CHAOS: Combined CT and MRI Abdominal Organ Segmentation Challenge Data

40 total cases of abdominal CT and MRI scans from a 2019 segmentation challenge provide ground truth masks for healthy organs. The dataset includes 20 training and 20 testing cases for both CT and MRI modalities, with MRI scans featuring T1-Dual and T2 SPIR sequences. Researchers from Dokuz Eylül University released this data to support the development of medical image segmentation algorithms.

Use Cases

• Train multi-modal segmentation models based on the provided CT and MRI image pairs.
• Benchmark segmentation algorithm performance for the liver, kidneys, and spleen using the provided ground truth masks.
• Develop and validate models for healthy organ delineation in abdominal scans.
• Explore the fusion of CT and MRI data for improved anatomical analysis.

Strengths

• Includes ground truth segmentation masks for training data, which are essential for supervised learning.
• Provides a multi-modal dataset with 20 CT and 20 MRI cases, each with separate training and testing splits.
• Specifies annotated organs: livers in CT, and livers, left/right kidneys, and spleens in MRI.

Limitations

• Ground truth masks for the test set are withheld for challenge validation and are not publicly available.
• The dataset scale is limited to 40 total cases (20 CT, 20 MRI), which may be small for training deep learning models without augmentation.
• Column-level documentation and specific file format details are not provided in the input metadata.

Provenance

Source

Dokuz Eylül University (via Zenodo)

Collection Method

Collected for the CHAOS (Combined Healthy Abdominal Organ Segmentation) Challenge.