人工智能辅助诊断儿童DDH的研究进展

Research advances on the artificial intelligence-based imaging diagnosis of pediatric DDH

儿童发育性髋关节发育不良(developmental dysplasia of the hip,DDH)是导致髋关节骨关节炎和下肢残疾的重要原因,治疗难度和治疗效果与早期准确诊断密切相关.传统的诊断方法对尚未出现股骨头次级骨化中心者首选髋关节超声,对已出现者选择骨盆正位X线片;但两种方法均有赖于临床医生的手动测量与经验判断,耗时费力、可重复性差.人工智能技术与医学影像的有效整合可改善儿童DDH的诊疗现状,提升临床诊治效率.对4~6月龄内婴儿通过局部特征提取的分割算法、基于图像搜索的分割算法及深度学习网络等技术能够快速分析髋关节超声图像、测算DDH指标及辅助诊断DDH;对4~6月龄以上者利用骨边缘检测与模块匹配算法、深度迁移学习算法、同步挖掘局部及全局结构特征的卷积神经网络等技术自动识别骨性解剖关键点、计算髋关节参数及诊断儿童DDH.然而,由于技术所限及研究者认识不足,现有的儿童DDH辅助诊断工具在实际应用中面临着一些问题.通过文献检索从诊断可靠性及合理性等方面探讨儿童DDH人工智能影像学辅助诊断方法的研究进展,并为今后实现真正智能化的自动诊断工具提供研究思路.

Developmental dysplasia of the hip(DDH)is a common skeletal malformation in children and the prominent cause of hip osteoarthritis and lower limb disability.The therapeutic difficulty and effect of DDH are closely related to an early and proper diagnosis.Hip ultrasonography and anteroposterior pelvic radiography are preferred depending on the presence of the secondary ossification center of the femoral head.Conventional diagnostic methods primarily relied on manual measurements and empirical judgments by clinicians,which were laborious and generally lacked reliability.The effective integration of medical imaging and artificial intelligence algorithms is expected to improve the diagnosis of pediatric DDH and enhance the efficiency of clinical diagnosis and treatment.Segmentation algorithms based on the extraction of local geometric features,3D map search-based segmentation algorithms,and deep learning networks were utilized to assist in analyzing hip ultrasound images,calculating key dysplasia indicators,and diagnosing DDH in infants under 4-6 months.Computer-aided techniques,such as bone edge detection and template matching algorithms,deep transfer learning algorithms,and local-global feature mining convolutional neural networks were used to automatically identify bony landmarks on pelvic radiographs for measuring hip parameters and evaluating DDH in children over 4-6 months.However,there were several crucial problems in the clinical application of the artificial intelligence model for the auxiliary diagnosis of DDH due to technical limitations and insufficient understanding of researchers.This paper aims to review the progress of application in the medical artificial intelligence technology for the clinical auxiliary diagnosis of DDH.The author also provides references for future research for truly intelligent diagnostic tools.