不同CT阈值下实性成分占比对小肺癌浸润性预测的影响.pdf

第41卷第3期2024年3月DOI:10.3969/j.issn.1005-202X.2024.03.009不同CT阈值下实性成分占比对小肺癌浸润性预测的影响中国医学物理学杂志Chinese Journal of Medical PhysicsVol.41 No.3March 2024医学影像物理-323-牛树国,周福兴,颜克松,赵润生,刘彬彬”,柴文晓1.武威市中医医院放射科，甘肃武威7 330 0 0；2.甘肃省人民医院肿瘤介人科，甘肃兰州7 30 0 0 0；3.武威市中医医院病理科，甘肃武威7 330 0 0【摘要】目的：利用人工智能辅助测量，比较在不同CT阈值下测得的肿瘤实性成分占比（CTR）对2 cm小肺癌浸润性预测的准确率，探讨预测肺癌浸润性的CTR阅值及对应的CT阅值。方法：收集2 0 2 1年1月至2 0 2 3年5月武威市中医医院就诊的59 例肺癌患者（共7 8 个肺结节）的临床资料，分析在CT阅值分别为-40 0、-350、-30 0、-2 50、-2 0 0-150 HU时测得的CTR对直径0.322时结节为微浸润腺癌或浸润性腺癌可能性较大。【关键词】肺结节；肺癌；CT阅值；实性成分占比；人工智能【中图分类号】R734.2;R816.4Predictive value of consolidation/tumor ratio at different CT thresholds for invasiveness insmall lungcancerNIU Shuguo,ZHOU Fuxing,YAN Kesong,ZHAO Runsheng,LIU Binbin,CHAI Wenxiao?1.Department of Radiology,Wuwei Hospital of Traditional Chinese Medicine,Wuwei 733000,China;2.Department of TumorIntervention,Gansu Provincial Hospital,Lanzhou 730000,China;3.Department of Pathology,Wuwei Hospital of Traditional ChineseMedicine,Wuwei 733000,ChinaAbstract:Objective To compare the accuracy of consolidation/tumor ratio(CTR)measured at different CT thresholds for theprediction of invasiveness in small lung cancer with diameter 2 cm using artificial intelligence-assisted measurements,andto explore the CTR thresholds and the corresponding CT thresholds for predicting lung cancer invasiveness.MethodsClinical data from 59 lung cancer patients(78 lung nodules in total)treated at Wuwei Hospital of Traditional ChineseMedicine from January 2021 to May 2023 were collected to analyze the prediction efficacy of CTR on invasiveness in smalllung cancer with diameter 2 cm measured at CT thresholds of-400,-350,-300,-250,-200,-150 HU.R0C curves wereplotted to determine the optimal critical value for invasiveness prediction,followed by the corresponding CT threshold.Results The highest diagnostic efficacy for the invasiveness of lung nodules was achieved at a CT threshold of-250 HU,withan area under the curve of 0.931,sensitivity of 77.5%,specificity of 100%,and an optimal CTR threshold of 0.322.Conclusion For small lung cancers with a maximum diameter 2 cm,CTR measured at a CT threshold of-250 HU canaccurately predict lung cancer invasiveness.At CTR 0.322,the nodule is more likely to be microinvasive or invasiveadenocarcinoma.Keywords:lung nodule;lung cancer;CT threshold;consolidation/tumor ratio;artificial intelligence【文献标志码】A【文章编号】10 0 5-2 0 2 X（2 0 2 4)0 3-0 32 3-0 4前言我国肺癌高危人群CT筛查发现肺结节的阳性【收稿日期】2 0 2 3-11-0 3【基金项目】甘肃省自然科学基金（2 2 JR5RA687）；武威市科技计划项目(WW2101168)【作者简介】牛树国，硕士，主治医师，研究方向：影像诊断与介人治疗,E-mail:【通信作者】柴文晓，硕士，主任医师，研究方向：肿瘤介入治疗，E-mail:率高达2 2.9%，其中约6.34%为恶性结节，而早期肺癌(Ia期)行肺叶切除手术治疗5年生存率可达到90%以上，原位腺癌及微浸润腺癌基本不出现淋巴结转移，手术后5年生存率接近10 0%2。因此早发现早治疗对早期肺癌患者获益明显3。肿瘤实性成分-324-占比(ConsolidationTumorRatio,CTR）是指在高分辨计算机断层扫描（HRCT)肺窗中肿瘤最大实性直径与肿瘤最大直径的比值4。近年的多项研究表明CTR可作为早期预测肺癌良恶性的重要指标，是早期肺癌复发、气腔播散及预后等独立相关因素，并成为肺叶切除或亚肺叶切除手术方式选择及是否行淋巴结清扫的主要参考指标，为早期肺癌患者治疗方案提供依据5-8。但是，目前CTR值的测定尚无统一标准，在不同CT阈值下测得的数值不同，限制了CTR在肺结节良恶性预测及指导治疗方面的应用9。人工智能(AI)在影像诊断方面的应用日新月异,尤其是肺结节辅助诊断方面的应用最为广泛,对肺结节识别、大小及CT值测量、恶性等级判断等方面均具有较高的准确率10-12。AI在准确测量结节大小、判断结节类型方面较人工具有明显优势，并可根据不同的CT阈值测得相应的CTR数据13。本研究利用AI辅助测量肺结节参数，通过比较不同CT阈值下测得的CTR对小肺癌浸润性预测的准确率,探讨预测肺癌浸润性的最佳CT阈值。资料与方法1.1研究对象收集2 0 2 1年1月至2 0 2 3年5月武威市中医医院符合纳人和排除条件的59 例患者，其中男39 例，女20例，年龄范围48 8 1岁，平均年龄（6 2.35.7)岁。经多层螺旋CT肺部平扫共检出7 8 个肺部结节，经手术或穿刺活检确诊为肺癌,其中不典型腺瘤样增生10个（12.8%），原位癌13个（16.7%），微浸润腺癌2 1个（2 6.9%），浸润性腺癌34个（43.6%）。纳入标准：有完整的最大吸气末HRCT影像资料，病变所在层面显示清晰、无呼吸及运动伪影；肺内亚实性结节,直径2.0 cm;病理确诊为肺癌，病理资料保存完整；患者或亲属知情同意。排除标准：病历资料不完整；存在呼吸及体外异物等干扰，影响CT图像质量；CT检查前行活检或放疗、化疗、靶向或免疫治疗等。本研究符合伦理要求并获得医院伦理委员会批准。1.2检查方法采用美国GE公司2 56 排RevolutionCT进行全肺扫描。检查前对患者进行屏气训练，在患者最大吸气末时屏气行肺尖到肺底的薄层扫描。管电压：120kV，管电流：150 mA，层厚：5mm，重建层厚：0.625mm，螺距：0.516 mm，扫描矩阵：512 512，FOV:350mmx350mm。所有肺结节均经手术或穿刺活检取得病理组织,由两名病理科医师分别对病理切片进行判读,将中国医学物理学杂志不典型腺瘤样增生和原位癌划分为非浸润性，将微浸润腺癌和浸润性腺癌划分为浸润性14。意见不一致时共同讨论确定。1.3CT阈值选择将扫描后获得的高分辨CT图像导入AI工作站数坤软件，数坤（北京)网络科技股份有限公司进行定量分析，自动识别并测量结节的三维体积、实性成分体积与非实性成分体积、CTR等15。在上述基础上，分别设置不同CT阈值并测量结节实性成分与体积，由软件计算出CTR。共设置6 个CT阈值：分别为-40 0、-350、-30 0、-2 50、-2 0 0、-150 HU。1.4诊断效能评价纳人研究的结节判断浸润性的CTR临界值设为0.25,若CTR0.25认为结节具有浸润性16。预测结果与病理诊断相对照,以判断是否准确为结局变量,以不同CT阈值下的CTR为检验变量,绘制受试者工作特征(ROC)曲线。通过最大曲线下面积(AUC)判断最佳CT阈值，选择Youden指数最大时的CTR值作为判断结节浸润性的CTR临界值。1.5统计学方法采用GraphPadPrism8.0.2软件对数据进行统计分析，并绘制统计图。计量资料用均数土标准差表示，计数资料用率(%)表示。采用卡方检验进行多组间率的比较。采用单因素方差分析检验组间平均值差异性。利用二元Logistic回归分析CTR变量是否影响预测结局。绘制ROC曲线，分析独立预测因素的诊断效能。P0.05为差异有统计学意义。2结果2.1不同CT阈值测得的肺结节CTR经单因素方差分析显示，各组间CTR值方差齐（P=0.996），各组CTR均数随CT阈值增加呈递减关系(P0.322的结节为浸润性腺癌的可能性大于CTR0.322者。这将为肺结节的诊断及治疗方案的制定提供参考依据。本研究仍存在一定的局限性，首先,纳入的样本量较小,且均为直径2 cm肺腺癌,不能反映所有肺结节的CT特点；其次,选择的CT阈值较少,仅能粗略测得诊断效能相对较高的CT阈值。后续的研究应纳人更全面的结节类型，设计更优的试验方案，以期探索CTR在临床中更广泛的应用价值。【参考文献】1 Ettinger DS,Wood DE,Aisner DL,et al.Non-small cell lung cancer,version 3.2022,NCCN clinical practice guidelines in oncologyJ.JNatl Compr Canc Netw,2022,20(5):497-530.2刘宝东，陈海泉,刘伦旭，等.肺结节多学科微创诊疗中国专家共识J.中国胸心血管外科临床杂志,2 0 2 3,30(8):10 6 1-10 7 4.Liu BD,Chen HQ,Liu LX,et al.Chinese expert consensus onmultidisciplinary minimally invasive diagnosis and treatment ofpulmonary nodules J.Chinese Journal of Clinical Thoracic andCardiovascular Surgery,2023,30(8):1061-1074.3 Subramanian M,MeMurry T,Meyers BF,et al.Long-term results forclinical stage IA lung cancer:comparing lobectomy and sublobarresectionJ.Ann Thorac Surg,2018,106(2):375-381.4 Ohde Y,Nagai K,Yoshida J,et al.The proportion of consolidation toground-glass opacity on high resolution CT is a good predictor for-326-distinguishing the population of non-invasive peripheraladenocarcinomaJ.Lung Cancer,2003,42(3):303-310.5 Xi JJ,Yin JC,Liang JQ,et al.Prognostic impact of radiologicalconsolidation tumor ratio in clinical stage IA pulmonary ground glassopacitiesJ.Front Oncol,2021,11:616149.6 Dong H,Wang XB,Qiu YG,et al.Establishment and visualization ofa model based on high-resolution CT qualitative and quantitativefeatures for prediction of micropapillary or solid components ininvasive lung adenocarcinomaJ.J Cancer Res Clin Oncol,2023,149(12):10519-10530.7 Ito H,Suzuki K,Mizutani T,et al.Long-term survival outcome afterlobectomy in patients with clinical T,N。l u n g c a n c e r J.J T h o r a cCardiovasc Surg,2021,161(1):281-290.8 Wu YM,Song WP,Wang DN,et al.Prognostic value of consolidation-to-tumor ratio on computed tomography in NSCLC:a meta-analysisJ.World J Surg Oncol,2023,21(1):190.9黄汉清，叶波.肿瘤实性成分占比在早期周围型肺癌诊疗中的研究进展J.中国肺癌杂志,2 0 2 2,2 5(10):7 6 4-7 7 0.Huang HQ,Ye B.Research progress in the effect of consolidationtumor ratio on the diagnosis and treatment of early-stage peripherallung cancer J.Chinese Journal of Lung Cancer,2022,25(10):764-770.【10 吴阶平医学基金会模拟医学部胸外科专委会。人工智能在肺结节诊治中的应用专家共识(2 0 2 2 年版)J.中国肺癌杂志,2 0 2 2,2 5(4):219-225.Thoracic Surgery Committee,Department of Simulated Medicine,WuJieping Medical Foundation.Chinese experts consensus on artificialintlligence assisted management for pulmonary nodule(2022 version)J.Chinese Journal of Lung Cancer,2022,25(4):219-225.11王新宇,赵静文，刘翔,等.人工智能在肺结节筛查和肺癌诊断中的应用J。中国医学物理学杂志，2 0 2 3,40(9)：118 2-118 8.Wang XY,Zhao JW,Liu X,et al.Applications of artificial intelligencein lung nodule detection and lung cancer diagnosis J.ChineseJournal of Medical Physics,2023,40(9):1182-1188.12 Gandhi Z,Gurram P,Amgai B,et al.Artificial intelligence and lungcancer:impact on improving patient outcomesJ.Cancers(Basel),2023,15(21):5236.13 Blanc D,Racine V,Khalil A,et al.Artificial intelligence solution toclassify pulmonary nodules on CTJ.Diagn Interv Imaging,2020,101(12):803-810.14 Moreira AL,Ocampo PSS,Xia YH,et al.A grading system for中国医学物理学杂志invasive pulmonary adenocarcinoma:a proposal from the internationalassociation for the study of lung cancer pathology committeeJ.JThorac Oncol,2020,15(10):1599-1610.15 Lin YW,Ouyang L.An effective lung lesion segmentation andclassification method for CT images based on a hybrid deep learningnetworkJJ.IEEE J Biomed Health Inform,2021,25(5):1427.16 Suzuki K,Koike T,Asakawa T,et al.A prospective radiological studyof thin-section computed tomography to predict pathologicalnoninvasiveness in peripheral clinical IA lung cancer(Japan clinicaloncology group 0201)J.J Thorac Oncol,2011,6(4):751-756.17 Asamura H,Hishida T,Suzuki K,et al.Radiographically determinednoninvasive adenocarcinoma of the lung:survival outcomes of Japanclinical oncology group 0201J.J Thorac Cardiovasc Surg,2013,146(1):24-30.18 Fang C,Xiang YW,Han WL.Preoperative risk factors of lymph nodemetastasis in clinical No lung adenocarcinoma of 3 cm or less indiameterJ.BMC Surg,2022,22(1):153.19 Lin MW,Chen LW,Yang SM,et al.CT-based deep-learning model forspread-through-air-spaces prediction in ground glass-predominant lungadenocarcinomaJ.Ann Surg Oncol,2024,31(3):1536-1545.20 Suzuki K,Watanabe SI,Wakabayashi M,et al.A single-arm study ofsublobar resection for ground-glass opacity dominant peripheral lungcancerJ.J Thorac Cardiovasc Surg,2022,163(1):289-301.e2.21 Pei Q,Luo YN,Chen YY,et al.Artificial intelligence in clinicalapplications for lung cancer:diagnosis,treatment and prognosisJ.Clin Chem Lab Med,2022,60(12):1974-1983.22 Ather S,Kadir T,Gleeson F.Artificial intelligence and radiomics inpulmonary nodule management:current status and future applicationsJ.Clin Radiol,2020,75(1):13-19.23 Chen JW,Cao R,Jiao SY,et al.Application value of a computer-aideddiagnosis and management system for the detection of lung nodulesJ.Quant Imaging Med Surg,2023,13(10):6929-6941.24 Chiu HY,Chao HS,Chen YM.Application of artificial intelligence inlung cancerJ.Cancers(Basel),2022,14(6):1370.25 Koike S,Shimizu K,Ide S,et al.Is using a consolidation tumor ratio0.5 as criterion feasible in daily practice?Evaluation of interobservermeasurement variability of consolidation tumor ratio of lung cancerless than 3 cm in sizeJ.Thorac Cancer,2022,13(21):3018-3024.（编辑：黄开颜）第41卷