| 朱红伟,董晓雷,马士华.基于 CT灌注成像与纹理分析技术建立孤立性肺结节良恶性预测模型[J].安徽医药,2025,29(7):1427-1431. |
| 基于 CT灌注成像与纹理分析技术建立孤立性肺结节良恶性预测模型 |
| Prediction model for benign and malignant isolated pulmonary nodules was established based on CT perfusion imaging and texture analysis |
| |
| DOI：10.3969/j.issn.1009-6469.2025.07.033 |
| 中文关键词: 孤立性肺结节 良性肿瘤 恶性肿瘤 CT灌注成像 纹理分析技术 预测模型 |
| 英文关键词: Isolated pulmonary nodules Benign tumor Malignant tumor CT perfusion imaging Texture analysis techniques Predictive models |
| 基金项目:保定市科技计划项目( 18ZF275) |
|
| 摘要点击次数: 240 |
| 全文下载次数: 62 |
| 中文摘要: |
| 目的分析基于 CT灌注成像与纹理分析技术建立孤立性肺结节( SPN)良恶性预测模型的应用价值。方法回顾 2020年 8月至 2022年 7月保定市第二医院收治的表现为 SPN并经病理诊断为恶性病变(恶性组， n=61)及良性病变(良性组， n=122)病人的临床资料，比较两组病人的基线资料、 CT灌注成像参数［血流量、血容量、通透性］、纹理分析参数(熵值、熵和、熵差、对比度、相关、均和)采用多因素 logistic回归分析法分析 SPN恶性的影响因素，并据此构建 logistic回归模型、绘制诺莫图，并进行模型内外部验证和，受试者操作特征曲线( ROC曲线)预测效能的评价。结果恶性组年龄［( 53.94±9.35)岁］、结节长径［( 12.06±2.15)mm］、血容量［( 6.13±1.10)mL/100 g］、血流量［( 142.37±31.35)mL·(100 g).1·min.1］、通透性［( 22.35±6.14)mL·(100 g).1 ·min.1］、熵值( 1.84±0.25)、熵和( 1.21±0.08)、熵差( 1.05±0.08)高于良性组［( 50.10±10.22)岁、(10.17±2.25)mm、(3.47± 0.50)mL/100 g、(75.38±26.84)mL·(100 g).1 ·min.1、(7.54±2.39)mL·(100 g).1 ·min.1、1.53±0.27、1.14±0.07、0.95±0.07］(P<0.05)；恶性组与良性组对比度、相关、均和比较，差异无统计学意义( P>0.05)。多因素 logistic回归分析显示，结节长径较长、血容量、血流量、通透性、熵值、熵差升高均是导致 SPN恶性结节的独立危险因素( P<0.05)构建模型方程， logit(p)=.0.291+0.551×结节长径 +0.912×血容量 +0.653×血流量 +1.134×通透性 +0.857×熵值 +1.311×熵差。利用本，次数据资料(恶性 61例，良性 122例)进行模型的内部验证，绘制 ROC曲线结果显示， ROC曲线下面积( AUC)为 0.94(P<0.05)，灵敏度 91.80%、特异度 82.79%、约登指数 0.75；校准曲线与理想曲线拟合良好( P>0.05)。将风险模型应用于 2022年 8月至 2023年 4月该院收治的 66例 SPN结节(恶性 21枚、良性 45枚)病人进行外部验证，绘制 ROC曲线结果显示 AUC为 0.88(P<0.05)，灵敏度 90.48%、特异度 86.67%、约登指数 0.77；绘制 Calibration曲线结果显示校准曲线与理想曲线拟合良好( P>0.05)。结论结节长径较长、血容量、血流量、通透性、熵值、熵差升高是预测 SPN恶性的危险因素，据此构建的风险模型有较高区分度、拟合度和真实度，绘制的可视化诺莫图对临床判断 SPN良恶性有重要参考价值。 |
| 英文摘要: |
| Objective To analyze the application value of establishing a benign and malignant prediction model for isolated pulmo-nary nodules (SPN) based on CT perfusion imaging and texture analysis techniques.Methods The clinical data of patients who pre-sented with SPN and were pathologically diagnosed as malignant lesions (malignant group, n=61) and benign lesions (benign group, n= 122) admitted to Baoding Second Hospital from August 2020 to July 2022 were reviewed. Baseline data, CT perfusion imaging parame-ters [blood flow, blood volume, permeability], and texture analysis parameters (entropy value, entropy sum, entropy difference, contrast,correlation, mean sum) were compared between the two groups. Multifactor logistic regression analysis was used to analyse the influenc-ing factors of SPN malignancy. And accordingly, Logistic regression model was constructed, Nomogram was drawn, and internal and ex-ternal validation of the model and evaluation of the predictive efficacy of the subjects' job characteristics curves (ROC curves) were car-ried out.Results In the malignant group, age [(53.94±9.35) years], nodule length [(12.06±2.15) mm], blood volume [(6.13±1.10) mL/100 g], blood flow [(142.37±31.35) mL·(100 g).1·min.1], permeability [(22.35±6.14) mL·(100 g).1·min.1], entropy (1.84±0.25), entropysum (1.21±0.08), entropy difference (1.05±0.08) were higher than those in benign group [(50.10±10.22) years, (10.17±2.25) mm, (3.47±0.50) mL/100 g, (75.38±26.84) mL·(100 g).1·min.1, (7.54±2.39) mL·(100 g).1·min.1, 1.53±0.27, 1.14±0.07, 0.95±0.07] (P < 0.05). Thedifferences in contrast, correlation and mean sum between the malignant and benign groups were not statistically significant (P > 0.05).Multivariate logistic regression analysis showed that longer nodule length, increased blood volume, blood flow, permeability, entropyvalue and entropy difference were all independent risk factors for malignant nodules in SPN (P < 0.05). The model equation was con-structed with logit (p) = .0.291 + 0.551 × nodule length + 0.912 × blood volume + 0.653 × blood flow + 1.134 × permeability + 0.857 ×entropy value + 1.311 × entropy difference. The internal validation of the model was carried out using the current dataset (61 malignantand 122 benign cases), and the results of plotting the ROC curve showed that the area under the ROC curve (AUC) was 0.94 (P<0.05),with a sensitivity of 91.80%, specificity of 82.79% and Yorden index of 0.75, and the calibration curve fitted well with the ideal curve(P > 0.05). The model was applied to 66 patients with SPN nodules (21 malignant and 45 benign) admitted from August 2022 to April2023 for external validation, and the results of plotting the ROC curve showed that the AUC was 0.88 (P<0.05), sensitivity was 90.48%,specificity was 86.67%, and Yodan index was 0.77. Calibration curve showed that the calibration curve fitted well with the ideal curve(P > 0.05).Conclusion Longer nodal length, increased blood volume, blood flow, permeability, entropy value and entropy differenceare risk factors for predicting the malignancy of SPN. The risk model constructed accordingly has high discrimination, goodness of fitand realism, and the visualized Nomogram drawn is an important reference value for clinical judgment of benign and malignant SPN. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
