Evaluation of the Invasion of Pulmonary Subsolid Nodules by the Artificial Intelligence Volumetric Density Method
-
摘要: 目的：探讨人工智能（AI）体积密度法判断肺亚实性结节（SSNs）浸润性的价值。方法：回顾性分析106例患者的108枚SSNs的CT和病理结果，将结节分为腺体前驱病变组和腺癌组。通过肺结节AI软件测量并比较两组的最大CT值、最小CT值、平均CT值、峰度、偏度、Perc.25％、Perc.50％、Perc.75％、Perc.95％、结节体积、结节平均径等CT定量参数。使用Medcalc软件得出受试者工作特征曲线（ROC），评价诊断SSNs浸润性的敏感度、特异度、阳性预测值及阴性预测值，用逻辑回归分析评估他们的诊断性能。结果：SSNs的多数CT定量参数差异存在统计学意义，其中，诊断效能最高的是Perc.25％，AUC达0.797；其次为Perc.50％ 和平均CT值，AUC均为0.787。Logistic回归分析显示，将诊断效能最高的Perc.25％ 分别与Perc.50％ 和平均CT值两两建立联合诊断模型1，其中Perc.25％ 与平均CT值的模型诊断效能最高，且联合诊断模型诊断效能高于Perc.25％ 与平均CT值单独的诊断效能。Medcalc软件分析显示，Perc.25％≥-578 HU和平均CT值≥-468 HU的SSNs病理表现为腺癌的可能性大。将Perc.25％ 与结节平均径结合，可获得对判断SSNs浸润性非常有价值的联合诊断模型2。结论：AI体积密度法对SSNs的浸润性有较高的诊断价值，联合使用Perc.25％ 与平均CT值比单独使用更能准确地判断浸润性。Abstract: Objective: To explore the value of the artificial intelligence (AI) volumetric density method in determining the invasion of pulmonary hyposolid nodules (SSNs). Methods: A total of 108 SSNs and the pathological results of 106 patients were reviewed, and these were divided into a glandular prodromal lesions group and an adenocarcinoma group. Pulmonary nodule AI software was used to measure and compare the CT quantitative parameters of the two groups, including the maximum CT value, minimum CT value, average CT value, kurtosis, skewness, Perc.25%, Perc.50%, Perc.75%, Perc.90%, nodule volume, and mean nodule diameter. Moreover, a receiver operating characteristic curve (ROC) was obtained by MedCalc software to evaluate the sensitivity, specificity, positive predictive value, and negative predictive value for the diagnosis of SSN infiltration, and their diagnostic performance was evaluated by logistic regression analysis. Results: There were significant differences in most CT quantitative parameters of SSNs. The highest diagnostic efficiency was Perc.25% and the AUC was 0.797, while the AUC was 0.787 for Perc.50% and the mean CT value. Logistic regression analysis showed that Perc.25% with the highest diagnostic efficiency was combined with Perc.50% and the mean CT value. The model with Perc.25% and the mean CT value had the highest diagnostic efficiency, and the combined diagnostic model had a higher diagnostic efficiency than Perc.25% and the mean CT value alone. According to MedCalc software, SSNs with Perc.25% ≥−578 HU and mean CT values ≥ −468 HU were more likely to be in the adenocarcinoma group. In this study, Perc.25% was combined with the mean diameter of nodules, and a very valuable combined diagnostic model II was obtained to judge the infiltration of SSNs. Conclusion: The AI volume density method has a high diagnostic value for SSN invasion. Moreover , the combination of Perc.25% and mean CT value can accurately judge the invasion than the use of average CT value alone, providing a quantitative basis for the clinical management of SSNs.
-
Keywords:
- CT /
- artificial intelligence /
- volume density method /
- pulmonary subsolid nodules
-
-
![]()
图 1 女，48岁，CT显示右肺上叶后段SSN（蓝色箭头）。结节平均经约9.6 mm，术后病理诊断为AIS
Figure 1. A 48-year-old woman with SSN (blue arrow) of the posterior upper lobe of the right lung. The average length of the nodules was approximately 9.6 mm on CT, and the postoperative pathological diagnosis was AIS
![]()
图 2 腺体前驱病变组与腺癌组的灰度直方图参数ROC图，其中Perc.25％ 诊断效能最佳
Figure 2. The gray histogram parameters of the adenocar- cinoma group and the glandular precursor lesion group were ROC plots, in which Perc.25％ had the best diagnostic efficiency
![]()
图 3 腺体前驱病变组与腺癌组的Perc.25％、平均CT值及结节平均直径联合模型的ROC比较
Figure 3. Perc.25％, mean CT value and ROC comparison between the glandular precursors and adenocarcinoma groups
表 1 腺体前驱病变组与腺癌组临床资料比较
Table 1 Comparison of clinical data between the glandular prodromal disease group and adenocarcinoma group
临床资料 组别 统计检验 腺体前驱病变组（25例） 腺癌组（83例） t/χ2 P 年龄/岁 60.93±8.72 60.78±9.68 0.088 0.930 性别 男 7 32 0.928 0.335 女 18 51 结节分布 右上 13 34 4.263 0.370 右中 1 5 右下 5 14 左上 3 19 左下 3 11 结节类型 pGGNs 5 13 0.260 0.760 mGGNs 20 70 
下载: 导出CSV
表 2 腺体前驱病变组、腺癌组的CT值分布直方图纹理参数及结节体积、结节平均径比较
Table 2 Comparison of the CT value distribution histogram texture parameters, nodule volume and mean nodule diameter between the glandular precursor lesion group and adenocarcinoma group
参数 组别 P 腺体前驱病变组（n＝25） 腺癌组（n＝83） 偏度/HU 0.45（0.25～0.70） 0.215（0.06～0.45） 0.001 峰度/HU -0.64（-1.03～-0.02） -0.95（-1.10～-0.53） 0.260 CT最大值/HU 22.00（-122.00～126.00） 167.00（41.00～338.00） <0.001 CT最小值/HU -757.00（-782.00～-711.00） -653.00（-752.25～-597.00） <0.001 平均CT值/HU -536.67±99.18 -390.16±165.59 <0.001 Perc.25％/HU -605.00（-659.00～-496.00） -416.00（-532.25～-233.50） <0.001 Perc.50％/HU -580.00（-650.00～-522.00） -418.50（-545.00～-296.00） <0.001 Perc.75％/HU -508.00（-673.00～-433.00） -449.00（-610.00～-208.75） 0.003 Perc.95％/HU -329.00（-722.00～-107.00） -528.50（-685.00～-5.25） 0.819 结节体积/mm3 646.86（278.79～1647.36） 1467.00（534.50～3270.94） <0.001 结节平均直径/mm 9.90（8.40～13.50） 15.65（11.12～19.65） <0.001
下载: 导出CSV
表 3 ROC分析结果
Table 3 ROC analysis results
参数 AUC 最佳临界值 敏感度/％ 特异性/％ 阳性预测
值/％阴性预测
值/％95％ CI 下限 上限 偏度 0.689 0.24 56.63 76.00 88.7 34.5 0.593 0.775 CT最大值/HU 0.725 39.00 77.11 64.00 87.7 45.7 0.631 0.807 CT最小值/HU 0.731 -692.00 62.65 88.00 94.5 41.5 0.737 0.812 平均CT值/HU 0.787 -468.00 72.29 84.00 97.9 47.7 0.697 0.860 Perc.25％/HU 0.797 -578.00 86.75 68.00 90.0 60.7 0.709 0.869 Perc.50％/HU 0.787 -500.00 71.08 84.00 93.7 46.7 0.698 0.860 Perc.75％/HU 0.678 -474.00 57.83 76.00 88.9 35.2 0.581 0.765 体积/mm3 0.701 777.38 73.49 64.00 87.1 42.1 0.606 0.785 平均径/mm 0.718 13.50 63.86 76.00 89.9 38.8 0.623 0.800 联合模型1 0.814 － 75.90 76.00 91.3 48.7 0.728 0.883 联合模型2 0.816 － 75.90 76.00 91.7 48.3 0.730 0.884
下载: 导出CSV
-
[1] SUNG H, FERLAY J, SIEGEL R L, et al. Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries[J]. CA: A Cancer Journal for Clinicians, 2021, 71(3): 209−249. DOI: 10.3322/caac.21660.
[2] ZHANG Y, FU F, CHEN H, et al. Management of ground-glass opacities in the lung cancer spectrum[J]. The Annals of Thoracic Surgery, 2020, 110(6): 1796−1804. DOI: 10.1016/j.athoracsur.2020.04.094.
[3] QIU T, RU X, YIN K, at al. Two nomograms based on CT features to predict tumor invasiveness of pulmonary adenocarcinoma and growth in pure GGN: A retrospective analysis[J]. Japanese Journal of Radiology, 2020, 38(8): 761−770. DOI: 10.1007/s11604-020-00957-x.
[4] SUN Y, LI C, JIN L, et al. Radiomics for lung adenocarcinoma manifesting as pure ground-glass nodules: Invasive prediction[J]. European Radiology, 2020, 30(7): 3650-3659. DOI: 10.1007/s00330-020-06776-y.
[5] ZHU N, ZHANG D, WANG W, et al. A novel coronavirus from patients with pneumonia in China, 2019[J]. The New England Journal of Medicine, 2020, 382(8): 727−733. DOI: 10.1056/NEJMoa2001017.
[6] LI X, ZHANG W, YU Y, et al. CT features and quantitative analysis of subsolid nodule lung adenocarcinoma for pathological classification prediction[J]. British Medical Council Cancer, 2020, 20(1): 60. DOI: 10.1186/s12885-020-6556-6.
[7] 李艳伶. 病灶直径和CT值联合评估浸润性肺腺癌的诊断价值[J]. 医学信息, 2021, 34(17): 172-174. LI Y L. Diagnosis value of invasive lung adenocarcinoma combined with tumor diameter and CT value[J]. Medical Informatics, 2021, 34(17): 172-174. (in Chinese).
[8] 虞梁, 王俊, 李洪, 等. 肺磨玻璃结节CT影像征象鉴别诊断肺浸润性腺癌与微浸润腺癌[J]. 南京医科大学学报(自然科学版), 2020,40(2): 248−251. YU L, WANG J, LI H, et al. Differential diagnosis of pulmonary infiltrating adenocarcinoma and microinfiltrating adenocarcinoma with CT features of ground glass nodules[J]. Journal of Nanjing Medical University (Natural Science), 2020, 40(2): 248−251. (in Chinese).
[9] 刘娜, 赵正凯, 邹佳瑜, 等. 基于人工智能的胸部CT肺结节检出及良恶性诊断效能评估[J]. CT理论与应用研究, 2021,30(6): 709−715. DOI: 10.15953/j.1004-4140.2021.30.06.06. LIU N, ZHAO Z K, ZOU J Y, et al. Detection of pulmonary nodules and evaluation of benign and malignant diagnosis in chest CT based on artificial intelligence[J]. CT Theory and Applications, 2021, 30(6): 709−715. DOI: 10.15953/j.1004-4140.2021.30.06.06. (in Chinese).
[10] 温德英, 潘雪琳, 姚辉, 等. 探讨CT扫描剂量对人工智能检测肺结节效能的影[J]. CT理论与应用研究, 2021,30(4): 455−465. DOI: 10.15953/j.1004-4140.2021.30.04.06. WEN D Y, PAN X L, YAO H, et al. Effects of CT scanning dose on the performance of artificial intelligence in detecting pulmonary nodules[J]. CT Theory and Applications, 2021, 30(4): 455−465. DOI: 10.15953/j.1004-4140.2021.30.04.06. (in Chinese).
[11] CHEN X, WEI X, ZHANG Z, et al. Differentiation of true-progression from pseudoprogression in glioblastoma treated with radiation therapy and concomitant temozolomide by GLCM texture analysis of conventional MRI[J]. Clinical Imaging, 2015, 39(5): 775-80.
[12] 蔡雅倩, 周小君, 张正华, 等. CT直方图分析鉴别肺良恶性纯磨玻璃结节的价值[J]. 放射学实践, 2020,35(8): 949−952. DOI: 10.13609/j.cnki.1000-0313.2020.08.001. CAI Y Q, ZHOU X J, ZHANG Z H, et al. Value of CT histogram analysis in differentiating lung benign and malignant pure ground glass nodules[J]. Practice Radiology, 2020, 35(8): 949−952. DOI: 10.13609/j.cnki.1000-0313.2020.08.001. (in Chinese).
[13] 毛海霞, 徐蒙莱, 冯银波, 等. CT图像纹理分析对小于10 mm纯磨玻璃结节侵袭性的诊断价值[J]. 中华临床医师杂志(电子版), 2019,13(5): 367−370. MAO H X, XU M L, FENG Y B, et al. Diagnostic value of CT image texture analysis for nodular invasivity in pure ground glass smaller than 10 mm[J]. Chinese Journal of Clinical Physicians (Electronic edition), 2019, 13(5): 367−370. (in Chinese).
[14] YAGI T, YAMAZAKI M, OHASHI R, et al. HRCT texture analysis for pure or part-solid ground-glass nodules: Distinguishability of adenocarcinoma in situ or minimally invasive adenocarcinoma from invasive adenocarcinoma[J]. Japanese Journal of Radiology, 2018, 36(2): 113-121. DOI: 10.1007/s11604-017-0711-2.
[15] 张宏, 丁必彪, 魏恒乐, 等. 高分辨率CT对肺纯磨玻璃结节侵袭性的预测价值[J]. 临床放射学杂志, 2019,38(3): 436−440. DOI: 10.13437/j.cnki.jcr.2019.03.017. ZHANG H, DING B B, WEI H L, et al. High resolution computed tomography (CT) for lung pure glass grinding nodules of invasive predictive value[J]. Journal of Clinical Radiology, 2019, 38(3): 436−440. DOI: 10.13437/j.cnki.jcr.2019.03.017. (in Chinese).
[16] 徐小东, 李君权, 吴向飞, 等. CT密度直方图对肺内纯磨玻璃样结节病理分级的预测价值[J]. 医学影像学杂志, 2021,31(1): 37−39, 42. XU X D, LI J Q, WU X F, et al. The prognostic value of CT density histogram in the pathological grading of pure ground glass nodules in lung[J]. Journal of Medical Imaging, 2021, 31(1): 37−39, 42. (in Chinese).
[17] KITAMI A, SANO F, HAYASHI S, et al. Correlation between histological invasiveness and the computed tomography value in pure ground-glass nodules[J]. Surgery Today, 2016, 46(5): 593−598. DOI: 10.1007/s00595-015-1208-1.
[18] IKEDA K, AWAI K, MORI T, et al. Differential diagnosis of ground-glass opacity nodules: CT number analysis by three-dimensional computerized quantification[J]. Chest, 2007, 132(3): 984−990. DOI: 10.1378/chest.07-0793.
[19] HAMMER M M, PALAZZO L L, KONG C Y, et al. Cancer risk in subsolid nodules in the national lung screening trial[J]. Radiology, 2019, 293(2): 441−448. DOI: 10.1148/radiol.
[20] WANG H, WENG Q, HUI J, et al. Value of TSCT features for differentiating preinvasive and minimally invasive adenocarcinoma from invasive adenocarcinoma presenting as subsolid nodules smaller than 3 cm[J]. Academic Radiology, 2020, 27(3): 395−403. DOI: 10.1016/j.acra.10.1016/j.acra.2019.05.005.
[21] 步玉兰, 李云, 戚元刚, 等. 纯磨玻璃密度结节高分辨率CT征象与病理组织学相关性研究[J]. 临床放射学杂志, 2018,5(2): 247−250. DOI: 10.13437/j.cnki.jcr.2018.02.016. BU Y L, LI Y, QI Y G, et al. Study on the relationship between density nodules and histopathology in high resolution CT of pure ground glass[J]. Journal of Clinical Radiology, 2018, 5(2): 247−250. DOI: 10.13437/j.cnki.jcr.2018.02.016. (in Chinese).
[22] LIU L H, LIU M, WEI R, at al. CT findings of persistent pure ground glass opacity: Can we predict the invasiveness?[J] Asian Pacific Journal of Cancer Prevention, 2015: 16(5): 1925-1928. DOI: 10.7314/apjcp.
-
期刊类型引用(6)
1. 谢玮，夏勇，杨骞，毕臣臣，吕慧，雷朝阳. 基于绕射波的地震属性研究进展. 石化技术. 2024(04): 216-218 . 
百度学术
2. 梁瑶，霍守东，李学良，舒梦珵，杨晓，石太昆. 利用绕射信息在裂缝型地层中进行钻前风险评估. 地球物理学报. 2023(01): 46-53 . 百度学术
3. 肖广锐，李尧，张羽茹，徐德奎. 绕射波成像在潜山裂缝储层预测中的应用——以渤中A气田为例. 石油物探. 2022(05): 812-820+829 . 百度学术
4. 田涛，李少轩，高阳，韦红. 变质岩潜山裂缝型储层精细预测技术——以渤海海域A油田为例. 石油地质与工程. 2022(06): 8-13 . 百度学术
5. 张志军，肖广锐，李尧. 渤中19-6油田变质岩潜山内幕裂缝地震响应特征及预测技术. 石油地球物理勘探. 2021(04): 845-852+675 . 百度学术
6. 周鹏，肖曦，陶杰，刘方，梁瑶，霍守东，舒梦珵. 绕射信息提取技术及其在致密砂岩断裂系统识别中的应用. 石油物探. 2020(02): 276-282 . 百度学术
其他类型引用(0)
计量
- 文章访问数: 320
- HTML全文浏览量: 161
- PDF下载量: 32
- 被引次数: 6
