A computed tomography-based lung radiomics nomogram to identify acute exacerbation of chronic obstructive pulmonary...

Diagnosing acute exacerbation of chronic obstructive pulmonary disease (AECOPD) remains challenging in clinical practice. Computed tomography (CT) plays a great role in characterizing and quantifying changes in lung structure and function of AECOPD. This study aimed to explore the performance of CT-based whole-lung radiomics in identifying AECOPD.

From 2013 to 2022, 475 chronic obstructive pulmonary disease (COPD) patients were retrospectively enrolled from three hospitals, including 257 non-AECOPD and 218 AECOPD. Whole-lung imaging features were extracted. Eleven machine learning (ML) algorithms were utilized to construct the AECOPD identification model. Least absolute shrinkage and selection operator logistic regression (LR) was applied for feature selection and radiomic signature construction. A radiomic nomogram was then established by combining the radiomic score and clinical factors. Receiver operating characteristic curve analysis and decision curve analysis assessed the predictive performance of the radiomic nomogram in the training, internal, and independent external validation cohorts.

Sixty-one radiomic features were collected to construct a radiomic score model. Among the 11 ML algorithms, the LR model achieved the best diagnostic performance, with the area under the curve (AUC) values of 0.974, 0.836, and 0.944 in the training, internal, and external validation sets, respectively. The diagnostic performance surpassed that of the clinical model, with AUC values of 0.630, 0.636, and 0.734 in the respective sets. Notably, the AECOPD radiomic nomogram demonstrated nearly equivalent diagnostic performance to the radiomic score model. The AUC values for the radiomic nomogram were 0.974 in the training set, 0.849 in the internal validation set, and 0.957 in the external validation set.

The CT-based whole-lung radiomic nomogram accurately identifies AECOPD and offers a robust tool for clinical diagnosis and treatment planning.