chehablab/AOMIC_t1w

This dataset is derived from the Amsterdam Open MRI Collection (AOMIC), a large-scale neuroimaging dataset of healthy individuals. The original dataset contains high-quality structural MRI scans acquired under standardized research protocols. Here, we provide a 2D slice-based version of T1-weighted (T1w) MRI volumes, extracted only from healthy subjects. The dataset is designed for efficient deep learning experiments, including self-supervised learning, representation learning, anomaly detection, and slice-level modeling.

📦 Dataset Structure

We have three splits as from the original dataset:

• ID1000
• PIOP1
• PIOP2

Each entry corresponds to a single 2D slice from a 3D T1-weighted MRI volume:

• volume_id → Unique identifier for the subject/volume
• slice_id → Index of the slice within the volume
• run → Acquisition run identifier (Volumes in the 'ID1000' split have multiple scans for same participant)
• type → 'T1w'
• image → 2D T1-weighted MRI slice

⚙️ Preprocessing

• Only T1-weighted (T1w) volumes were used
• All subjects are healthy (no pathology labels)
• Volumes were converted into 2D axial slices

Caveat

• Late slices (Majorly black images) in each volume have noise in them

🚀 Usage

from datasets import load_dataset
import matplotlib.pyplot as plt

ds = load_dataset("chehablab/AOMIC_t1w", split="ID1000")

sample = ds[314]
img = sample["image"]

plt.imshow(img, cmap="gray")
plt.title(f"{sample['volume_id']} | Slice {sample['slice_id']}")
plt.axis("off")
plt.show()

🧪 Use Cases

• Self-supervised learning (SSL) on brain MRI
• Learning anatomical representations of healthy brains
• Anomaly detection using healthy-only distribution
• Pretraining for downstream medical imaging tasks
• Slice-level classification or reconstruction

📚 Citation

If you use this dataset, please acknowledge our lab Chehab Lab and cite the original AOMIC dataset:

@article{Snoek2021,
  author  = {Snoek, Lukas and van der Miesen, Maite M. and Beemsterboer, Tinka and van der Leij, Andries and Eigenhuis, Annemarie and Scholte, H. Steven},
  title   = {The Amsterdam Open MRI Collection, a set of multimodal MRI datasets for individual difference analyses},
  journal = {Scientific Data},
  year    = {2021},
  volume  = {8},
  number  = {1},
  pages   = {85},
  doi     = {10.1038/s41597-021-00870-6},
}

📜 License

This dataset is released under the Creative Commons CC0 1.0 Universal (CC0 1.0) Public Domain Dedication. You may copy, modify, distribute, and use the data, even for commercial purposes, without asking permission.

Chehab Lab @ 2026