人工智能在头影测量自动定点算法上的研究进展  

Progress of research on using artificial intelligence for cephalometric automatic-landmarking algorithms

作　　者：汪云毅 朱珠 张峰[1,2] Wang Yunyi;Zhu Zhu;Zhang Feng(Dept.of Stomatology,Children’s Hospital,Zhejiang University School of Medicine,Hangzhou 310000,China;National Clinical Research Center for Child Health,Hangzhou 310000,China;Dept.of Data and Information,Children’s Hospital,Zhejiang University School of Medicine,Hangzhou 310000,China;Sino-Finland Joint AI Laboratory for Child Health of Zhejiang Province,Hangzhou 310000,China)

机构地区：[1]浙江大学医学院附属儿童医院口腔科,杭州310000 [2]国家儿童健康与疾病临床医学研究中心,杭州310000 [3]浙江大学医学院附属儿童医院数据信息部,杭州310000 [4]浙江-芬兰儿童健康人工智能联合实验室,杭州310000

出　　处：《国际口腔医学杂志》2024年第5期630-641,共12页International Journal of Stomatology

基　　金：浙江省科学技术厅“尖兵”“领雁”研发攻关计划(2023C03101)。

摘　　要：头影测量是正畸诊断和治疗过程中不可或缺的分析手段。高精度定位头影测量的标志点对于确保正畸临床诊断的准确性和治疗目标的正确性至关重要。随着计算机辅助技术特别是人工智能的发展,头影测量标志点从手动标注逐渐进展到自动定点,并已应用于临床实践。从基于知识方法到基于模型和模板匹配方法,再到现在的机器学习及深度学习方法,人工智能在传统头颅侧位片上标志点检测的准确率已有显著提高,但在图像数据更精确的三维图像上,自动定点尚处于起步阶段。本文旨在综述人工智能在头影测量自动定点算法方面的研究进展,并对其未来研究方向进行展望。Cephalometry is an indispensable tool in the diagnosis and treatment process of orthodontics.The high-precision localization of cephalometric landmarks is crucial to ensuring the accuracy of clinical orthodontic diagnosis and the correctness of treatment objectives.In clinical practice,doctors usually need to manually annotate these landmarks,which is time consuming and subjective.Over the past few decades,the concept of automated cephalometric landmark identification has emerged with the development of computer-assisted technology,particularly artificial intelligence.This automated identification is gradually being applied in clinical practice.From the knowledge-based method to the model-based and template-based matching method,and subsequently to the current machine-learning and deep-learning methods,the accuracy of artificial intelligence in landmark detection on traditional lateral cephalogram has significantly improved.However,in three-dimensional images with more accurate image data,automatic landmarking remains in its infancy.This article aims to review the progress of research on artificial intelligence in automatic-landmarking algorithm for cephalometry,as well as to look forward to its future research directions.

关 键 词：人工智能 头影测量标志点 自动定点 机器学习 深度学习 自动化分析 计算机视觉 

分 类 号：R783.5[医药卫生—口腔医学]