Integrating CT radiomics and clinical data with machine learning to predict fibrosis progression in coalworker pneumoconiosis.

OBJECTIVE

This study aims to develop a machine learning (ML) model that integrates computed tomography (CT) radiomics with clinical features to predict the progression of pulmonary interstitial fibrosis in patients with coalworker pneumoconiosis (CWP).

METHODS

Clinical and imaging data from 297 patients diagnosed with CWP at The First Affiliated Hospital of Chongqing Medical and Pharmaceutical College between December 2021 and December 2023 were analyzed. Of these patients, 170 developed pulmonary interstitial fibrosis over a 3-year follow-up and were classified as the progression group, while 127 patients showed stable conditions and were classified as the stable group. The patients were divided into a training cohort ( = 207) and a test cohort ( = 90). Radiomic features were extracted from CT images of lung fibrosis lesions in the training cohort. These features were reduced in dimensionality to construct morphological biomarkers. ML methods were then used to develop three models: a clinical model, a radiomics model, and a multimodal joint model. The performance of these models was evaluated in the test cohort using receiver operating characteristic (ROC) curves and decision curve analysis (DCA).

RESULTS

In the training cohort, the area under the curve (AUC) for the clinical, radiomics, and joint models were 0.835, 0.879, and 0.945, respectively. In the test cohort, the AUC values for these models were 0.732, 0.750, and 0.845, respectively. The joint model demonstrated the highest predictive performance and clinical benefit in both the training and test cohorts.

CONCLUSION

The multimodal model, combining CT radiomics and clinical features, offers an effective and accurate tool for predicting the progression of pulmonary fibrosis in CWP.