CT空间分辨率对影像组学模型鉴别肺磨玻璃结节浸润性效能的影响
| 作者: |
1房坚,
2孙炎冰,
2陶广昱
1 上海市闵行区虹桥社区卫生服务中心，上海 201103 2 上海市胸科医院，上海交通大学附属胸科医院放射科，上海 200030 |
| 通讯: |
孙炎冰
Email: xkradiology@163.com |
| DOI: | 10.3978/j.issn.2095-6959.2022.05.024 |
| 基金: | 国家自然科学基金(81871353，82071873)；上海市卫生与计划生育委员会科研课题(20184Y0219)；上海市数字媒体处理与传输重点实验室开放课题(STCSM18DZ2270700)；上海市卫生健康委先进适宜技术推广项目(2019SY063)；徐汇区人工智能医疗院地合作项目(2020-010)。 |
摘要
目的：探讨超高分辨率计算机断层扫描(ultra high resolution computed tomography，UHRCT)靶扫描技术对肺磨玻璃结节(ground-glass nodule，GGN)浸润性鉴别的影像组学模型的效能影响。方法：回顾性分析1 101例GGN患者的UHRCT资料。分别使用常规扫描数据和靶扫描数据建立logistic回归模型并比较二者效能。结果：基于常规扫描数据和基于靶扫描数据的影像组学模型的曲线下面积、正确率、敏感度和特异度分别为0.80 vs 0.83、78.7% vs 82.3%、81.6% vs 85.7%和60.7% vs 61.3%，其中基于靶扫描CT数据的组学模型在预测GGN的浸润性方面明显优于常规扫描下的组学模型(DeLong检验，P=0.01)。结论：超高分辨率靶扫描CT可以提高组学模型鉴别GGN浸润性的效能。
关键词：
计算机断层扫描；磨玻璃结节；浸润性；影像组学
Effect of CT spatial resolution on the efficacy of imaging omics model in distinguishing lung ground-glass nodules for infiltration
CorrespondingAuthor: SUN Yanbing Email: xkradiology@163.com
DOI: 10.3978/j.issn.2095-6959.2022.05.024
Foundation: This work was supported by the National Natural Science Foundation (81871353, 82071873), Shanghai Municipal Commission of Health and Family Planning Project (20184Y0219), Shanghai Key Laboratory Open Project (STCSM18DZ2270700), Advanced Appropriate Technology Promotion Project of Shanghai Municipal Health Commission (2019SY063), and Xuhui District Artificial Intelligence Medical Hospital Cooperation Project (2020-010), China.
Abstract
Objective: The purpose of this study was to investigate the effect of ultra-high-resolution computed tomography (CT) scanning technique on the efficacy of a radiomics model in distinguishing the invasiveness of lung ground-glass nodules (GGNs). Methods: The ultra-high-resolution CT (UHRCT) data of 1 101 GGN patients were retrospectively analyzed. Several logistic regression models were established using routine scan data and target scan data, and their efficacy was compared. Results: The area under the curve, accuracy, sensitivity, and specificity on the routine data set were 0.80, 0.78, 0.81, and 0.60, respectively, while the area under the curve, accuracy, sensitivity, and specificity on the target scanning data set were 0.83, 0.82, 0.86, and 0.61, respectively. The radiomics model based on target scan images was significantly better at predicting the invasiveness of GGN than when based on the radiomics features of routine CT (P=0.01 by DeLong test). Conclusion: Reducing the pixel size through the use of the target scanning method can improve the efficiency of the radiomics model in determining the invasiveness of GGNs.
Keywords：
computed tomography; ground-glass nodule; invasiveness; radiomics