分享:
分享到微信朋友圈
X
综述
人工智能联合影像技术在转移性颈部淋巴结检测中的应用进展
陈莉军 王冰 王琳

本文引用格式:陈莉军, 王冰, 王琳. 人工智能联合影像技术在转移性颈部淋巴结检测中的应用进展[J]. 磁共振成像, 2025, 16(11): 209-215. DOI:10.12015/issn.1674-8034.2025.11.032.


[摘要] 转移性颈部淋巴结(metastatic cervical lymph nodes, MCLN)在众多头颈部肿瘤的诊断分期和临床决策制订中至关重要。传统的CT、MRI、PET-CT等影像学检查虽普遍应用于临床中,但在精准区分MCLN的敏感性和特异性较差。人工智能(artificial intelligence, AI)近两年迅速发展,尤其是深度学习(deep learning, DL)在医学影像分析方面取得了突破性成就。本文对不同模态影像技术联合AI(CT增强配以自动分割、MRI高软组织对比结合自动分割、PET-CT代谢图像融合AI模型、超声结合DL的实时自动辅助诊断等)在头颈部MCLN的最新研究进行概括性综述,阐述了AI在MCLN诊断和疗效和预后评估中的应用,总结目前研究中存在的不足及技术困境,并提出未来的发展方向。本综述旨在为未来研究协同、模型优化及临床应用提供参考。
[Abstract] Metastatic cervical lymph nodes (MCLN) are crucial in the diagnosis, staging, and clinical decision-making processes for various head and neck tumors. Despite the widespread use of conventional imaging modalities such as computed tomography (CT), magnetic resonance imaging (MRI), and positron emission tomography-computed tomography (PET-CT) in clinical practice, their sensitivity and specificity in accurately identifying all instances of MCLN remain suboptimal. In recent years, artificial intelligence (AI), and deep learning (DL) in particular, have made significant advancements in the field of medical image analysis. This review provides a comprehensive review of the latest research on the combined use of different modal imaging techniques and AI (CT enhancement combined with automatic segmentation, MRI high soft tissue contrast combined with automatic segmentation, PET-CT metabolic image fusion AI model, ultrasound combined with DL for real-time automatic auxiliary diagnosis, etc.) in head and neck MCLN. It elaborates on the application of AI in the diagnosis, therapeutic effect and prognosis assessment of MCLN, summarizes the existing shortcomings and technical challenges in current research, and proposes future development directions. This review aims to provide a reference for future research collaboration, model optimization and clinical application.
[关键词] 头颈部肿瘤;淋巴结转移;人工智能;深度学习;影像技术;磁共振成像
[Keywords] head and neck neoplasms;lymph node metastasis;artificial intelligence;deep learning;imaging technology;magnetic resonance imaging

陈莉军 1, 2   王冰 3   王琳 4*  

1 甘肃中医药大学第一临床医学院,兰州 730000

2 甘肃省人民医院放射科,兰州 730050

3 白银市第一人民医院骨科,白银 730900

4 甘肃中医药大学附属医院放射科,兰州 730000

通信作者:王琳,E-mail:tedyong@163.com

作者贡献声明:王琳统筹综述的整体构思与方向,进行文献分析与内容归纳,对全文进行关键性修改;陈莉军起草和撰写稿件,进行文献的收集、分析、整理;王冰进行文献整理与分析,对稿件重要内容进行了修改;王冰获得甘肃省科技计划项目资助,陈莉军获得甘肃省人民医院院内项目资助;全体作者都同意发表最后的修改稿,同意对本研究的所有方面负责,确保本研究的准确性和诚信。


基金项目: 甘肃省科技计划项目 25JRRD002 甘肃省人民医院院内项目 20GSSY3-9
收稿日期:2025-06-08
接受日期:2025-09-30
中图分类号:R445.2  R739.91 
文献标识码:A
DOI: 10.12015/issn.1674-8034.2025.11.032
本文引用格式:陈莉军, 王冰, 王琳. 人工智能联合影像技术在转移性颈部淋巴结检测中的应用进展[J]. 磁共振成像, 2025, 16(11): 209-215. DOI:10.12015/issn.1674-8034.2025.11.032.

0 引言

       转移性颈部淋巴结(metastatic cervical lymph nodes, MCLN)的检出是影响原发恶性肿瘤患者治疗决策与预后的关键因素之一[1]。头颈癌(headand neck carcinoma, HNC),包括口腔、咽部和喉部等的恶性肿瘤,具有侵袭性和局部淋巴结转移(lymph node metastasis, LNM)的特征[2]。在头颈部及其他部位原发恶性肿瘤患者中,MCLN的识别可通过多种影像学技术实现,为疾病的分期与治疗决策提供重要依据。其中,计算机断层扫描(computed tomography, CT)、磁共振成像(magnetic resonance imaging, MRI)、超声及正电子发射断层扫描-计算机断层扫描(positron emission tomography-computed tomography, PET-CT)等是常见的检测工具[3]。在传统诊疗模式中,影像学检查结果的解读主要依赖放射科医生或临床医师的阅片经验,以获取诊断信息并指导治疗决策。然而,该过程高度依赖专业人员的知识储备和主观判断,在一致性和准确性方面存在一定的局限,同时解读过程通常耗时较长,加重临床工作负担。AI模型的出现实现了自动的计算机图像分析、特征提取及定量分析,在辅助提高病变定位和检测的敏感性和特异性、减少误差、提高诊断的一致性方面具有一定优势[4, 5]

       虽然已有部分综述[6, 7]涉及AI与医学图像的关系,但缺乏对AI与多种医学图像组合应用于MCLN检测的系统性评价和综述。本综述从AI联合CT、MRI、PET-CT、超声成像应用在MCLN鉴别中的研究报道入手,系统归纳不同模型的特征及其评价指标,评述其临床意义,并探讨当前存在的技术问题与未来的发展趋势。本综述旨在通过总结最新研究进展,揭示方法学优势与不足,为后续跨学科研究与临床探索提供参考。

1 聚焦于MCLN检测的AI概述

       AI泛指计算机模仿人类大脑解决复杂问题的方法[8, 9],近年来被广泛应用于医学影像领域。尤其是深度学习(deep learning, DL)技术,在MCLN检测和识别方面的相关研究已初见成效。相比传统影像的经验式诊断,AI能够通过标准化的量化分析提升诊断结果的统一性,并通过深度特征建模提高与真实病理结果的符合度,从而实现更高的精确性[10, 11]

       卷积神经网络(convolutional neural network, CNN)是最常用的医学图像分析DL框架之一,通过卷积和非线性变换等方法逐层变换图像空间语义特征,在图像分类、目标检测和语义分割等领域得到应用[12]。在头颈部的影像学研究中,CNN可自动定位CT[13]、MRI[14]或PET-CT[15]图像中的病灶和可疑的淋巴结,减少人工阅读时间及个人主观偏差。U-Net[16]是基于对称编码-解码结构的神经网络经典结构之一,对进行多尺度特征提取和融合,适用于小样本医学图像的像素级分割。U-Net能够精准勾画病灶边界,在复杂解剖结构、组织重叠或病变对比度低的情况下获得较高敏感度与特异度,attentionU-Net[17]、ResUNet[18]等变体提升了U-Net对多尺度信息的利用和上下文信息的建模。残差网络(residual network, ResNet)结构通过加入残差连接缓解了深层网络存在的梯度消失问题,能够构建更为深层的网络模型,并提升了其表征能力和分类精确性,在影像特征提取、多模态融合辅助诊断中被广泛应用于复杂异质性肿瘤结构判别等过程中,具有稳定的分类性能和泛化性能[19]

       此外,AI模型可用以结合影像组学对多模态影像(PET-CT融合图像)提取的高维度纹理、形态及功能特征进行建模,刻画肿瘤的异质性,达到分层病灶、风险评分及预测患者预后的目的[20, 21]。由AI辅助引导的影像组学流程能够识别人眼无法识别的微转移灶,能够提高淋巴结早期受累检出。

       相较于传统依赖人工经验的特征提取流程,DL模型具备端到端自动学习的优势,能显著降低前期预处理与手工特征设计的复杂度,同时,研究表明,ResNet、UNet及其变体在不同中心的PET-CT或CT数据集上分割表现稳定,AUC或Dice值较高,提示其具有一定的迁移性与泛化能力[22]。从而为其在头颈癌患者的临床分型、术前分期、疗效评估与预后判断等环节的应用奠定了方法学基础,展现出广阔的发展前景。

2 AI联合CT技术在MCLN检测中的应用

       CT是评估大多数MCLN时的首选影像学方法[23],在诊断LNM中表现优良,敏感度和特异度分别可达84.9%与91.5%[24]。常规CT通过形态特征来区分转移性和非转移性LN,双能CT(dual-energy computed tomography, DECT)[25]作为功能成像的新兴手段,为MCLN的检测提供了有益的补充。碘浓度(iodine concentration, IC)和归一化碘浓度(normalized iodine concentration, NIC)是源自碘图谱的常用DECT的参数,已被广泛用于检测恶性肿瘤患者的LNM[26]。多项研究表明,DECT在检测MCLN方面具有良好的性能[27]

       甲状腺乳头状癌(papillary thyroid carcinoma, PTC)占甲状腺癌(thyroid carcinoma, TC)的90%[28]。LNM发生在60%~70%的PTC患者中[29],这与术前局部复发和癌症特异性死亡率高度相关[30]。PTC患者的局部区域复发风险为15%~30%[31]。上述证据都提倡对潜在的LNM进行严格的术前筛查,这有助于选择适当的手术策略。ONOUE等[32]基于ResNet-101架构,采用迁移学习策略构建深度学习模型,用于MCLN的影像学分析,识别增强CT图像中三类囊性淋巴结。通过图像裁剪、增强与五折交叉验证提升模型稳健性,准确率达76%。WANG等[33]构建AI系统融合CT影像、影像组学与临床特征预测PTC患者MCLN,曲线下面积(area under the curve, AUC)达0.84/0.81(内外部测试),优于单一模型。该研究凸显多模态特征融合(形态+影像组学+临床)的策略在模型构建中的方法学价值。JIN等[34]基于DL融合DECT多参数特征预测PTC患者的MCLN,分别提取动脉期常规单能量CT图像、静脉期常规单能量CT 图像、动脉期碘基物质分解图像、静脉期碘基物质分解图像及常规CT特征建立多模型,并在多中心数据上进行验证。结果显示,组合模型整合多参数特征后性能显著优于单一模型,开发集与独立测试集的AUC分别达到0.890和0.865,体现出光谱参数与DL融合在提升诊断性能及临床应用潜力方面的优势。上述研究表明,AI算法结合CT、增强CT和DECT图像可能有助于检测PTC患者CT图像的LNM,并且显示出了良好的性能。

       头颈部鳞状细胞癌(head and neck squamous cell carcinoma, HNSCC)是一类常见的恶性肿瘤,具有侵袭性生长和局部LNM的倾向。部分患者在疾病累及颈部区域淋巴结时被确诊,淋巴结受累是其重要的不良预后因素之一[35, 36]。准确分期,尤其是对转移性淋巴结的识别与评估,对于制订最佳治疗方案至关重要。BARDOSI等[37]结合增强CT与影像组学的消除特征选择算法,对HNSCC患者252个颈部淋巴结提取超过3万特征,实现约90%分类准确率。该方法凸显高维特征筛选与影像组学融合在病理性淋巴结精准分类中的优势。

       虽然CT、强化CT和DECT等结合AI辅助手段可减少人为阅片偏差,提高淋巴结小病灶的检出能力,但还存在以下不足:一方面,针对淋巴结小结节的诊疗,目前多集中在术前诊断和风险分级上,而对其术后的观察随访、治疗效果的监测以及全程管理的应用尚少;另一方面,DECT等功能参数虽展现出一定增量价值,但其在小病灶、微转移检出中的敏感性仍有待进一步验证。未来研究可重点关注以下方面:第一,扩大AI在术后监测、效果评价及复发预估等环节的应用,促进其介入到头颈肿瘤MCLN“诊断—治疗—随访”全病程管理;第二,继续细化完善以及验证DECT功能参数提取以及AI模型,并在多中心进行验证,明确其在小病灶、微转移的检测应用价值。总体而言,基于AI的CT类影像在MCLN检出中具有不错应用前景,但有待多层面改良方能更好地转化并推广应用至精准医疗领域中。

3 AI联合MRI技术在MCLN检测中的应用

       MRI能够提供极高的软组织对比度[38]。在多种MRI技术中,钆对比剂增强MRI可显示异常血管模式,间接反映病理变化[39]。扩散加权成像(diffusion-weighted imaging, DWI)及表观扩散系数(apparent diffusion coefficient, ADC)可通过反映水分子扩散受限情况区分良恶性淋巴结,转移性病灶常表现为ADC值降低;动态对比增强MRI(dynamic contrast-enhanced magnetic resonance imaging, DCE-MRI)可量化血管通透性与灌注特征[40],磁共振波谱(magnetic resonance spectroscopy, MRS)能够检测提示恶性病变的代谢谱变化。AI技术与MRI结合,可在多参数影像中自动提取高维形态与功能特征,融合DWI、ADC、DCE及MRS信息实现定量化与标准化分析,不仅提升MCLN的检出率与分型精度,还能减少人工判读的主观差异,为个体化诊疗和预后评估提供坚实依据。

       ZHOU等[39]基于CNN融合MRI开发了颈部淋巴结检测与分割模型,采用多尺度特征提取与端到端训练,平均精度(mean average precision, mAP)达74.64%,准确率达83.6%、AUC达0.834。该方法集成形态与位置特征,兼顾检测与分割性能,提升头颈部恶性肿瘤淋巴结诊断精度。YUAN等[41]应用T1WI与T2WI提取了16个纹理特征,比较了6种机器学习(machine learning, ML)方法预测早期口腔舌鳞状细胞癌(oral tongue squamous cell cancer, OTSCC)隐匿性MCLN。朴素贝叶斯结合10折交叉验证识别准确率达74.1%,表现最佳。该方法体现了多模态特征融合与ML算法筛优的预测优势。WANG等[42]基于MRI影像组学特征融合ADC值与淋巴结尺寸构建CLN预测模型,AUC达0.83。该方法结合形态、功能及定量特征,实现多源信息互补,提升颈部淋巴结恶性风险评估的准确性与稳定性。WANG等[43]利用MRI影像组学特征预测舌癌LNM,CRprim+10模型在训练集与测试集AUC分别达0.995与0.872。特征与病理DOI显著相关,体现影像表型与组织学指标的关联性,提升术前风险分层与精准治疗决策能力。WANG等[44]构建MRI影像组学与DL联合模型预测舌鳞癌MCLN,筛选9个影像组学与3个DL特征。影像组学中多层感知器模型AUC最高(0.747),DL中逻辑回归AUC为0.655,显示特征融合较单一DL具更优预测能力。LIU等[45]提出的多模态MRI影像组学ML模型预测OTSCC患者MCLN,比较6种序列组合。最佳模型(T1WI、FS-T2WI、T2WI、CE-MRI)在两组的AUC分别为0.881与0.868。结果提示序列增加未必显著提升性能,强调序列优化优于单纯堆叠。

       目前AI与MRI检测MCLN多表现为提高敏感度、特异度,尤其是功能成像(包括DWI,ADC,DCE,MRS)和智能分析(包括影像组学、DL等)结合的应用;CNN+ML与MRI的结合,已经在诊断能力上取得了一定进展,且显示出多参数的特征提取能力[46]。上述研究也均存在一定的局限性:影像采集与特征选取不统一,结果存在显著差异,没有可比性与重复性;差异的诊断性能也表明序列优化优于单纯堆叠,仍然需要进一步系统评估[47]。下一步的研究应该重点在发挥MRI功能成像优势的基础上,尽力做影像采集与特征提取标准化,通过多中心的外推验证提升稳健度[48];加大正则化与扩充样本等提高泛化能力;加强影像组学、DL与临床以及分子生物标记物结合提高预测性能;同时也引入解释性AI方法,并在前瞻性临床试验中验证临床实用性[49]

4 AI联合超声技术在MCLN检测中的应用

       超声因其非侵入性、实时成像及成本效益,仍是评估MCLN的重要手段[50]。然而,传统超声依赖操作经验,敏感度与重复性存在差异。近年来,高分辨率超声(high-resolution ultrasound, HRUS)、对比增强超声(contrast-enhanced ultrasound, CEUS)及超声弹性成像(UltrasoundElastography, USE)显著提升了淋巴结的空间分辨率、血流灌注评估及组织硬度测量能力[51]。在此基础上,AI技术通过自动特征提取与模式识别,可减少人为因素影响,提升多参数超声在MCLN检测中的准确性与稳定性,为头颈肿瘤分期与治疗决策提供更可靠的影像学依据。

       ZHOU等[52]基于多机构超声数据,构建融合甲状腺结节影像与临床变量的深度CNN模型预测CLNM。经多机构内部、外部验证,AUC值分别为0.86、0.77,准确度、敏感度与特异度均优于放射科医生平均水平,体现了多模态特征融合与跨中心泛化优势。CHANG等[53]基于PTC患者多中心数据,结合DL提取的超声特征与临床变量,构建多变量逻辑回归列线图预测CLNM。纳入预测因子包括模型预测值、倍数、位置、微钙化、长短径比及US报告淋巴结状态。模型在训练、内部及外部验证集的AUC分别为0.812、0.809和0.829,表现出稳健的跨中心泛化能力。WANG等[54]收集488例细针穿刺确诊PTC患者的临床与病理资料,基于多特征集成的CNN预测模型评估CLNM风险。模型在训练集与测试集AUC分别为0.89和0.78;在结节<1 cm亚组中,AUC分别为0.87和0.76,显示出在微小病灶预测中的稳定性与适用性。

       AI和超声成像在对患者的颈部LNM,特别是PTC患者检测方面有一定程度的改善。上述研究证明在预测PTC患者MCLN检测方面的性能精度、敏感度和特异度方面,AI均优于放疗科医师,且AI模型在训练集或验证集之间的预测性能比较稳定,提示其有着良好的推广价值和临床应用价值[55, 56]。但目前研究也存在一定局限性:目前大多数研究均在PTC进行,缺少其他头颈部恶性肿瘤的系统性验证;而且尽管在多模态信息整合上得到广泛应用,可以提高预测性能,但如何对不同特征的差异进行科学配比融合尚无定论。下一步的研究可以将AI的范围扩展到更多头颈部恶性肿瘤类型上,并进一步改进多模态特征融合策略;应推动模型的简化和轻量化,使AI辅助超声诊断的便携式超声设备能在资源缺乏地区得以应用,提高AI辅助超声诊断在世界的推广与应用前景。

5 AI联合PET-CT技术在MCLN检测中的应用

       PET-CT结合代谢功能与解剖结构信息,是评估头颈部恶性肿瘤(尤其是鳞状细胞癌与PTC)MCLN的关键影像学工具[57]。近年来,PET-CT在检测小体积病灶、识别微小代谢异常及多灶病变方面表现出较高的敏感性与特异性。AI技术的引入,可在全身代谢影像中实现自动化病灶分割、SUV定量及异质性分析,提升MCLN检测的准确性与效率,并有助于风险分层与预后预测,为精准分期和个体化治疗提供更为坚实的影像支持。

       MYRONENKO等[58]基于多阶段ResNet的SegResNet模型对3D PET-CT头颈部肿瘤及淋巴结进行自动分割,采用统一重采样、区域裁剪、5折交叉验证及多模型融合策略,在多中心数据集上实现平均Dice系数0.798 9,分割精度居HECKTOR2022首位。SUN等[59]基于CNN模型对CT与PET-CT中标注肿瘤及转移LN的ROI进行分割训练,比较EF-60、EF-90、nnUNet与nnFormer的性能,LN分割Dice系数分别达0.832 4、0.833 5、0.832 5与0.830 4,体现了多网络结构在跨模态淋巴结分割中的稳定性与高精度。JIANG等[60]基于nnU-Net对头颈部肿瘤合并MCLN的CT/PET-CT图像进行自动分割,第三次试验在采用标准化数据及后处理优化后,mDice值提升至0.769 38,显示了模型在多模态融合与后处理策略结合下的精度优势。REBAUD等[61]基于全自动nnU-Net架构对CT/PET-CT图像中的肿瘤原发灶总体积(Gross Tumor Volume of the primary tumor, GTVp)与转移淋巴结总体积(gross tumor volume of the metastatic lymph node, GTVn)进行分割,并采用交叉验证评估模型稳健性,其中GTVn的Dice系数达到0.789,体现了模型在多解剖结构精确分割中的稳定性能。SALAHUDDIN等[62]基于3D-UNet架构对头颈部肿瘤合并LNM患者的PET-CT图像进行五折交叉验证分割,在综合形态与代谢信息的自动分割任务中mDice系数达0.768 0,体现了3D卷积在复杂多模态数据融合及空间特征提取方面的优势。WANG等[63]构建nnU-Net与Transfiner分割模型,对头颈部肿瘤患者PET-CT及增强CT图像进行肿瘤及转移淋巴结分割,经五折交叉验证,mDice系数分别达0.769 7和0.765 2,显示两种架构在多模态影像自动分割中的稳健性能与可比精度。JAIN等[64]对PET-CT图像进行预处理,构建并评估3DU-Net、nnM-Net、SwinUNet及nnU-Net(2D/3D)等DL框架,实现颈部GTVp与GTVn的自动分割。结果显示:3DnnU-Net、nnM-Net及SwinUNet的mDice系数分别为0.740、0.698和0.669,体现3DnnU-Net在多模态分割中的优越性。KUDOH等[65]基于18F-FDG PET影像,利用LASSO回归构建CLNM预测模型。通过特征筛选与建模,AUC达0.79、准确率达68%、敏感度达65%、特异度达70%,体现了PET代谢信息在模型预测中的价值。

       上述各种研究表明,基于ResNet、nnU-Net、3D-UNet及基于Transfiner等的DL模型与PET-CT成像手段相结合,在自动分隔头颈部肿瘤及相应转移性淋巴结方面,平均Dice值在0.74~0.83,模型在不同数据集内的交叉验证结果表现相对一致,具有较为稳定的鲁棒性及潜在的临床实际运用价值。特别是nnU-Net及其变体因为模型可自适应性地、稳定地输出,在该方向被应用得最广[66]。但现有研究仍存在一定不足:不同解剖位置及不同患者间分割精度差异显著,易受PET-CT图像质量、手动标注差异、肿瘤大小及代谢活性不同等因素影响;同时,图像融合仍是PET与CT联合应用的一大难点,尤其在病灶与邻近组织边界模糊时,模型鉴别存在困难。今后工作可从以下方面开展:(1)开发更加有效的多模态融合网络,最大程度学习PET和CT的互补关系,如利用注意力机制或图卷积网络;(2)寻找基于PET-CT纵向随访影像的动态分析及疗效预判模型,为MCLN患者疾病发展的全过程提供帮助。

6 小结与展望

       AI特别是DL和影像组学在提高MCLN检出、分割方面具有较高的应用潜力,可显著提高检测的准确度、效率及可重复性。在多种影像模式(超声、MRI、CT、PET-CT)上应用AI模型的研究中取得了良好的表现。同时,将形态、功能和代谢成像特征结合的多模态融合方法也逐渐受到关注,并取得了积极效果。

       但现阶段AI技术应用于临床还需要突破以下几大障碍:其一,目前多数研究仍基于回顾性、单中心、有限样本,跨中心、跨设备的泛化能力不足;其二,不同研究在影像采集、特征提取与模型构建上缺乏统一标准,影响结果的可比性与临床推广;其三,当前DL模型普遍为“黑盒”模型,可解释性差,降低了临床医生的接受度;其四,临床验证不足,缺乏大规模、多中心、前瞻性试验支撑;其五,AI工具在与PACS、电子病历及放射科工作流程的融合方面仍存在实际应用障碍。需要特别指出的是,一些难点成为热点研究并取得了一定进展,例如多机构统一标准图像数据库的构建、基于模型迁移学习的跨设备方法、基于可解释AI的图像不确定性量化等,有可能解决上述难点问题。

       未来研究可重点在以下方面开展:(1)建立标准的参考数据集提高模型的稳定性和可重复性;(2)创建解析解释性、迁移性的模型;(3)创建适应临床医生需求的多模态、任务特定模型。此外,还需要加强放射科医生、肿瘤科专家、计算专家与监管人员的合作,促进AI在头颈部淋巴结诊断中的有效、安全应用。通过各方共同努力,才能真正实现AI在临床肿瘤学的革新性应用。

[1]
GAO W J, HU Y H, ZHU D, et al. Extranodal extension in bilateral cervical metastases: a predictor of undesirable survival outcomes despite aggressive salvage treatment in oral cancer patients[J]. J Cancer, 2021, 12(19): 5848-5863. DOI: 10.7150/jca.60152.
[2]
MAHMOOD H, SHABAN M, RAJPOOT N, et al. Artificial Intelligence-based methods in head and neck cancer diagnosis: an overview[J]. Br J Cancer, 2021, 124(12): 1934-1940. DOI: 10.1038/s41416-021-01386-x.
[3]
LIAO L J, LO W C, HSU W L, et al. Detection of cervical lymph node metastasis in head and neck cancer patients with clinically N0 neck-a meta-analysis comparing different imaging modalities[J/OL]. BMC Cancer, 2012, 12: 236 [2025-06-07]. https://pubmed.ncbi.nlm.nih.gov/22691269/. DOI: 10.1186/1471-2407-12-236.
[4]
ALABI R O, ELMUSRATI M, LEIVO I, et al. Artificial intelligence-driven radiomics in head and neck cancer: current status and future prospects[J/OL]. Int J Med Inform, 2024, 188: 105464 [2025-06-07]. https://pubmed.ncbi.nlm.nih.gov/38728812/. DOI: 10.1016/j.ijmedinf.2024.105464.
[5]
PENG Z Y, WANG Y M, WANG Y X, et al. Application of radiomics and machine learning in head and neck cancers[J]. Int J Biol Sci, 2021, 17(2): 475-486. DOI: 10.7150/ijbs.55716.
[6]
SHAMSHAD F, KHAN S, ZAMIR S W, et al. Transformers in medical imaging: a survey[J/OL]. Med Image Anal, 2023, 88: 102802 [2025-06-07]. https://pubmed.ncbi.nlm.nih.gov/37315483/. DOI: 10.1016/j.media.2023.102802.
[7]
KOH D M, PAPANIKOLAOU N, BICK U, et al. Artificial intelligence and machine learning in cancer imaging[J/OL]. Commun Med (Lond), 2022, 2: 133 [2025-06-07]. https://pubmed.ncbi.nlm.nih.gov/36310650/. DOI: 10.1038/s43856-022-00199-0.
[8]
BARRAGÁN-MONTERO A, JAVAID U, VALDÉS G, et al. Artificial intelligence and machine learning for medical imaging: a technology review[J/OL]. Phys Med, 2021, 83: 242-256 [2025-06-07]. https://pubmed.ncbi.nlm.nih.gov/33979715/. DOI: 10.1016/j.ejmp.2021.04.016.
[9]
LEI Y M, YIN M, YU M H, et al. Artificial intelligence in medical imaging of the breast[J/OL]. Front Oncol, 2021, 11: 600557 [2025-06-07]. https://pubmed.ncbi.nlm.nih.gov/34367938/. DOI: 10.3389/fonc.2021.600557.
[10]
GIANNITTO C, MERCANTE G, AMMIRABILE A, et al. Radiomics-based machine learning for the diagnosis of lymph node metastases in patients with head and neck cancer: Systematic review[J]. Head Neck, 2023, 45(2): 482-491. DOI: 10.1002/hed.27239.
[11]
VALIZADEH P, JANNATDOUST P, PAHLEVAN-FALLAHY M T, et al. Diagnostic accuracy of radiomics and artificial intelligence models in diagnosing lymph node metastasis in head and neck cancers: a systematic review and meta-analysis[J]. Neuroradiology, 2025, 67(2): 449-467. DOI: 10.1007/s00234-024-03485-x.
[12]
ZHOU S K, GREENSPAN H, DAVATZIKOS C, et al. A review of deep learning in medical imaging: Imaging traits, technology trends, case studies with progress highlights, and future promises[J]. Proc IEEE Inst Electr Electron Eng, 2021, 109(5): 820-838. DOI: 10.1109/JPROC.2021.3054390.
[13]
RINNEBURGER M, CAROLUS H, IUGA A I, et al. Automated localization and segmentation of cervical lymph nodes on contrast-enhanced CT using a 3D foveal fully convolutional neural network[J/OL]. Eur Radiol Exp, 2023, 7(1): 45 [2025-06-07]. https://pubmed.ncbi.nlm.nih.gov/37505296/. DOI: 10.1186/s41747-023-00360-x.
[14]
REINDERS F C J, SAVENIJE M H F, DE RIDDER M, et al. Automatic segmentation for magnetic resonance imaging guided individual elective lymph node irradiation in head and neck cancer patients[J/OL]. Phys Imaging Radiat Oncol, 2024, 32: 100655 [2025-06-07]. https://pubmed.ncbi.nlm.nih.gov/39502445/. DOI: 10.1016/j.phro.2024.100655.
[15]
NIKULIN P, ZSCHAECK S, MAUS J, et al. A convolutional neural network with self-attention for fully automated metabolic tumor volume delineation of head and neck cancer in [18 F] FDG PET/CT[J]. Eur J Nucl Med Mol Imaging, 2023, 50(9): 2751-2766. DOI: 10.1007/s00259-023-06197-1.
[16]
GAO Y X, JIANG Y, PENG Y H, et al. Medical image segmentation: a comprehensive review of deep learning-based methods[J/OL]. Tomography, 2025, 11(5): 52 [2025-06-07]. https://pubmed.ncbi.nlm.nih.gov/40423254/. DOI: 10.3390/tomography11050052.
[17]
FIEBELKORN I C, KASTNER S. Functional specialization in the attention network[J/OL]. Annu Rev Psychol, 2020, 71: 221-249 [2025-06-07]. https://pubmed.ncbi.nlm.nih.gov/31514578/. DOI: 10.1146/annurev-psych-010418-103429.
[18]
ALOM M Z, YAKOPCIC C, HASAN M, et al. Recurrent residual U-Net for medical image segmentation[J/OL]. J Med Imaging (Bellingham), 2019, 6(1): 014006 [2025-06-07]. https://pubmed.ncbi.nlm.nih.gov/30944843/. DOI: 10.1117/1.JMI.6.1.014006.
[19]
JIANG Y, CHEN L, ZHANG H, et al. Breast cancer histopathological image classification using convolutional neural networks with small SE-ResNet module[J/OL]. PLoS One, 2019, 14(3): e0214587 [2025-06-07]. https://pubmed.ncbi.nlm.nih.gov/30925170/. DOI: 10.1371/journal.pone.0214587.
[20]
ORTEGA C, ESHET Y, PRICA A, et al. Combination of FDG PET/CT radiomics and clinical parameters for outcome prediction in patients with Hodgkin's lymphoma[J/OL]. Cancers, 2023, 15(7): 2056 [2025-06-07]. https://pubmed.ncbi.nlm.nih.gov/37046717/. DOI: 10.3390/cancers15072056.
[21]
DUAN C L, AN L, YANG Y F, et al. The role of artificial intelligence and radiomics in the management of lymphomas by PET/CT: the clairvoyance in clinic[J/OL]. Cancer Manag Res, 2025, 17: 1457-1475 [2025-06-07]. https://pubmed.ncbi.nlm.nih.gov/40704298/. DOI: 10.2147/CMAR.S529589.
[22]
SHIRI I, RAZEGHI B, VAFAEI SADR A, et al. Multi-institutional PET/CT image segmentation using federated deep transformer learning[J/OL]. Comput Methods Programs Biomed, 2023, 240: 107706 [2025-06-07]. https://pubmed.ncbi.nlm.nih.gov/37506602/. DOI: 10.1016/j.cmpb.2023.107706.
[23]
朱小珊, 彭娟. 能谱CT在头及颈部疾病诊断中的应用[J]. 医疗卫生装备, 2019, 40(4): 98-102. DOI: 10.19745/j.1003-8868.2019103.
ZHU X S, PENG J. Application of spectral CT in diagnosis of head and neck diseases[J]. Chin Med Equip J, 2019, 40(4): 98-102. DOI: 10.19745/j.1003-8868.2019103.
[24]
魏培英, 蒋念东, 韩志江, 等. CT强化程度对甲状腺乳头状癌颈部淋巴结转移的诊断价值[J]. 中华内分泌外科杂志, 2020, 14(2): 5. DOI: 10.3760/cma.j.issn.115807-20190414-00081.
WEI P Y, JIANG N D, HAN Z J, et al. Diagnostic value of CT enhancement degree in cervical lymph node metastasis of papillary thyroid carcinoma[J]. Chin J Endocr Surg, 2020, 14(2): 5. DOI: 10.3760/cma.j.issn.115807-20190414-00081.
[25]
CHEN M T, JIANG Y D, ZHOU X H, et al. Dual-energy computed tomography in detecting and predicting lymph node metastasis in malignant tumor patients: a comprehensive review[J/OL]. Diagnostics (Basel), 2024, 14(4): 377 [2025-06-07]. https://pubmed.ncbi.nlm.nih.gov/38396416/. DOI: 10.3390/diagnostics14040377.
[26]
LIU Y, ZHANG X, WANG J, et al. Dual-energy CT quantitative parameters for predicting lymph node metastasis in papillary thyroid carcinoma [J/OL]. Eur J Radiol, 2020, 130: 109153 [2025-06-07]. https://pubmed.ncbi.nlm.nih.gov/32717577/. DOI: 10.1016/j.ejrad.2020.109153.
[27]
FORGHANI R, CHATTERJEE A, REINHOLD C, et al. Head and neck squamous cell carcinoma: prediction of cervical lymph node metastasis by dual-energy CT texture analysis with machine learning[J]. Eur Radiol, 2019, 29(11): 6172-6181. DOI: 10.1007/s00330-019-06159-y.
[28]
QIU Y X, FEI Y, LIU J Y, et al. Prevalence, risk factors and location of skip metastasis in papillary thyroid carcinoma: a systematic review and meta-analysis[J/OL]. Cancer Manag Res, 2019, 11: 8721-8730 [2025-06-07]. https://pubmed.ncbi.nlm.nih.gov/31576174/. DOI: 10.2147/CMAR.S200628.
[29]
ROTSTEIN L. The role of lymphadenectomy in the management of papillary carcinoma of the thyroid[J]. J Surg Oncol, 2009, 99(4): 186-188. DOI: 10.1002/jso.21234.
[30]
KITAHARA C M, SOSA J A. The changing incidence of thyroid cancer[J]. Nat Rev Endocrinol, 2016, 12(11): 646-653. DOI: 10.1038/nrendo.2016.110.
[31]
ROBENSHTOK E, FISH S, BACH A, et al. Suspicious cervical lymph nodes detected after thyroidectomy for papillary thyroid cancer usually remain stable over years in properly selected patients[J]. J Clin Endocrinol Metab, 2012, 97(8): 2706-2713. DOI: 10.1210/jc.2012-1553.
[32]
ONOUE K, FUJIMA N, ANDREU-ARASA V C, et al. Cystic cervical lymph nodes of papillary thyroid carcinoma, tuberculosis and human papillomavirus positive oropharyngeal squamous cell carcinoma: utility of deep learning in their differentiation on CT[J/OL]. Am J Otolaryngol, 2021, 42(5): 103026 [2025-06-07]. https://pubmed.ncbi.nlm.nih.gov/33862564/. DOI: 10.1016/j.amjoto.2021.103026.
[33]
WANG C, YU P Y, ZHANG H C, et al. Artificial intelligence-based prediction of cervical lymph node metastasis in papillary thyroid cancer with CT[J]. Eur Radiol, 2023, 33(10): 6828-6840. DOI: 10.1007/s00330-023-09700-2.
[34]
JIN D, NI X Q, ZHANG X D, et al. Multiphase dual-energy spectral CT-based deep learning method for the noninvasive prediction of head and neck lymph nodes metastasis in patients with papillary thyroid cancer[J/OL]. Front Oncol, 2022, 12: 869895 [2025-06-07]. https://pubmed.ncbi.nlm.nih.gov/35515110/. DOI: 10.3389/fonc.2022.869895.
[35]
MOORE B A, WEBER R S, PRIETO V, et al. Lymph node metastases from cutaneous squamous cell carcinoma of the head and neck[J]. Laryngoscope, 2005, 115(9): 1561-1567. DOI: 10.1097/01.mlg.0000173202.56739.9f.
[36]
FERRIS R L, XI L Q, RAJA S, et al. Molecular staging of cervical lymph nodes in squamous cell carcinoma of the head and neck[J]. Cancer Res, 2005, 65(6): 2147-2156. DOI: 10.1158/0008-5472.can-04-3717.
[37]
BARDOSI Z R, DEJACO D, SANTER M, et al. Benchmarking eliminative radiomic feature selection for head and neck lymph node classification[J/OL]. Cancers (Basel), 2022, 14(3): 477 [2025-06-07]. https://pubmed.ncbi.nlm.nih.gov/35158745/. DOI: 10.3390/cancers14030477.
[38]
CHOI Y J, LEE I S, SONG Y S, et al. Diagnostic performance of diffusion-weighted (DWI) and dynamic contrast-enhanced (DCE) MRI for the differentiation of benign from malignant soft-tissue tumors[J]. J Magn Reson Imaging, 2019, 50(3): 798-809. DOI: 10.1002/jmri.26607.
[39]
ZHOU Z W, XUE J W, WU Y, et al. Automated detection of metastatic lymph nodes in head and neck malignant tumors on high - resolution MRI images using an improved convolutional neural network[J/OL]. Int J Med Inform, 2025, 200: 105904 [2025-06-07]. https://pubmed.ncbi.nlm.nih.gov/40220628/. DOI: 10.1016/j.ijmedinf.2025.105904.
[40]
ALBANO D, BRUNO F, AGOSTINI A, et al. Dynamic contrast-enhanced (DCE) imaging: state of the art and applications in whole-body imaging[J]. Jpn J Radiol, 2022, 40(4): 341-366. DOI: 10.1007/s11604-021-01223-4.
[41]
YUAN Y, REN J L, TAO X F. Machine learning-based MRI texture analysis to predict occult lymph node metastasis in early-stage oral tongue squamous cell carcinoma[J]. Eur Radiol, 2021, 31(9): 6429-6437. DOI: 10.1007/s00330-021-07731-1.
[42]
WANG Y P, YU T H, YANG Z H, et al. Radiomics based on magnetic resonance imaging for preoperative prediction of lymph node metastasis in head and neck cancer: Machine learning study[J]. Head Neck, 2022, 44(12): 2786-2795. DOI: 10.1002/hed.27189.
[43]
WANG F, TAN R K, FENG K, et al. Magnetic resonance imaging-based radiomics features associated with depth of invasion predicted lymph node metastasis and prognosis in tongue cancer[J]. J Magn Reson Imaging, 2022, 56(1): 196-209. DOI: 10.1002/jmri.28019.
[44]
WANG D W, HE X, HUANG C M, et al. Magnetic resonance imaging-based radiomics and deep learning models for predicting lymph node metastasis of squamous cell carcinoma of the tongue[J]. Oral Surg Oral Med Oral Pathol Oral Radiol, 2024, 138(1): 214-224. DOI: 10.1016/j.oooo.2024.01.016.
[45]
LIU S, ZHANG A H, XIONG J J, et al. The application of radiomics machine learning models based on multimodal MRI with different sequence combinations in predicting cervical lymph node metastasis in oral tongue squamous cell carcinoma patients[J]. Head Neck, 2024, 46(3): 513-527. DOI: 10.1002/hed.27605.
[46]
LAN T J, KUANG S J, LIANG P S, et al. MRI-based deep learning and radiomics for prediction of occult cervical lymph node metastasis and prognosis in early-stage oral and oropharyngeal squamous cell carcinoma: a diagnostic study[J]. Int J Surg, 2024, 110(8): 4648-4659. DOI: 10.1097/JS9.0000000000001578.
[47]
XIE K, JIANG H, CHEN X W, et al. Multiparameter MRI-based model integrating radiomics and deep learning for preoperative staging of laryngeal squamous cell carcinoma[J/OL]. Sci Rep, 2025, 15(1): 16239 [2025-06-07]. https://pubmed.ncbi.nlm.nih.gov/40346120/. DOI: 10.1038/s41598-025-01270-1.
[48]
YIMIT Y, YASIN P, TUERSUN A, et al. Multiparametric MRI-based interpretable radiomics machine learning model differentiates medulloblastoma and ependymoma in children: a two-center study[J]. Acad Radiol, 2024, 31(8): 3384-3396. DOI: 10.1016/j.acra.2024.02.040.
[49]
ZHANG D Y, SHEN M Y, ZHANG L, et al. Establishment of an interpretable MRI radiomics-based machine learning model capable of predicting axillary lymph node metastasis in invasive breast cancer[J/OL]. Sci Rep, 2025, 15(1): 26030 [2025-06-07]. https://pubmed.ncbi.nlm.nih.gov/40676103/. DOI: 10.1038/s41598-025-10818-0.
[50]
CHENG Z Q, MA J J, YIN L, et al. Non-invasive molecular imaging for precision diagnosis of metastatic lymph nodes: opportunities from preclinical to clinical applications[J]. Eur J Nucl Med Mol Imaging, 2023, 50(4): 1111-1133. DOI: 10.1007/s00259-022-06056-5.
[51]
MUGASA H, DUA S, KOH J E W, et al. An adaptive feature extraction model for classification of thyroid lesions in ultrasound images[J/OL]. Pattern Recognit Lett, 2020, 131: 463-473 [2025-06-07]. https://www.sciencedirect.com/science/article/abs/pii/S0167865520300490?via%3Dihub. DOI: 10.1016/j.patrec.2020.02.009.
[52]
ZHOU L Q, ZENG S E, XU J W, et al. Deep learning predicts cervical lymph node metastasis in clinically node-negative papillary thyroid carcinoma[J/OL]. Insights Imaging, 2023, 14(1): 222 [2025-06-07]. https://pubmed.ncbi.nlm.nih.gov/38117404/. DOI: 10.1186/s13244-023-01550-2.
[53]
CHANG L C, ZHANG Y Q, ZHU J L, et al. An integrated nomogram combining deep learning, clinical characteristics and ultrasound features for predicting central lymph node metastasis in papillary thyroid cancer: a multicenter study[J/OL]. Front Endocrinol (Lausanne), 2023, 14: 964074 [2025-06-07]. https://pubmed.ncbi.nlm.nih.gov/36896175/. DOI: 10.3389/fendo.2023.964074.
[54]
WANG Z Z, QU L M, CHEN Q T, et al. Deep learning-based multifeature integration robustly predicts central lymph node metastasis in papillary thyroid cancer[J/OL]. BMC Cancer, 2023, 23(1): 128 [2025-06-07]. https://pubmed.ncbi.nlm.nih.gov/36750791/. DOI: 10.1186/s12885-023-10598-8.
[55]
FAN F J, LI F, WANG Y X, et al. Deep learning based on ultrasound images predicting cervical lymph node metastasis in postoperative patients with differentiated thyroid carcinoma[J]. Br J Radiol, 2025, 98(1170): 920-928. DOI: 10.1093/bjr/tqaf047.
[56]
SHEN P C, YANG Z Y, SUN J J, et al. Explainable multimodal deep learning for predicting thyroid cancer lateral lymph node metastasis using ultrasound imaging[J/OL]. Nat Commun, 2025, 16(1): 7052 [2025-06-07]. https://pubmed.ncbi.nlm.nih.gov/40750786/. DOI: 10.1038/s41467-025-62042-z.
[57]
BHATTACHARYA K, MAHAJAN A, VAISH R, et al. Imaging of neck nodes in head and neck cancers - a comprehensive update[J]. Clin Oncol (R Coll Radiol), 2023, 35(7): 429-445. DOI: 10.1016/j.clon.2023.03.012.
[58]
MYRONENKO A, SIDDIQUEE M M R, YANG D, et al. Automated head and neck tumor segmentation from 3D PET/CT HECKTOR 2022 challenge report[M]. Springer, 2022: 31-37. DOI: 10.1007/978-3-031-27420-6_2.
[59]
SUN X, AN C, WANG L. A coarse-to-fine ensembling framework for head and neck tumor and lymph segmentation in CT and PET images[M]. Springer, 2022: 38-46. DOI: 10.1016/j.acra.2023.05.009.
[60]
JIANG H, HAIMERL J, GU X, et al. A general web-based platform for automatic delineation of head and neck gross tumor volumes in PET/CT images[M]. Springer, 2022: 47-53. DOI: 10.1007/978-3-031-27420-6_4.
[61]
REBAUD L, ESCOBAR T, KHALID F, et al. Simplicity is all you need: out-of-the-box nnUNet followed by binary-weighted radiomic model for segmentation and outcome prediction in head and neck PET/CT[M]. Springer, 2022: 121-134. DOI: 10.1007/978-3-031-27420-6_13.
[62]
SALAHUDDIN Z, CHEN Y, ZHONG X, et al. HNT-AI: an automatic segmentation framework for head and neck primary tumors and lymph nodes in FDG-PET/CT images[M]. Springer, 2022: 212-220. DOI: 10.1007/978-3-031-27420-621.
[63]
WANG A, BAI T, NGUYEN D, et al. Octree boundary transfiner: efficient transformers for tumor segmentation refinement[M]. Springer, 2022: 54-60. DOI: 10.1007/978-3-031-27420-65.
[64]
JAIN A, HUANG J, RAVIPATI Y, et al. Head and neck primary tumor and lymph node auto-segmentation for PET/CT scans[M]. Springer, 2022: 61-69. DOI: 10.1007/978-3-031-27420-6_6.
[65]
KUDOH T, HAGA A, KUDOH K, et al. Radiomics analysis of [(18)F]-fluoro-2-deoxyglucose positron emission tomography for the prediction of cervical lymph node metastasis in tongue squamous cell carcinoma[J]. Oral Radiol, 2023, 39(1): 41-50. DOI: 10.1007/s11282-022-00600-7.
[66]
KOVACS DG, LADEFOGED CN, ANDERSEN KF, et al. Clinical Evaluation of Deep Learning for Tumor Delineation on (18)F-FDG PET/CT of Head and Neck Cancer[J]. J Nucl Med, 2024, 65(4): 623-629. DOI: 10.2967/jnumed.123.266574.

上一篇 基于磁共振成像的脑氧代谢定量评估技术及其临床应用进展
下一篇 扩张型心肌病精准诊疗进展:聚焦遗传学与心脏磁共振成像
  
诚聘英才 | 广告合作 | 免责声明 | 版权声明
联系电话:010-67113815
京ICP备19028836号-2