Radiomics-driven prediction of postoperative infection risk in maxillofacial fractures: multidimensional risk factor analysis and model construction based on absorbable plate fixation.

OBJECTIVES: This study proposes a predictive model that integrates radiological features with multidimensional clinical factors for the accurate prediction of postoperative infection in maxillofacial fracture patients treated with absorbable fixation plates. METHODS: This model incorporates a fusion algorithm based on K-means clustering and fully connected neural networks to perform in-depth analysis of non-contrast and reconstructed maxillofacial CT images. It also includes frequently overlooked clinical variables, such as surgery-related factors and individual patient characteristics, thereby enhancing the model's interpretability and generalizability. The study involved 1200 maxillofacial CT images from 100 patients at the First Affiliated Hospital of Dalian Medical University. The dataset was randomly split into training and testing sets in an 8:2 ratio for validation. RESULTS: The combined model achieved an area under the AUC-ROC curve of 0.96. The Results demonstrated that the model exhibited consistent and excellent performance in testing, reliably identifying high-risk factors for postoperative infection. The establishment of this model holds positive significance for promoting the application of absorbable materials and advancing personalized treatment for trauma patients. CONCLUSION: The integrated habitat-clinical model demonstrates improved predictive performance. Combining habitat analysis with clinical features offers a promising approach for the prediction of postoperative infection of absorbable bone plates. CLINICAL RELEVANCE: The establishment of this model holds positive significance for promoting the application of absorbable materials and advancing personalized treatment for trauma patients.