深度学习在骨科领域的应用研究进展  

Advances in the application of deep learning in orthopedics

作　　者：刘恩情 谭雪梅 蒋子月 黄成颉 张伟聪 吕海 苏志海 LIU Enqing;TAN Xuemei;JIANG Ziyue;HUANG Chengjie;ZHANG Weicong;LÜ Hai;SU Zhihai(Department of Spinal Surgery,The Third Affiliated Hospital of Southern Medical University,Guangzhou 510000,China;Department of Spinal Surgery,Fifth Affiliated Hospital of Sun Yat-sen University,Zhuhai 519000,China)

机构地区：[1]南方医科大学第三附属医院脊柱外科,广东广州510000 [2]中山大学附属第五医院脊柱外科,广东珠海519000

出　　处：《分子影像学杂志》2022年第1期151-156,共6页Journal of Molecular Imaging

基　　金：珠海市创新创业团队项目(ZH0406190031PWC)。

摘　　要：深度学习是当前人工智能发展最为迅速的一个分支。深度学习可以在大样本数据中自动提取良好的特征表达,有效提升各种机器学习的任务性能,广泛应用于图像信号处理、计算机视觉和自然语言处理等领域。随着数字影像的发展,深度学习凭借自动提取特征,高效处理高维度医学图像数据的优点,已成为医学图像分析在临床应用的重要技术之一。目前这项技术在分析某些医学影像方面已达到放射科医生水平,如肺结节的检出识别以及对膝关节退变进行级别分类等,这将为计算机科学发展在医疗应用的提供一个新机遇。由于骨科领域疾病种类繁多,图像数据特征清晰,内容复杂丰富,相关的学习任务与应用场景对深度学习提出了新要求。本文将从骨关节关键参数测量、病灶检测、疾病分级、图像分割以及图像配准五大临床图像处理分析任务对深度学习在骨科领域的应用研究进展进行综述,并对其发展趋势进行展望,以供从事骨科相关研究人员作参考。Deep learning is currently one of the most rapidly developing branches of artificial intelligence.Deep learning can automatically extract good feature expressions in large sample data,effectively improve the performance of various machine learning tasks,which is widely used in image signal processing,computer vision,and natural language processing.With the development of digital imaging,deep learning has become one of the essential techniques for the clinical application of medical image analysis with the advantages of automatically extracting features and efficiently processing high-dimensional medical image data.At present,this technology has reached the level of radiologists in analyzing certain medical images,such as detecting and recognizing lung nodules and classifying knee joint degeneration,which will provide a new chance for the development of computer science in medical applications.Due to the wide variety of diseases in the orthopedics field,clear image data features,and complex and rich contents in the orthopedic field,related learning tasks and application scenarios put forward a new challenge for deep learning.This article reviews the application research progress of deep learning in orthopedics from five clinical image processing and analysis tasks including bone and joint critical parameter measurement,lesion detection,disease grading,image segmentation,and image registration and outlook the development for the reference of orthopedics-related researchers.

关 键 词：深度学习 人工智能 卷积神经网络 骨科 医学图像处理分析 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TP391.41[自动化与计算机技术—控制科学与工程] R68[医药卫生—骨科学]