Frontiers in medicine · 2026
Development and validation of an automatic machine learning-based myocardial contusion prediction model in rib fracture patients: a retrospective multicenter cohort study.
СтатьяНазвание и аннотация: Перевод в очереди
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Yang Yan, Liu Ping, Jia Ni, Xu Yisha, Wu Peng, Zhang Zhigang, Han Tingyong, Jiang Yan, Huang Xianpeng, Li Long
Аннотация
OBJECTIVE: This study aims to construct an automatic machine learning-based prediction model to accurately identify the risk of myocardial contusion in traumatic rib fracture patients, addressing the core limitations of insufficient sensitivity in traditional diagnostic methods and difficulties in clinical translation of existing prediction models.
METHODS: A multicenter retrospective study design was adopted, including 1,038 rib fracture patients from 8 hospitals (training set: 808 cases; testing set: 230 cases) from June 2024 to June 2025. An automatic machine learning framework (AutoML) was developed based on the Improved Crested Porcupine Optimization algorithm (ICPO). A dual-stage optimization process achieved key feature screening and hyperparameter tuning, with Synthetic Minority Over-sampling Technique (SMOTE) addressing class imbalance. Model performance was systematically evaluated using six core metrics (Sensitivity/SEN, Specificity/SPE, F1-score, AUC, etc.), calibration curves, and Decision Curve Analysis (DCA). Feature robustness was validated via LASSO regression, and SHAP interpretability models analyzed key variable contributions. A clinical decision support interactive tool was developed on the MATLAB platform.
RESULTS: The AutoML model demonstrated superior performance on the testing set (classification threshold = 0.5; TP = 138, TN = 64, FP = 19, FN = 9; PRE = 0.8790, SEN = 0.9388, SPE = 0.7711, ACC = 0.8783, F1 = 0.9079, AUC = 0.9537), significantly outperforming traditional models such as logistic regression and support vector machine. SHAP analysis identified sternal fracture, anterolateral upper thoracic fracture, pneumomediastinum, and ISS score as key contributing features in the predictive model. SHAP analysis further suggested potential interaction patterns between pneumomediastinum and ISS > 20, and between sternal fracture and pneumothorax within the model.
CONCLUSION: The AutoML prediction model exhibits high accuracy and strong interpretability. Its integrated visual interactive tool represents an initial step toward addressing clinical translation barriers, providing intelligent decision support for early warning and stratified intervention of myocardial contusion in rib fracture patients.
METHODS: A multicenter retrospective study design was adopted, including 1,038 rib fracture patients from 8 hospitals (training set: 808 cases; testing set: 230 cases) from June 2024 to June 2025. An automatic machine learning framework (AutoML) was developed based on the Improved Crested Porcupine Optimization algorithm (ICPO). A dual-stage optimization process achieved key feature screening and hyperparameter tuning, with Synthetic Minority Over-sampling Technique (SMOTE) addressing class imbalance. Model performance was systematically evaluated using six core metrics (Sensitivity/SEN, Specificity/SPE, F1-score, AUC, etc.), calibration curves, and Decision Curve Analysis (DCA). Feature robustness was validated via LASSO regression, and SHAP interpretability models analyzed key variable contributions. A clinical decision support interactive tool was developed on the MATLAB platform.
RESULTS: The AutoML model demonstrated superior performance on the testing set (classification threshold = 0.5; TP = 138, TN = 64, FP = 19, FN = 9; PRE = 0.8790, SEN = 0.9388, SPE = 0.7711, ACC = 0.8783, F1 = 0.9079, AUC = 0.9537), significantly outperforming traditional models such as logistic regression and support vector machine. SHAP analysis identified sternal fracture, anterolateral upper thoracic fracture, pneumomediastinum, and ISS score as key contributing features in the predictive model. SHAP analysis further suggested potential interaction patterns between pneumomediastinum and ISS > 20, and between sternal fracture and pneumothorax within the model.
CONCLUSION: The AutoML prediction model exhibits high accuracy and strong interpretability. Its integrated visual interactive tool represents an initial step toward addressing clinical translation barriers, providing intelligent decision support for early warning and stratified intervention of myocardial contusion in rib fracture patients.