深度学习在甲状腺结节超声诊断中的研究进展  

Research progress of deep learning in ultrasound diagnosis of thyroid nodules

作　　者：姚巧丽 叶菁菁 高翠霞[2] YAO Qiaoli;YE Jingjing;GAO Cuixia(The First Clinical Medical College of Gansu University of Chinese Medicine,Lanzhou 730000,China;Department of Ultrasound,Gansu Provincial Hospital,Lanzhou 730000,China)

机构地区：[1]甘肃中医药大学第一临床医学院,甘肃兰州730000 [2]甘肃省人民医院超声医学科,甘肃兰州730000

出　　处：《分子影像学杂志》2024年第9期990-996,共7页Journal of Molecular Imaging

基　　金：甘肃省自然科学基金(20JR5RA153)。

摘　　要：甲状腺结节在人群中较为常见,其评价主要依据甲状腺影像报告和数据系统。超声检查中甲状腺结节诊断的准确性与超声医生的检查技能、临床经验和思考分析能力密切相关。近年来,甲状腺癌的发病率迅速上升,如何快速有效地提高我国甲状腺结节的术前诊断能力已成为一个亟待解决的问题。深度学习作为人工智能领域的一项新技术,逐渐应用于医学影像领域,并在甲状腺超声诊断中倍受关注。本文将从甲状腺结节超声图像分割、结节良恶性鉴别、结节组织病理学预测以及颈部淋巴结智能评估等方面来介绍深度学习在甲状腺结节超声诊断中的研究进展及应用价值,旨在通过对既往相关研究的归纳总结,辅助医师深入了解该领域的发展现状,探讨新的研究方向,以期未来能为甲状腺结节的临床诊疗决策提供更准确,更全面的参考信息。Thyroid nodules are common in the population,and their evaluation is mainly based on thyroid Imaging reporting and data system.The accuracy of thyroid nodule diagnosis in ultrasound examination is closely related to the examination skills,clinical experience and thinking and analysis ability of sonographers.In recent years,the incidence of thyroid cancer has increased rapidly.How to improve the ability of preoperative diagnosis of thyroid nodules quickly and effectively in our country has become an urgent problem to be solved.As a new technology in the field of artificial intelligence,deep learning has been gradually applied in the field of medical imaging,and has attracted much attention in thyroid ultrasound diagnosis.This paper will introduce the research progress and application value of deep learning in ultrasound diagnosis of thyroid nodules from the aspects of ultrasound image segmentation of thyroid nodules,differentiation of benign and malignant nodules,histopathological prediction of nodules,and intelligent evaluation of cervical lymph nodes.The purpose of this review is to summarize the previous relevant studies to assist physicians to deeply understand the development status of this field and explore new research directions,so as to provide more accurate and comprehensive reference information for the clinical diagnosis and treatment of thyroid nodules in the future.

关 键 词：深度学习 超声诊断 甲状腺结节 人工智能 

分 类 号：R445.1[医药卫生—影像医学与核医学] R581[医药卫生—诊断学]