结直肠癌图像分析数据集

该数据集为结直肠癌组织病理学苏木精-伊红染色图像数据集（EBHI-Seg），包含2,855张内窥镜活检组织图像及对应真实标签，涵盖正常、息肉、低级别/高级别上皮内瘤变、锯齿状腺瘤和腺癌六类病变，适用于图像分割任务。数据按4:4:2划分为训练、验证和测试集，支持传统机器学习与深度学习实验。数据集规模适中，标注准确，可用于开发和验证结直肠癌自动识别算法，助力医疗大模型训练与临床辅助诊断。

This dataset is the colorectal cancer histopathological hematoxylin-eosin stained image dataset (EBHI-Seg). It contains 2,855 endoscopic biopsy tissue images and their corresponding ground truth labels, covering six lesion categories: normal, polyp, low-grade/high-grade intraepithelial neoplasia, serrated adenoma, and adenocarcinoma. It is suitable for image segmentation tasks. The data is split into training, validation and test sets at a ratio of 4:4:2, supporting experiments with both traditional machine learning and deep learning approaches. With a moderate scale and accurate annotations, this dataset can be used to develop and validate automatic colorectal cancer recognition algorithms, and assist in the training of medical large language models and clinical auxiliary diagnosis.

搜集汇总

背景与挑战

背景概述

该数据集是包含5,170张结直肠癌组织病理学图像(EBHI-Seg)的公开数据集，涵盖六种病变类型并附带真实标签，专为医学图像分割算法研究设计，实验证明深度学习模型在该数据上能达到96.5%的Dice分割精度。