IAEUS Dataset

This dataset consists of endoscopic ultrasound images of pancreas. Its main purposes are the binary classification of images containing tumors and segmentation of pancreatic tumors.

Dataset description

The dataset contains a total of 7825 images of endoscopic ultrasound of pancreas. They are classified into two different classes: with tumor (positive class) or without tumor (negative class).

Structure

The dataset has been split into two parts: training and test. They are the splits used for the creation of benchmarks in the original paper description. In addition to classification, segmentation masks for the positive tumors from the training and test split are also available in a separate segmentation folder.

    DATASET
    |- train
    |   |- pos
    |   |- neg
    |- test
    |   |- pos
    |   |- neg
    |- segmentation 
    |   |- train 
    |   |- test 
    |   IA_EUS_TRAIN_COCO.json
    |   IA_EUS_TEST_COCO.json
    attributes.csv
            

In addition to the endoscopic ultrasound images, at the root of a folder there is a CSV with the 5 following attributes qualifying the tumors in the positive class:

• Visibility of the tumoral mass
• Echogenicity of a tumor
• Irregularity of the brim
• Hetegerogeneity of the mass
• Visibility of ductal structure

License

This dataset is made available under CC-BY-NC-SA.

BY: credit must be given to the creator.
NC: Only noncommercial uses of the work are permitted.
SA: Adaptations must be shared under the same terms.

How to Request the Dataset

To access the dataset, please contact the authors directly. Kindly send your request by email, including your name, affiliation, and the purpose for which you intend to use the dataset.

Cite us