Vindr-Mammo

该数据集是一个大规模的全场数字乳腺摄影数据集，包含了来自5000项越南患者研究的20,000次扫描。这些扫描被分为两类：可疑良性（BI-RADS 2）和可疑恶性（BI-RADS 4和5），标签分布不平衡。具体来说，数据集中有2,831个同侧视图样本，这些样本被进一步划分为训练集（包括1,870个良性病例和395个恶性病例）和测试集（包括467个良性病例和99个恶性病例）。该数据集的规模为20,000次扫描，任务是对乳腺摄影进行分类。

This dataset is a large-scale full-field digital mammography cohort containing 20,000 scans derived from 5,000 Vietnamese patient studies. These scans are categorized into two classes: suspicious benign (BI-RADS 2) and suspicious malignant (BI-RADS 4 and 5), with an imbalanced class distribution. Specifically, there are 2,831 ipsilateral view samples in this dataset, which are further split into a training set and a test set. The training set comprises 1,870 benign cases and 395 malignant cases, while the test set includes 467 benign cases and 99 malignant cases. The primary task of this dataset is mammographic classification.

搜集汇总

背景与挑战

背景概述

VinDr-Mammo是一个大规模的全视野数字乳腺X线摄影基准数据集，包含5,000个四视图检查，每个检查都提供了乳腺级别的评估和发现标注，并经过独立双读和仲裁以确保数据质量。该数据集旨在支持乳腺癌筛查的计算机辅助检测和诊断工具开发，可从Physionet公开获取。