under the CC BY-NC 4.0 license.