Development of a machine learning model for predicting severe AECOPD based on non-contrast CT imaging of accessory respiratory muscles(PDF)
《中国医学物理学杂志》[ISSN:1005-202X/CN:44-1351/R]
- Issue:
- 2025年第7期
- Page:
- 892-900
- Research Field:
- 医学影像物理
- Publishing date:
Info
- Title:
- Development of a machine learning model for predicting severe AECOPD based on non-contrast CT imaging of accessory respiratory muscles
- Author(s):
- YE Zhe1; PAN Qiong2; GAO Shiyuan2; DAI Yakang3; GENG Chen3; LIAN Yixin2; YU Weibo1
- 1. School of Electrical and Electrical Engineering, Changchun University of Technology, Changchun 130012, China; 2. Department ofRespiratory and Critical Care Medicine, the Second Affiliated Hospital of Soochow University, Suzhou 215004, China 3. SuzhouInstitute of Biomedical Engineering and Technology, Chinese Academy of Sciences, Suzhou 215163, China
- Keywords:
- accessory respiratory muscle; non-contrast CT; radiomics; machine learning; severity stratification of acuteexacerbation of chronic obstructive pulmonary disease
- PACS:
- R318
- DOI:
- DOI:10.3969/j.issn.1005-202X.2025.07.008
- Abstract:
- Abstract: Regarding the challenge of early identification of critically ill patients with acute exacerbation of chronic obstructive pulmonary disease (AECOPD), a radiomics-clinical fusion model is proposed based on non-contrast CT images of accessory respiratory muscles to predict life-threatening conditions. A retrospective study is conducted involving 233 AECOPD patients (153 non-life-threatening and 80 life-threatening cases). Patients are divided into a training set (n=186)and a test set (n=47) at a 4:1 ratio. A total of 1 874 radiomic features are extracted from the erector spinae and pectoralis muscle regions delineated by radiologists on non-contrast CT images, and the features selection is performed using maximum relevance minimum redundancy and least absolute shrinkage and selection operator (LASSO) algorithms. Meanwhile, clinical data are analyzed with t-test and LASSO for variable screening. The selected features are input into C-support vector classification, Logistic regression, random forest, adaptive boosting (AdaBoost), and extreme gradient boosting (XGBoost) to construct radiomics model, clinical model, and fusion model. Predictive performance and clinical practicality are evaluated in the test set using receiver operating characteristic curve, area under the curve (AUC), and decision curve analysis. The radiomics-clinical fusion model built with XGBoost outperformed standalone radiomics and clinical models, achieving an AUC of 0.902 (95% CI 0.846, 0.994), with accuracy, sensitivity, specificity, and precision of 0.837, 0.933, 0.786, and 0.7, respectively. Results demonstrate that the fusion model based on the non-contrast CT radiomics of accessory respiratory muscles and clinical data exhibits promising diagnostic performance, highlighting its potential clinical significance for stratified management and preemptive critical care intervention in AECOPD patients.
Last Update: 2025-07-25