Establishment and Verification of Benign and Malignant Prediction Model of
Subcentimeter Pulmonary Ground Glass Nodules Based on HRCT.
10.3779/j.issn.1009-3419.2023.101.15
- Author:
Zhengwei CHEN
1
;
Gaoxiang WANG
2
;
Hanran WU
2
;
Mingsheng WU
2
;
Xianning WU
2
;
Meiqing XU
2
;
Mingran XIE
2
Author Information
1. Wannan Medical College, Wuhu 241001, China.
2. Department of Thoracic Surgery, The First Affiliated Hospital of University of
Science and Technology of China, Hefei 230001, China.
- Publication Type:Journal Article
- Keywords:
Benign and malignant lesions;
High resolution computed tomography;
Nomogram;
Prediction model;
Subcentimeter pulmonary ground glass nodules
- MeSH:
Humans;
Retrospective Studies;
Lung Neoplasms/surgery*;
Adenocarcinoma;
China;
Hospitals;
Multiple Pulmonary Nodules
- From:
Chinese Journal of Lung Cancer
2023;26(5):377-385
- CountryChina
- Language:Chinese
-
Abstract:
BACKGROUND:Pre-operative accuracy of subcentimeter ground glass nodules (SGGNs) is a difficult problem in clinical practice, but there are few clinical studies on the benign and malignant prediction model of SGGNs. The aim of this study was to help identify benign and malignant lesions of SGGNs based on the imaging features of high resolution computed tomography (HRCT) and the general clinical data of patients, and to build a risk prediction model.
METHODS:This study retrospectively analyzed the clinical data of 483 patients with SGGNs who underwent surgical resection and were confirmed by histology from the First Affiliated Hospital of University of Science and Technology of China from August 2020 to December 2021. The patients were divided into the training set (n=338) and the validation set (n=145) according to 7:3 random assignment. According to the postoperative histology, they were divided into adenocarcinoma group and benign lesion group. The independent risk factors and models were analyzed by univariate analysis and multivariate Logistic regression. The receiver operator characteristic (ROC) curve was constructed to evaluate the model differentiation, and the calibration curve was used to evaluate the model consistency. The clinical application value of the decision curve analysis (DCA) evaluation model was drawn, and the validation set data was substituted for external verification.
RESULTS:Multivariate Logistic analysis screened out patients' age, vascular sign, lobular sign, nodule volume and mean-CT value as independent risk factors for SGGNs. Based on the results of multivariate analysis, Nomogram prediction model was constructed, and the area under ROC curve was 0.836 (95%CI: 0.794-0.879). The critical value corresponding to the maximum approximate entry index was 0.483. The sensitivity was 76.6%, and the specificity was 80.1%. The positive predictive value was 86.5%, and the negative predictive value was 68.7%. The benign and malignant risk of SGGNs predicted by the calibration curve was highly consistent with the actual occurrence risk after sampling 1,000 times using Bootstrap method. DCA showed that patients showed a positive net benefit when the predictive probability of the predicted model probability was 0.2 to 0.9.
CONCLUSIONS:Based on preoperative medical history and preoperative HRCT examination indicators, the benign and malignant risk prediction model of SGGNs was established to have good predictive efficacy and clinical application value. The visualization of Nomogram can help to screen out high-risk groups of SGGNs, providing support for clinical decision-making.
