Enteroscope Biopsy Histopathological Hematoxylin and Eosin Image Dataset for Image Segmentation Tasks (EBHI-Seg)

EBHI-Seg is provided a new publicly availabl Image Database with a
total of 5170 images.
The image input size is 224 × 224 pixels, and the format is * .png

Each sub-database has six types of images: normal, polyp, low-grade intraepithelial neoplasia, high- grade intraepithelial neoplasia, serrated adenoma, andadenocarcinoma.

We state the problem of data splitting as follows： The ratio of the training, validation and test sets is split 4:4:2 in our experiments. For the traditional machine learning experiment part, since we randomly shuffle the order of the images when creating the data set, we use the first 40% as the training set and the next 20% as the test set. This part does not involve the validation set, but we have to choose this way to ensure the accuracy of the comparative experiment. For the deep learning experiment part, we randomly select 80% of the images, and randomly divide them into two equal parts as the training set and the validation set, and the remaining 20% is the test set.

If you need it or have any questions, please contact us.

First release: 11-11-2022.

Nearst update: 11-11-2022.

Any questions: Prof. Dr.-Ing. Chen Li, lichen201096@hotmail.com, lichen@bmie.neu.edu.cn

Related people: Chen Li, Liyu Shi, Weiming Hu