TN5000: An Ultrasound Image Dataset for Thyroid Nodule Detection and Classification

An open-access ultrasound image dataset for thyroid nodule detection and classification, i.e. the TN5000, is proposed, which contains 5,000 B-mode ultrasound images of thyroid nodule, as well as complete annotations and biopsy confirmations of the thyroid nodules from expert radiologists. More detailed, the dataset has been randomly split at the image level with a ratio of 7:1:2 into a training-validation-test set. And all the images are pre-processed into a VOC-compatible schema. The benchmark methods for thyroid nodules detection and classification have also been provided here.

本研究提出一款用于甲状腺结节检测与分类的开源超声图像数据集TN5000，该数据集包含5000张甲状腺结节B型超声图像，同时附带放射学专家标注的完整结节注释信息与活检验证结果。具体而言，该数据集以图像为单位按7:1:2的比例随机划分为训练集、验证集与测试集；所有图像均已预处理为兼容VOC（Visual Object Classes）数据集的格式规范。本数据集同时提供了用于甲状腺结节检测与分类的基准测试方法。

搜集汇总

背景与挑战

背景概述

TN5000是一个包含5000张甲状腺结节B型超声图像的开放数据集，用于结节检测和分类，提供专家标注和活检确认，并按7:1:2比例划分为训练集、验证集和测试集，所有图像预处理为VOC兼容格式。