基于CT影像组学和临床特征预测单发肺结节生长的临床意义

doi:10.3877/cma.j.issn.1674-6902.2026.01.009

目的

探讨计算机断层扫描(computed tomography, CT)影像组学和临床特征预测单发肺结节生长的临床意义。

方法

选择2021年5月至2023年5月我院收治的胸部高分辨率CT(high-resolution computed tomography, HRCT)示单发肺结节患者165例为对象，根据7︰3随机分为训练集115例，验证集50例。通过开源Python包Pyradiomics提取基于CT影像组学特征变量。随访2年观察肺结节生长情况，使用最大相关性最小冗余性方法及LASSO分析筛选影像学特征变量，构建模型，采用受试者工作特征(receiver operating characteristic, ROC)曲线对比模型预测意义。

结果

165例中生长型肺结节99枚(60.00%)，稳定型肺结节66枚(40.00%)。LASSO分析筛选出5个特征original_shape_MajorAxisLength、log.sigma.1.0.mm.3D_glrlm_ShortRunLowGrayLevelEmphasis、log.sigma.5.0.mm.3D_firstorder_90Percentile、wavelet.LHH_gldm_LargeDependenceHighGrayLevelEmphasis、wavelet.LLL_glcm_JointEntropy为预测肺结节生长影像组学标签，建立影像组学评分Rad-score＝0.4535973-0.3413027×original_shape_MajorAxisLength-0.113593×log.sigma.1.0.mm.3D_glrlm_ShortRunLowGrayLevelEmphasis＋0.5150988×log.sigma.5.0.mm.3D_firstorder_90Percentile-0.7143545×wavelet.LHH_gldm_LargeDependenceHighGrayLevelEmphasis＋0.10314×wavelet.LLL_glcm_JointEntropy。多因素Logistic回归分析显示，年龄大(OR＝1.066，95%CI：1.018~1.117)、女性(OR＝0.313，95%CI：0.113~0.870)、结节类型为实性(OR＝0.495，95%CI：0.279~0.878)、Rad-score高(OR＝1.038，95%CI：1.020~1.057)为肺结节生长危险因素。训练集组合模型(年龄、性别、结节类型、Rad-score)ROC曲线下面积(area under the curve, AUC)0.871(95%CI：0.804~0.939)，验证集AUC 0.873(95%CI：0.776~0.971)。Hosmer-Lemeshow检验显示训练集P＝0.664，验证集P＝0.502，组合模型预测概率与实际概率偏差无统计学意义。决策曲线分析显示，训练集高风险阈值0.60~0.95，验证集0.60~0.93，组合模型具有临床应用价值。

结论

结合胸部HRCT影像组学特征和年龄、性别、结节类型模型可预测肺结节短期生长风险，有助于临床制对肺结节的判断和治疗策略。

关键词: 单发肺结节, 计算机断层扫描影像组学, 临床特征, 短期生长, 列线图

Objective

To explore the clinical significance of computed tomography (CT) radiomics and clinical features in predicting the growth of solitary pulmonary nodules.

Methods

A total of 165 patients with solitary pulmonary nodules shown by chest high-resolution computed tomography (HRCT) admitted to our hospital from May 2021 to May 2023 were selected as subjects. They were randomly divided into a training set (115 cases) and a validation set (50 cases) at a ratio of 7: 3. Radiomics feature variables based on CT images were extracted using the open-source Python package Pyradiomics. The growth of pulmonary nodules was observed during a 2-year follow-up. The maximum relevance minimum redundancy method and LASSO analysis were used to screen imaging feature variables, and a model was constructed. The receiver operating characteristic (ROC) curve was used to compare the predictive value of the model.

Results

Among 165 cases, there were 99 growing pulmonary nodules (60.00%) and 66 stable pulmonary nodules (40.00%). LASSO analysis identified 5 features (original_shape_MajorAxisLength, log.sigma.1.0.mm.3D_glrlm_ShortRunLowGrayLevelEmphasis, log.sigma.5.0.mm.3D_firstorder_90Percentile, wavelet.LHH_gldm_LargeDependenceHighGrayLevelEmphasis, wavelet.LLL_glcm_JointEntropy) as the radiomics signature for predicting pulmonary nodule growth, and a radiomics score was established (Rad-score=0.4535973-0.3413027×original_shape_MajorAxisLength-0.113593×log.sigma.1.0.mm.3D_glrlm_ShortRunLowGrayLevelEmphasis+ 0.5150988×log.sigma.5.0.mm.3D_firstorder_90Percentile-0.7143545×wavelet.LHH_gldm_LargeDependenceHighGrayLevelEmphasis+ 0.10314×wavelet.LLL_glcm_JointEntropy). Multivariate logistic regression analysis showed that older age (OR=1.066, 95%CI: 1.018~1.117), female gender (OR=0.313, 95%CI: 0.113~0.870), solid nodule type (OR=0.495, 95%CI: 0.279~0.878), and high Rad-score (OR=1.038, 95%CI: 1.020~1.057) were risk factors for pulmonary nodule growth. The combined model (age, gender, nodule type+ Rad-score) had good discriminative ability in both the training set and the validation set area under the curve(AUC) of ROC 0.871 (95%CI: 0.804~0.939) vs. 0.873 (95%CI: 0.776~0.971). The Hosmer-Lemeshow test results for the training set and validation set were P=0.664 and P=0.502, respectively, indicating no statistically significant difference between the predicted probability and the actual probability of the combined model. Decision curve analysis showed that the combined model had clinical application value within the high-risk threshold range training set 0.60~0.95, validation set 0.60~0.93.

Conclusion

The nomogram model combining chest HRCT-derived radiomics features with age, gender, and nodule type has the potential to accurately predict the short-term growth risk of pulmonary nodules. This model has shown certain clinical value in helping clinical doctors develop diagnosis and treatment strategies.

Key words: Single pulmonary nodule, Computed tomography radiomics, Clinical features, Short-term growth, Nomogram model

图1 肺结节患者典型胸部CT图。图A为右肺中叶外侧段Se3，Im(291-307，0.5 mm)见部分实性结节，实性占比约为47%，大小约9 mm×6 mm，平均CT值约-435.8HU；图B为随访2年后CT示结节大小约9 mm×6 mm

表1 训练集肺结节患者临床资料结果比较

临床资料	生长型肺结节(n＝69)	稳定型肺结节(n＝46)	Z/t/χ²	P值
年龄[岁，(±s)]	61.54±10.16	54.59±10.77	3.508	0.001
性别[n(%)]			6.763	0.009
女	38(55.07)	14(30.43)
男	31(44.93)	32(69.57)
吸烟史[n(%)]	21(30.43)	17(36.96)	0.531	0.466
结节最大直径[mm，M₅₀(P₂₅，P₇₅)]	22.00(18.00，26.00)	22.00(19.00，27.00)	0.611	0.542
结节类型[n(%)]			8.387	0.015
实性	14(20.29)	21(45.65)
部分实性	13(18.84)	6(13.04)
磨玻璃	42(60.87)	19(41.30)
结节位置[n(%)]			2.236	0.693
右肺上叶	20(28.99)	15(32.61)
右肺中叶	2(2.90)	3(6.52)
右肺下叶	12(17.39)	6(13.04)
左肺上叶	18(26.09)	17(36.96)
左肺下叶	13(18.84)	7(15.22)
胸部CT征象[n(%)]
毛刺征	10(14.49)	8(17.39)	0.176	0.675
分叶	22(31.88)	21(45.65)	2.235	0.135
空泡征	14(20.29)	8(17.39)	0.150	0.699
胸膜凹陷征	25(36.23)	19(41.30)	0.301	0.584
空气支气管征	15(21.74)	9(19.57)	0.079	0.779
钙化	5(7.25)	3(6.52)		0.999^a
血管集束征	12(17.39)	6(13.04)	0.395	0.530
边界模糊	16(23.19)	14(30.43)	0.752	0.386

表1 训练集肺结节患者临床资料结果比较

临床资料	生长型肺结节(n＝69)	稳定型肺结节(n＝46)	Z/t/χ²	P值
年龄[岁，(±s)]	61.54±10.16	54.59±10.77	3.508	0.001
性别[n(%)]			6.763	0.009
女	38(55.07)	14(30.43)
男	31(44.93)	32(69.57)
吸烟史[n(%)]	21(30.43)	17(36.96)	0.531	0.466
结节最大直径[mm，M₅₀(P₂₅，P₇₅)]	22.00(18.00，26.00)	22.00(19.00，27.00)	0.611	0.542
结节类型[n(%)]			8.387	0.015
实性	14(20.29)	21(45.65)
部分实性	13(18.84)	6(13.04)
磨玻璃	42(60.87)	19(41.30)
结节位置[n(%)]			2.236	0.693
右肺上叶	20(28.99)	15(32.61)
右肺中叶	2(2.90)	3(6.52)
右肺下叶	12(17.39)	6(13.04)
左肺上叶	18(26.09)	17(36.96)
左肺下叶	13(18.84)	7(15.22)
胸部CT征象[n(%)]
毛刺征	10(14.49)	8(17.39)	0.176	0.675
分叶	22(31.88)	21(45.65)	2.235	0.135
空泡征	14(20.29)	8(17.39)	0.150	0.699
胸膜凹陷征	25(36.23)	19(41.30)	0.301	0.584
空气支气管征	15(21.74)	9(19.57)	0.079	0.779
钙化	5(7.25)	3(6.52)		0.999^a
血管集束征	12(17.39)	6(13.04)	0.395	0.530
边界模糊	16(23.19)	14(30.43)	0.752	0.386

表2 肺结节生长多因素Logistic回归分析

影响因素	β	SE	Walds值	OR(95%CI)	P值
年龄(连续变量，每增加1岁)	0.064	0.024	7.387	1.066(1.018~1.117)	0.007
性别(0＝女，1＝男)	1.160	0.521	4.957	0.313(0.113~0.870)	0.026
结节类型(1＝磨玻璃，2＝部分实性，3＝实性)	0.703	0.292	5.780	0.495(0.279~0.878)	0.016
Rad-score(每增加1个单位)	0.037	0.009	17.312	1.038(1.020~1.057)	0.000

表2 肺结节生长多因素Logistic回归分析

影响因素	β	SE	Walds值	OR(95%CI)	P值
年龄(连续变量，每增加1岁)	0.064	0.024	7.387	1.066(1.018~1.117)	0.007
性别(0＝女，1＝男)	1.160	0.521	4.957	0.313(0.113~0.870)	0.026
结节类型(1＝磨玻璃，2＝部分实性，3＝实性)	0.703	0.292	5.780	0.495(0.279~0.878)	0.016
Rad-score(每增加1个单位)	0.037	0.009	17.312	1.038(1.020~1.057)	0.000

图2 组合模型列线图

表3 肺结节生长预测的ROC曲线分析

预测模型	AUC(95%CI)	精确度(95%CI)	灵敏度(95%CI)	特异度(95%CI)	PPV(95%CI)	NPV(95%CI)	最佳截断值
临床模型	0.786(0.703~0.870)	0.704(0.612~0.786)	0.870(0.772~0.967)	0.594(0.478~0.710)	0.588(0.471~0.705)	0.872(0.777~0.968)	0.695
CT影像组学模型	0.801(0.719~0.883)	0.748(0.658~0.824)	0.761(0.638~0.884)	0.739(0.636~0.843)	0.660(0.533~0.788)	0.823(0.727~0.918)	35.136
组合模型	0.871(0.804~0.939)	0.817(0.735~0.883)	0.848(0.744~0.952)	0.797(0.702~0.892)	0.736(0.617~0.855)	0.887(0.808~0.966)	0.650

表3 肺结节生长预测的ROC曲线分析

预测模型	AUC(95%CI)	精确度(95%CI)	灵敏度(95%CI)	特异度(95%CI)	PPV(95%CI)	NPV(95%CI)	最佳截断值
临床模型	0.786(0.703~0.870)	0.704(0.612~0.786)	0.870(0.772~0.967)	0.594(0.478~0.710)	0.588(0.471~0.705)	0.872(0.777~0.968)	0.695
CT影像组学模型	0.801(0.719~0.883)	0.748(0.658~0.824)	0.761(0.638~0.884)	0.739(0.636~0.843)	0.660(0.533~0.788)	0.823(0.727~0.918)	35.136
组合模型	0.871(0.804~0.939)	0.817(0.735~0.883)	0.848(0.744~0.952)	0.797(0.702~0.892)	0.736(0.617~0.855)	0.887(0.808~0.966)	0.650

1	王小平，赵会玲，敖文安. HRCT鉴别孤立性肺结节性质的价值[J]. 现代医用影像学，2024, 33(3): 419-422.
2	Li TZ, Xu K, Krishnan A, et al. Performance of lung cancer prediction models for screening-detected, incidental, and biopsied pulmonary nodules[J]. Radiol Artif Intell, 2025, 7(2): e230506.
3	Liu H, Yao X, Xu B, et al. Efficacy and safety analysis of multislice spiral CT-guided transthoracic lung biopsy in the diagnosis of pulmonary nodules of different sizes[J]. Comput Math Methods Med, 2022, 2022: 8192832.
4	Kim J, Dabiri B, Hammer MM. Micronodular lung disease on high-resolution CT: patterns and differential diagnosis[J]. Clin Radiol, 2021, 76(6): 399-406.
5	Zhu L, Liu J, Zeng L, et al. Thoracic high resolution computed tomography evaluation of imaging abnormalities of 108 lung cancer patients with different pulmonary function[J]. Cancer Imaging, 2024, 24(1): 78.
6	Desai SR, Sivarasan N, Johannson KA, et al. High-resolution CT phenotypes in pulmonary sarcoidosis: a multinational Delphi consensus study[J]. Lancet Respir Med, 2024, 12(5): 409-418.
7	Qi SA, Wu Q, Chen Z, et al. High-resolution metabolomic biomarkers for lung cancer diagnosis and prognosis[J]. Sci Rep, 2021, 11(1): 11805.
8	Yao Y, Wang X, Guan J, et al. Metabolomic differentiation of benign vs malignant pulmonary nodules with high specificity via high-resolution mass spectrometry analysis of patient sera[J]. Nat Commun, 2023, 14(1): 2339.
9	Gambardella C, Messina G, Pica DG, et al. Intraoperative lung ultrasound improves subcentimetric pulmonary nodule localization during VATS: Results of a retrospective analysis[J]. Thorac Cancer, 2023, 14(25): 2558-2566.
10	Yang R, Hui D, Li X, et al. Prediction of single pulmonary nodule growth by CT radiomics and clinical features-a one-year follow-up study[J]. Front Oncol, 2022, 12：1034817.
11	van GriethuysenJJM, Fedorov A, Parmar C, et al. Computational radiomics system to decode the radiographic phenotype[J]. Cancer Res, 2017, 77(21): e104-e107.
12	苏大同，王颖. 肺结节CT随访中的生长特性评估[J]. 中国肺癌杂志，2017, 20(8): 584-588.
13	Chen Y, Hou X, Yang Y, et al. A novel deep learning model based on multi-scale and multi-view for detection of pulmonary nodules[J]. J Digit Imaging, 2023, 36(2): 688-699.
14	Chen X, Zhu X, Yan W, et al. Serum lncRNA THRIL predicts benign and malignant pulmonary nodules and promotes the progression of pulmonary malignancies[J]. BMC Cancer, 2023, 23(1): 755.
15	Mankidy BJ, Mohammad G, Trinh K, et al. High risk lung nodule: A multidisciplinary approach to diagnosis and management[J]. Respir Med, 2023, 214: 107277.
16	Guerrini S, Del RoscioD, Zanoni M, et al. Lung cancer imaging: Screening result and nodule management[J]. Int J Environ Res Public Health, 2022, 19(4): 2460.
17	Tomassini S, Falcionelli N, Sernani P, et al. Lung nodule diagnosis and cancer histology classification from computed tomography data by convolutional neural networks: A survey[J]. Comput Biol Med, 2022, 146: 105691.
18	Farjah F, Monsell SE, Greenlee RT, et al. Patient and nodule characteristics associated with a lung cancer diagnosis among individuals with incidentally detected lung nodules[J]. Chest, 2023, 163(3): 719-730.
19	Wu Z, Wang F, Cao W, et al. Lung cancer risk prediction models based on pulmonary nodules: A systematic review[J]. Thorac Cancer, 2022, 13(5): 664-677.
20	Ping X, Jiang N, Meng Q, et al. Prediction of the benign or malignant nature of pulmonary pure ground-glass nodules based on radiomics analysis of high-resolution computed tomography images[J]. Tomography, 2024, 10(7): 1042-1053.
21	MacMahon H, Naidich DP, Goo JM, et al. Guidelines for management of incidental pulmonary nodules detected on CT images: From the fleischner society 2017[J]. Radiology, 2017, 284(1): 228-243.
22	Abbaspour E, Karimzadhagh S, Monsef A, et al. Application of radiomics for preoperative prediction of lymph node metastasis in colorectal cancer: A systematic review and meta-analysis[J]. Int J Surg, 2024, 110(6): 3795-3813.
23	Qi YJ, Su GH, You C, et al. Radiomics in breast cancer: Current advances and future directions[J]. Cell Rep Med, 2024, 5(9): 101719.
24	Warkentin MT, Al-Sawaihey H, Lam S, et al. Radiomics analysis to predict pulmonary nodule malignancy using machine learning approaches[J]. Thorax, 2024, 79(4): 307-315.
25	Heuvelmans MA, Walter JE, Vliegenthart R, et al. Disagreement of diameter and volume measurements for pulmonary nodule size estimation in CT lung cancer screening[J]. Thorax, 2018, 73(8): 779-781.
26	Murchison JT, Ritchie G, Senyszak D, et al. Validation of a deep learning computer aided system for CT based lung nodule detection, classification, and growth rate estimation in a routine clinical population[J]. PLoS One, 2022, 17(5): e0266799.
27	Zwanenburg A, Vallières M, Abdalah MA, et al. The image biomarker standardization initiative: Standardized quantitative radiomics for high-throughput image-based phenotyping[J]. Radiology, 2020, 295(2): 328-338.
28	蔡玉琳，牛丹，马丹，等. 不同CT重建算法对人工智能肺结节诊断的意义[J/OL]. 中华肺部疾病杂志(电子版), 2024, 17(6): 931-935.
29	Hou X, Wu M, Chen J, et al. Establishment and verification of a prediction model based on clinical characteristics and computed tomography radiomics parameters for distinguishing benign and malignant pulmonary nodules[J]. J Thorac Dis, 2024, 16(3): 1984-1995.
30	Yang J, Yee PL, Khan AA, et al. Intelligent lung cancer MRI prediction analysis based on cluster prominence and posterior probabilities utilizing intelligent Bayesian methods on extracted gray-level co-occurrence (GLCM) features[J]. Digit Health, 2023, 9: 20552076231172632.

[1]	董凯华, 姚艳青, 苗敏. Omicron变异株流行期间不同年龄段儿童感染新型冠状病毒的临床特征及核酸转阴时间影响因素[J/OL]. 中华实验和临床感染病杂志(电子版), 2025, 19(06): 353-361.
[2]	徐世伟, 廖杜荣, 张镐, 叶辉, 陈志平, 雒洪志. 基于一种新炎症-营养指标构建结直肠癌术前淋巴结转移预测模型[J/OL]. 中华普通外科学文献(电子版), 2025, 19(06): 383-389.
[3]	张剑飞, 沈鹤, 邱建宏, 赵新鸿. 男性和女性腺性膀胱炎临床特征及复发率的比较[J/OL]. 中华腔镜泌尿外科杂志(电子版), 2025, 19(05): 628-632.
[4]	腾鹏, 田景昌, 鄂春翔, 向阳, 田伯宇. 肌层浸润性膀胱癌患者根治性膀胱切除术预后列线图模型的构建及验证[J/OL]. 中华腔镜泌尿外科杂志(电子版), 2025, 19(05): 645-652.
[5]	严一杰, 张军, 孟繁杰, 关志宇. 肺结节-胸膜关系预测CT引导下肺穿刺活检后气胸风险的作用研究[J/OL]. 中华肺部疾病杂志(电子版), 2026, 19(01): 36-41.
[6]	高瞻, 唐莎莎, 秦蘅, 王关嵩, 张雯. 慢性阻塞性肺疾病合并肺癌的临床特征及危险因素分析[J/OL]. 中华肺部疾病杂志(电子版), 2025, 18(05): 685-690.
[7]	黄少坚, 梁汉标, 李清平, 唐善华, 李青妍, 李芷西, 黄灿, 王小振, 陈灿辉, 王恺, 李川江. 基于影像组学和临床特征构建肝癌新辅助/转化治疗后病理学完全缓解预测模型[J/OL]. 中华肝脏外科手术学电子杂志, 2025, 14(06): 860-867.
[8]	赵俊宇, 林航宇, 李会灵, 王显飞, 游川. 肝癌肝切除术后大量腹水预测模型的建立与验证[J/OL]. 中华肝脏外科手术学电子杂志, 2025, 14(05): 740-747.
[9]	李毅, 姚略, 侯相见. 颅脑损伤去骨瓣减压术后新发出血的风险预测列线图模型构建[J/OL]. 中华神经创伤外科电子杂志, 2025, 11(05): 307-313.
[10]	慈娟娟, 吴俊成, 朱琴琴. 肝硬化合并食管胃静脉曲张破裂出血患者内镜治疗后再出血风险列线图的构建与验证[J/OL]. 中华消化病与影像杂志(电子版), 2026, 16(01): 47-52.
[11]	刘燕, 马亦旻. 消化道早癌患者内镜黏膜下剥离术后局部复发预测模型构建与验证[J/OL]. 中华消化病与影像杂志(电子版), 2025, 15(06): 576-582.
[12]	赵莉, 张敏伟, 郭浩东, 于海侠, 王光明, 李德春. 乳腺X线模型构建列线图预测乳腺癌HR及HER2的表达[J/OL]. 中华临床医师杂志(电子版), 2025, 19(10): 747-757.
[13]	王超, 张晓会, 李晓帆, 赵海丹. 维持性血液透析患者感染与心脑血管死亡风险的比较及联合预测模型构建[J/OL]. 中华临床医师杂志(电子版), 2025, 19(09): 675-681.
[14]	于少华, 苏飞, 芦永斌, 袁芳芸, 阚晓燕, 张涛, 侯小明. 基于血清电解质水平构建列线图模型对广泛期小细胞肺癌患者的预测价值[J/OL]. 中华临床医师杂志(电子版), 2025, 19(08): 574-581.
[15]	皇立媛, 浦洁, 王苏贵, 陈婷婷, 朱德慧, 胡雪. 中青年脑卒中患者应激障碍风险预测模型的构建与验证[J/OL]. 中华临床医师杂志(电子版), 2025, 19(07): 504-512.

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

Clinical significance of predicting the growth of solitary pulmonary nodules based on CT radiomics and clinical features

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

Clinical significance of predicting the growth of solitary pulmonary nodules based on CT radiomics and clinical features