Abstract
Previous studies on primary blast lung injury have mostly been small-sample simulation experiments, primarily relying on morphological identification and lacking imaging-based classification of severity. Herein, we have established a large-sample model of goats exposed to real natural field explosions and employed CT radiomics to assess the severity of lung injury. By extracting 1288 radiomics features and combining baseline data, baseline, radiomics, and comprehensive models were built. Results showed that the radiomics and comprehensive models outperformed the baseline model. Decision curve analysis indicated better clinical benefits with models incorporating rad-scores. A nomogram established with multiple factors demonstrated individualized predictive performance. The addition of CT radiomics features improved assessment accuracy and is expected to support clinical decision-making.