早期膝骨关节炎影像学检查与深度学习融合应用

doi:10.3877/cma.j.issn.1674-134X.2026.01.012

本文旨在梳理早期膝关节退行性疾病（膝骨关节炎）影像检查技术的发展现状，以及这些技术与人工智能深度学习技术的融合应用情况，为临床诊断优化和相关学术研究提供参考。本文通过检索两个英文文献数据库，采用相关关键词收集相关研究文献，经剔除重复、关联性低、质量不达标的文献后，最终选取46篇高质量文献进行系统分析。临床常用的影像检查技术各有特点，可帮助发现疾病早期的部分病变特征，而部分新型影像技术能更精准地捕捉病变的细微变化；深度学习技术与这些影像检查结合后，在自动识别病变部位、判断疾病严重程度、预测病情发展趋势等方面展现出良好效果，显著提升了诊断的效率和准确性，但目前相关技术仍存在泛化能力不足、标准化程度有待提高等问题。早期膝关节退行性疾病的影像检查技术持续发展，与深度学习的融合为临床实现精准、高效的疾病评估提供了新路径，未来需进一步优化技术的稳定性和临床适配性，推动其更广泛地应用于临床实践。

关键词: 骨关节炎，膝, 诊断显像, 深度学习, 早期诊断

This review aimed to sort out the current development status of imaging examination techniques for early knee degenerative disease (knee osteoarthritis) and the integrated application of these techniques with artificial intelligence deep learning technology, so as to provide references for the optimization of clinical diagnosis and relevant academic research. By searching two English literature databases, relevant research literatures were collected using appropriate keywords; after excluding duplicate, low-relevance, and low-quality literatures, 46 high-quality literatures were finally selected for systematic analysis. Clinically common imaging examination techniques each have their own characteristics and can help detect some lesion features in the early stage of the disease, while some novel imaging techniques can capture subtle changes of lesions more accurately. The combination of deep learning technology with these imaging examinations has exhibited good performance in automatically identifying lesion locations, judging disease severity, and predicting disease progression trends, which significantly improves the efficiency and accuracy of diagnosis. However, the related technologies currently still have problems such as insufficient generalization ability and the need for improved standardization. Imaging examination techniques for early knee degenerative disease are continuously developing, and their integration with deep learning provides a new path for precise and efficient disease assessment in clinical practice. In the future, it is necessary to further optimize the stability and clinical adaptability of the technologies to promote their wider application in clinical practice.

Key words: Osteoarthritis, knee, Diagnostic imaging, Deep learning, Early diagnosis

图1 文献检索与筛选流程图

Figure 1 Flow chart for literature retrieval and screening

图2 KOA（膝骨关节炎）患者X线片。图A为双下肢全长站立位片，示双膝内翻，红色箭头示内侧关节间隙狭窄；图B为右膝负重侧位片，黄色箭头示骨端边缘骨赘形成，绿色箭头示软骨下骨硬化；图C为右膝负重正位片，蓝色箭头示髁间突变尖，红色箭头示内侧关节间隙狭窄

Figure 2 Radiographs of patients with KOA（knee osteoarthritis）. A is full-length anteroposterior radiograph of both lower extremities at standing position, showing genu varum of both knees with narrowing of the medial joint space (red arrows); B is lateral radiograph of the right knee at weight-bearing position, yellow arrows reveal osteophyte formation at the marginal ends of the bone and green arrow is pointing subchondral bone sclerosis; C is anteroposterior radiograph of the right knee at weight-bearing position, blue arrow demonstrates tapering of the intercondylar eminence and red arrow shows the medial joint space narrowing

图3 右膝MRI弥散加权成像（3.0T）。图A为矢状位T1WI图像，示软骨磨损和软骨下骨硬化（蓝色箭头），髌骨和骨端骨赘形成（红色箭头）；图B为矢状位T2WI图像，绿色箭头示内侧半月板后角撕裂（条状高信号影达关节面），黄色箭头示股骨内侧髁骨髓水肿（斑片状高信号影）；图C为冠状位T2WI图像，示前交叉韧带水肿显示不清（白色箭头），关节间隙狭窄（紫色箭头）；图D为轴位T2WI图像，示髌股关节对位异常、关节间隙宽窄不均（橙色箭头），局部可见异常高信号影，周围软组织信号欠均匀

Figure 3 MRI with diffusion-weighted imaging of the right knee (3.0T). A is sagittal image of T1WI, showing cartilage wear and subchondral bone sclerosis (blue arrow), as well as osteophyte formation at the patella and bony ends (red arrows); B is sagittal image of T2WI, green arrow showing a tear of the posterior horn of the medial meniscus (linear hyperintense foci extending to the articular surface), and yellow arrows showing bone marrow edema in the medial femoral condyle (patchy hyperintense foci); C is coronal image of T2WI, showing edema of the anterior cruciate ligament with poor visualization (white arrow), and the joint space narrowing (purple arrow); D is axial image of T2WI, showing patellofemoral joint malalignment with unequal joint space width (orange arrow), focal abnormal hyperintense foci are visible at the site, and the signal intensity of the peripheral soft tissues is inhomogeneous

图4 肌骨彩色多普勒超声，示肌骨区域的声像图、局部血流信号及血流频谱

Figure 4 Musculoskeletal color doppler ultrasound, showing the sonograms, local blood flow signals and blood flow spectra of the musculoskeletal region

图5 肌骨超声弹性成像图像注：左侧为肌骨区域灰阶声像图，右侧为对应区域的弹性成像图（以色彩区分组织硬度），并同步显示弹性质量指数等参数

Figure 5 Musculoskeletal ultrasound elastography imagesNote: the left part is grayscale sonogram of the musculoskeletal region, and the right part is elastography image of the corresponding regions (with tissue stiffness differentiated by color), which simultaneously display the elastic quality index and other parameters

[1]	Liang J, Wang Y, Yu F, et al. Evaluation of the osteoarthritis disease burden in China from 1990 to 2021: based on the Global Burden of Disease Study 2021[J/OL]. Front Public Health, 2024, 12: 1478710. DOI:10.3389/fpubh.2024.1478710.
[2]	Chen H, Wu J, Wang Z, et al. Trends and patterns of knee osteoarthritis in China: a longitudinal study of 17.7 million adults from 2008 to 2017[J/OL]. Int J Environ Res Public Health, 2021, 18（16）: 8864. DOI:10.3390/ijerph18168864.
[3]	Migliore A, Scirè CA, Carmona L, et al. The challenge of the definition of early symptomatic knee osteoarthritis: a proposal of criteria and red flags from an international initiative promoted by the Italian Society for Rheumatology[J]. Rheumatol Int, 2017, 37（8）: 1227-1236.
[4]	Zhao Z, Ito A, Kuroki H, et al. Analysis of molecular changes and features in rat knee osteoarthritis cartilage: progress from cellular changes to structural damage[J]. Cartilage, 2025, 16（2）: 232-249.
[5]	Takahashi I, Takeda K, Toyama T, et al. Histological and immunohistochemical analyses of articular cartilage during onset and progression of pre- and early-stage osteoarthritis in a rodent model[J/OL]. Sci Rep, 2024, 14（1）: 10568. DOI:10.1038/s41598-024-61502-8.
[6]	Ota S, Sasaki E, Sasaki S, et al. Relationship between abnormalities detected by magnetic resonance imaging and knee symptoms in early knee osteoarthritis[J/OL]. Sci Rep, 2021, 11（1）: 15179. DOI:10.1038/s41598-021-94382-3.
[7]	King LK, Liew JW, Mahmoudian A, et al. Multi-centre modified Delphi exercise to identify candidate items for classifying early-stage symptomatic knee osteoarthritis[J]. Osteoarthritis Cartilage, 2025, 33（1）: 155-165.
[8]	Mahmoudian A, Lohmander LS, Mobasheri A, et al. Early-stage symptomatic osteoarthritis of the knee—time for action[J]. Nat Rev Rheumatol, 2021, 17（10）: 621-632.
[9]	Piccolo CL, Mallio CA, Vaccarino F, et al. Imaging of knee osteoarthritis: a review of multimodal diagnostic approach[J]. Quant Imaging Med Surg, 2023, 13（11）: 7582-7595.
[10]	Mallio CA, Bernetti C, Agostini F, et al. Advanced MR imaging for knee osteoarthritis: a review on local and brain effects[J/OL]. Diagnostics, 2022, 13（1）: 54. DOI:10.3390/diagnostics13010054.
[11]	Jin M, Mi L, Li HW, et al. The value of super microvascular imaging and shear wave elastography in evaluating synovial lesions in knee osteoarthritis[J]. J Back Musculoskelet Rehabil, 2025, 38（6）: 1342-1351.
[12]	Tsuneki M. Deep learning models in medical image analysis[J]. J Oral Biosci, 2022, 64（3）: 312-320.
[13]	Cheung JC, Tam AY, Chan L-C, et al. Superiority of multiple-joint space width over minimum-joint space width approach in the machine learning for radiographic severity and knee osteoarthritis progression[J/OL]. Biology, 2021, 10（11）: 1107. DOI:10.3390/biology10111107.
[14]	Fan Z, Song W, Ke Y, et al. XGBoost-SHAP-based interpretable diagnostic framework for knee osteoarthritis: a population-based retrospective cohort study[J/OL]. Arthritis Res Ther, 2024, 26（1）: 213. DOI:10.1186/s13075-024-03450-2.
[15]	Li X, Roemer FW, Cicuttini F, et al. Early knee OA definition-what do we know at this stage? An imaging perspective[J/OL]. Ther Adv Musculoskelet Dis, 2023, 15: 1759720X231158204. DOI:10.1177/1759720X231158204.
[16]	Jerban S, Chang EY, Du J. Magnetic resonance imaging （MRI） studies of knee joint under mechanical loading: Review[J]. Magn Reson Imaging, 2020, 65: 27-36.
[17]	Ehmig J, Engel G, Lotz J, et al. MR-imaging in osteoarthritis: current standard of practice and future outlook[J/OL]. Diagnostics, 2023, 13（15）: 2586. DOI:10.3390/diagnostics13152586.
[18]	Oei EHG, Runhaar J. Imaging of early-stage osteoarthritis: the needs and challenges for diagnosis and classification[J]. Skeletal Radiol, 2023, 52（11）: 2031-2036.
[19]	Panwar P, Chaurasia S, Gangrade J, et al. Optimizing knee osteoarthritis severity prediction on MRI images using deep stacking ensemble technique[J/OL]. Sci Rep, 2024, 14（1）: 26835. DOI:10.1038/s41598-024-78203-x.
[20]	Nelson AE. Turning the page in osteoarthritis assessment with the use of ultrasound[J/OL]. Curr Rheumatol Rep, 2020, 22（10）: 66. DOI:10.1007/s11926-020-00949-w.
[21]	Wu J, Li Y, Zhang X, et al. Assessment of blood flow around the knee joint in patients with knee osteoarthritis by color Doppler ultrasound[J/OL]. Eur J Radiol, 2023, 166: 111005. DOI:10.1016/j.ejrad.2023.111005.
[22]	Zhang H, Ning E, Lu L, et al. Research progress of ultrasound in accurate evaluation of cartilage injury in osteoarthritis[J/OL]. Front Endocrinol, 2024, 15: 1420049. DOI:10.3389/fendo.2024.1420049.
[23]	Zhang J, Zheng X, Zhao Y. Analysis of the main indicators and risk factors of ultrasonic shear wave elastography for the diagnosis of osteoarthritis among adults[J/OL]. Front Med, 2024, 11: 1366793. DOI:10.3389/fmed.2024.1366793.
[24]	Li F, Wang ZY, Zhang ZJ, et al. In hamstring muscles of patients with knee osteoarthritis an increased ultrasound shear modulus indicates a permanently elevated muscle tonus[J/OL]. Front Physiol, 2022, 12: 752455. DOI:10.3389/fphys.2021.752455.
[25]	Liew JW, King LK, Mahmoudian A, et al. A scoping review of how early-stage knee osteoarthritis has been defined[J]. Osteoarthritis Cartilage, 2023, 31（9）: 1234-1241.
[26]	Cui XW, Li KN, Yi AJ, et al. Ultrasound elastography[J]. Endosc Ultrasound, 2022, 11（4）: 252-274.
[27]	Chalian M, Li X, Guermazi A, et al. The QIBA profile for MRI-based compositional imaging of knee cartilage[J]. Radiology, 2021, 301（2）: 423-432.
[28]	El-Ghany SA, Elmogy M, Abd El-Aziz AA. A fully automatic fine tuned deep learning model for knee osteoarthritis detection and progression analysis[J]. Egypt Inform J, 2023, 24（2）: 229-240.
[29]	Araújo ALD, Sperandio M, Calabrese G, et al. Artificial intelligence in healthcare applications targeting cancer diagnosis-part I: data structure, preprocessing and data organization[J]. Oral Surg Oral Med Oral Pathol Oral Radiol, 2025, 140（1）: 79-88.
[30]	Liao Y, Yang G, Pan W, et al. OA-HybridCNN （OHC）: an advanced deep learning fusion model for enhanced diagnostic accuracy in knee osteoarthritis imaging[J/OL]. PLoS One, 2025, 20（5）: e0322540. DOI:10.1371/journal.pone.0322540.
[31]	Abdusalomov A, Mirzakhalilov S, Umirzakova S, et al. Light weight early detection of knee osteoarthritis in athletes[J/OL]. Sci Rep, 2025, 15: 31413. DOI:10.1038/s41598-025-04095-0.
[32]	Ren X, Hou L, Liu S, et al. OA-MEN: a fusion deep learning approach for enhanced accuracy in knee osteoarthritis detection and classification using X-Ray imaging[J/OL]. Front Bioeng Biotechnol, 2025, 12: 1437188. DOI:10.3389/fbioe.2024.1437188.
[33]	Kijowski R, Fritz J, Deniz CM. Deep learning applications in osteoarthritis imaging[J]. Skeletal Radiol, 2023, 52（11）: 2225-2238.
[34]	Wang Y, Wang X, Gao T, et al. An automatic knee osteoarthritis diagnosis method based on deep learning: data from the osteoarthritis initiative[J/OL]. J Healthc Eng, 2021, 2021: 5586529. DOI:10.1155/2021/5586529.
[35]	Liu B, Luo J, Huang H. Toward automatic quantification of knee osteoarthritis severity using improved Faster R-CNN[J]. Int J Comput Assist Radiol Surg, 2020, 15（3）: 457-466.
[36]	Yoon JS, Yon C-J, Lee D, et al. Assessment of a novel deep learning-based software developed for automatic feature extraction and grading of radiographic knee osteoarthritis[J/OL]. BMC Musculoskelet Disord, 2023, 24（1）: 869. DOI:10.1186/s12891-023-06951-4.
[37]	Swiecicki A, Li N, O’Donnell J, et al. Deep learning-based algorithm for assessment of knee osteoarthritis severity in radiographs matches performance of radiologists[J/OL]. Comput Biol Med, 2021, 133: 104334. DOI:10.1016/j.compbiomed.2021.104334.
[38]	Li W, Xiao Z, Liu J, et al. Deep learning-assisted knee osteoarthritis automatic grading on plain radiographs: the value of multiview X-ray images and prior knowledge[J]. Quant Imaging Med Surg, 2023, 13（6）: 3587-3601.
[39]	Tiulpin A, Saarakkala S. Automatic grading of individual knee osteoarthritis features in plain radiographs using deep convolutional neural networks[J/OL]. Diagnostics, 2020, 10（11）: 932. DOI:10.3390/diagnostics10110932.
[40]	Martel-Pelletier J, Pelletier J-P. Next-level prediction of structural progression in knee osteoarthritis: a perspective[J/OL]. Int J Mol Sci, 2025, 26（10）: 4748. DOI:10.3390/ijms26104748.
[41]	Guan B, Liu F, Haj-Mirzaian A, et al. Deep learning risk assessment models for predicting progression of radiographic medial joint space loss over a 48-MONTH follow-up period[J]. Osteoarthritis Cartilage, 2020, 28（4）: 428-437.
[42]	Hu J, Zheng C, Yu Q, et al. Deep KOA: a deep-learning model for predicting progression in knee osteoarthritis using multimodal magnetic resonance images from the osteoarthritis initiative[J]. Quant Imaging Med Surg, 2023, 13（8）: 4852-4866.
[43]	Si L, Zhong J, Huo J, et al. Deep learning in knee imaging: a systematic review utilizing a Checklist for Artificial Intelligence in Medical Imaging （CLAIM）[J]. Eur Radiol, 2022, 32（2）: 1353-1361.
[44]	Ahmed SM, Mstafa RJ. A comprehensive survey on bone segmentation techniques in knee osteoarthritis research: from conventional methods to deep learning[J/OL]. Diagnostics, 2022, 12（3）: 611. DOI:10.3390/diagnostics12030611.
[45]	Martel-Pelletier J, Paiement P, Pelletier J-P. Magnetic resonance imaging assessments for knee segmentation and their use in combination with machine/deep learning as predictors of early osteoarthritis diagnosis and prognosis[J/OL]. Ther Adv Musculoskelet Dis, 2023, 15: 1759720X231165560. DOI:10.1177/1759720X231165560.
[46]	Joseph GB, McCulloch CE, Nevitt MC, et al. Machine learning models for clinical and structural knee osteoarthritis prediction: recent advancements and future directions[J/OL]. Osteoarthr Cartil Open, 2025, 7（3）: 100654. DOI:10.1016/j.ocarto.2025.100654.

[1]	江瑶, 蒋程, 余翔, 谭莹, 温昕, 温慧莹, 彭桂艳, 李胜利. 基于注意力机制改进的子宫解剖结构检测与分割多任务模型的性能评估[J/OL]. 中华医学超声杂志（电子版）, 2025, 22(08): 703-710.
[2]	中华医学会骨科学分会关节外科学组, 王坤正, 史占军, 曲铁兵. 中国膝骨关节炎关节腔注射治疗专家共识（2026版）[J/OL]. 中华关节外科杂志(电子版), 2026, 20(01): 1-14.
[3]	李亚鹏, 岳辰, 李峰, 张志杰, 成彦, 郭珈宜. 膝骨关节炎膝周肌肉硬度与功能的相关性评估[J/OL]. 中华关节外科杂志(电子版), 2026, 20(01): 32-40.
[4]	刘肖雅, 郭珈宜, 成彦, 李峰, 杨艳霞, 高起, 杨玉霞, 岳辰. 虚拟现实运动在膝骨关节炎患者中应用效果的Meta分析[J/OL]. 中华关节外科杂志(电子版), 2026, 20(01): 50-59.
[5]	杨浩, 王春岭, 席少华, 尹素然. 富血小板血浆联合关节镜手术治疗膝骨关节炎的疗效分析[J/OL]. 中华损伤与修复杂志(电子版), 2025, 20(06): 499-504.
[6]	李婷, 郭超, 李晨曦, 李柔演, 张一涵, 张雨晗, 陈琰, 斯琴高娃, 龚忠诚. 基于深度学习的上颌阻生尖牙自动分割与体积量化研究[J/OL]. 中华口腔医学研究杂志(电子版), 2026, 20(01): 9-16.
[7]	孟笑菲, 龚卓弘, 万耀文, 李沛达, 胡琪琪, 邱龙诗语, 刘恒毅, 谢伟丽. 深度学习与多模态大语言模型在笑线分类任务中的应用初探[J/OL]. 中华口腔医学研究杂志(电子版), 2026, 20(01): 17-24.
[8]	吕思怡, 王琰琪, 仇珺, 陈宇江, 高洁. 人工智能图像处理技术在口腔医学中的应用[J/OL]. 中华口腔医学研究杂志(电子版), 2026, 20(01): 40-46.
[9]	杨雯林, 吴元魁. 影像组学在胰腺神经内分泌瘤诊疗中的研究进展[J/OL]. 中华普通外科学文献(电子版), 2025, 19(06): 426-432.
[10]	薛兆强, 袁寅. 双镜联合保功能胃癌根治术治疗早期近端胃癌的临床研究[J/OL]. 中华普外科手术学杂志(电子版), 2025, 19(06): 628-632.
[11]	梅昊楠, 杨瑞, 刘修恒. 人工智能辅助病理学图像分析在前列腺癌诊断中的研究进展[J/OL]. 中华腔镜泌尿外科杂志(电子版), 2026, 20(01): 1-7.
[12]	胡博文, 胡亚兰, 梁辉. 前列腺癌早期筛查的常见方法及最新研究进展[J/OL]. 中华腔镜泌尿外科杂志(电子版), 2025, 19(06): 800-808.
[13]	唐玥, 陈家璐, 覃德龙, 李宗龙, 汤朝晖, 全志伟. 基于AI的多模态影像在肝癌诊治中应用及面临挑战[J/OL]. 中华肝脏外科手术学电子杂志, 2026, 15(01): 4-9.
[14]	彭单单, 汪敏, 李君, 曾冲, 曾玲, 胡玉莲. 脑血流动力学参数联合血清hs-CRP、IL-6水平对早产儿脑损伤的早期诊断价值[J/OL]. 中华脑科疾病与康复杂志(电子版), 2025, 15(06): 352-357.
[15]	薛怡宁, 兰雅迪, 刘兆宇, 史磊, 赵琪, 许洪伟. 基于图像的人工智能在胃癌中的研究进展[J/OL]. 中华消化病与影像杂志(电子版), 2025, 15(06): 670-675.

阅读次数

全文

摘要

选择文件类型/文献管理软件名称

选择包含的内容

Integrated applications of imaging examinations and deep learning for early knee osteoarthritis

模态框（Modal）标题

选择文件类型/文献管理软件名称

选择包含的内容

Integrated applications of imaging examinations and deep learning for early knee osteoarthritis