Differentiation of nodule or mass type pulmonary cryptococcosis and lung adenocarcinoma, lung tuberculosis based on plain CT scanning radiomics models
10.13929/j.issn.1003-3289.2020.06.011
- VernacularTitle: 基于CT平扫影像组学模型鉴别结节/肿块型肺隐球菌病及肺腺癌与肺结核
- Author:
Mengsi FAN
1
Author Information
1. Department of Radiology, The Second Affiliated Hospital of Anhui Medical University
- Publication Type:Journal Article
- Keywords:
Artificial intelligence;
Cryptococcosis;
Lung;
Lung neoplasms;
Radiomics;
Tomography, X-ray computed;
Tuberculosis
- From:
Chinese Journal of Medical Imaging Technology
2020;36(6):853-857
- CountryChina
- Language:Chinese
-
Abstract:
Objective: To explore the feasibility of differential diagnosis of nodule or mass pulmonary cryptococcosis (PC), lung adenocarcinoma and lung tuberculosis (TB) based on plain CT scanning radiomics prediction models. Methods: Plain CT data of 28 patients with nodule or mass type PC, 30 with pulmonary adenocarcinoma and 26 with lung TB were retrospectively analyzed. The texture features of lesions on CT images were extracted and selected to establish the optimized texture parameters between PC and lung adenocarcinoma, also between PC and lung TB. Then all samples were divided into training set and testing set according to ratio of 7:3. The random forest method was used to establish prediction model with the optimized texture parameters, and the model was used to evaluate training set data and verified with testing set data. The corresponding ROC curve was drawn, so as to evaluate the model's differential diagnosis efficiency. Results: After screening, 7 optimized feature parameters were obtained between PC and lung adenocarcinoma, while 4 were obtained between PC and lung TB. The AUC, sensitivity, specificity, accuracy of the model for differentiating PC from lung adenocarcinoma was 0.96, 1.00, 0.78 and 0.89,respectively, while for differentiating PC from lung TB was 0.99, 0.88, 0.89 and 0.88, respectively. Conclusion: Radiomics models based plain CT scanning can be used for differentiating and diagnosing nodule or mass PC from lung adenocarcinoma and lung TB.
