基于深度学习法构建适合平原和高原儿童的骨龄预测模型  

Construction and Validation of A Deep Learning⁃based Bone Age Prediction Model for Children Living in Both Plain and Highland Regions

作　　者：刘琪星 汪火根 次旦旺久[3] 土旦阿旺 杨美杰 普琼穷达 杨筱[6] 潘慧[7] 王凤丹[1] LIU Qixing;WANG Huogen;CIDAN Wangjiu;TUDAN Awang;YANG Meijie;PUQIONG Qiongda;YANG Xiao;PAN Hui;WANG Fengdan(Department of Radiology,Chinese Academy of Medical Sciences&Peking Union Medical College,Beijing 100730,China;College of Computer Science and Technology,Zhejiang University,Hangzhou 310000,China;Department of Radiology,Tibet Autonomous Region People's Hospital,Lhasa 850000,China;Department of Radiology,People's Hospital of Nyima County,Nagqu,Tibet 852600,China;Department of Radiology,People's Hospital of Nagqu,Nagqu,Tibet 852000,China;Department of Ultrasound,Peking Union Medical College Hospital,Chinese Academy of Medical Sciences&Peking Union Medical College,Beijing 100730,China;Department of Endocrinology,Peking Union Medical College Hospital,Chinese Academy of Medical Sciences&Peking Union Medical College,Beijing 100730,China)

机构地区：[1]中国医学科学院北京协和医院放射科,北京100730 [2]浙江大学计算机科学与技术学院,杭州310000 [3]西藏自治区人民医院放射科,拉萨850000 [4]尼玛县人民医院放射科,西藏那曲852600 [5]那曲市人民医院放射科,西藏那曲852000 [6]中国医学科学院北京协和医院超声医学科,北京100730 [7]中国医学科学院北京协和医院内分泌科,北京100730

出　　处：《协和医学杂志》2024年第6期1439-1446,共8页Medical Journal of Peking Union Medical College Hospital

基　　金：国家自然科学基金青年科学基金(82001900);中央高水平医院临床科研专项(2022⁃PUMCH⁃A⁃003);中国医学科学院医学与健康科技创新工程(2021⁃I2M⁃1⁃051)。

摘　　要：目的基于深度学习法构建适合平原和高原儿童的骨龄预测模型,并进行临床验证。方法本研究共纳入三个数据集[北美放射学会(Radiology Society of North America,RSNA)数据集,包括训练集12611例、验证集1425例、测试集200例;放射学手部姿势评估(Radiological Hand Pose Estimation,RHPE)数据集,包括训练集5491例、验证集713例和测试集79例;自建数据集,包括训练集825例和测试集351例],用于模型的训练和内部验证。自建数据集回顾性纳入北京协和医院(745例,均为汉族)和西藏自治区人民医院(431例,其中汉族114例、藏族317例)共1176例儿童的左手腕部X线影像。此外,研究还纳入了来自尼玛县人民医院的外部测试集(256例,均为藏族),用于模型的外部验证。应用深度学习法构建骨龄预测模型(ethnicity vision gender⁃bone age net,EVG⁃BANet),并采用平均绝对差异(mean absolute difference,MAD)和1岁以内准确率作为模型的评价指标。结果EVG⁃BANet模型在RSNA和RHPE测试集中的MAD分别为0.34岁和0.52岁。在自建数据集中,该模型的MAD为0.47(95%CI:0.43~0.50)岁,1岁以内准确率为97.72%(95%CI:95.56%~99.01%);在外部测试集中,该模型的MAD为0.53(95%CI:0.48~0.58)岁,1岁以内准确率为89.45%(95%CI:85.03%~92.93%)。结论EVG⁃BANet模型在平原和高原儿童中均表现出较高的准确性,具有一定的推广应用价值。Objective To construct and validate a deep learning-based bone age prediction model for children living in both plain and highland regions.Methods A model named“ethnicity vision gender-bone age net(EVG-BANet)”was trained using three datasets,including the Radiology Society of North America(RSNA)dataset[training set(n=12611),validation set(n=1425),test set(n=200)],the Radiolog-ical Hand Pose Estimation(RHPE)dataset[training set(n=5491),validation set(n=713),test set(n=79)],and a self-established dataset[training set(n=825),test set(n=351)],and it was validated using an external test set.Self-established dataset retrospectively recruited 1176 left-hand DR images of children from Peking Union Medical College Hospital(n=745,all were Han)and Tibet Autonomous Region Peoples Hospital(n=431,114 were Han,317 were Tibetan).External test set included images from Peoples Hospital of Nagqu(n=256,all were Tibetan).Mean absolute difference(MAD)and accuracy within 1 year were used as indicators.Results EVG-BANet exhibited MAD of 034 and 052 years in RSNA and RHPE test sets,respectively.In the self-established test set,the model achieved MAD of 047 years(95%CI:043-050)with accuracy within 1 year of 9772%(95%CI:9556-9901%).For the external test set,MAD was 053 years(95%CI:048-058),with accuracy within 1 year of 8945%(95%CI:8503-9293).Conclusion EVG-BANet demonstrated high accuracy in bone age prediction,and therefore can be applied in children living in both plain and highland.

关 键 词：骨龄 深度学习 人工智能 高原 藏族 

分 类 号：R445.4[医药卫生—影像医学与核医学] R72[医药卫生—诊断学] TP3[医药卫生—临床医学]