中国临床医学影像杂志2025，Vol.36Issue(9)：616-620,5.DOI:10.12117/jccmi.2025.09.003
用于推动AI算法临床应用的甲状腺超声-超声造影数据集
A thyroid ultrasound and contrast-enhanced ultrasound dataset for advancing AI algorithms in clinical applications
迟剑宁 1李则蓝 1林庚 1陈佳慧 2黄瑛2
作者信息
- 1. 东北大学机器人科学与工程学院,辽宁 沈阳 110000
- 2. 中国医科大学附属盛京医院超声科,辽宁 沈阳 110004
- 折叠
摘要
Abstract
Objective:Data-driven deep learning algorithms in medical imaging face challenges such as limited datasets,difficulties in annotation,and modality constraints.This study proposes a multi-modal,multi-task dataset for AI-assisted diagnosis of thyroid nodules,combining ultrasound and contrast-enhanced ultrasound,aimed at advancing the application of deep learning algorithms in thyroid nodule disease diagnosis.Methods:A retrospective collection of patient data from Shengjing Hospital of China Medical University,was conducted from June 2016 to August 2023.The dataset included 498 cases of suspicious thyroid nodules initially suspected to be malignant in ultrasound examinations,with ultrasound image sequences and corresponding contrast-enhanced ultrasound videos.By combining ultrasound images with contrast videos,the dataset integrated multi-modal information on thyroid nodules,with annotations by experienced physicians for benign and malignant classification and nodule region segmentation,supporting both benign and malignant classification and segmentation tasks.Results:We trained and tested thyroid nodule benign and malignant diagnosis models and segmentation models using this dataset,comparing performance between single-modal and multi-modal datasets.The AI algorithms using the multi-modal dataset achieved an accuracy of over 82%for benign and malignant diagnosis and a Dice coefficient greater than 79%for segmentation accuracy.Conclusion:The experimental results demonstrate that the proposed multi-modal dataset plays a significant role in the development of AI-based thyroid nodule diagnosis and segmentation models,effectively assisting doctors in improving diagnostic efficiency.关键词
甲状腺结节/超声检查Key words
Thyroid Nodule/Ultrasonography分类
医药卫生引用本文复制引用
迟剑宁,李则蓝,林庚,陈佳慧,黄瑛..用于推动AI算法临床应用的甲状腺超声-超声造影数据集[J].中国临床医学影像杂志,2025,36(9):616-620,5.
