Injury · 2025
Development and validation of interpretable machine learning models for predicting the risk of necrosis after finger replantation: A retrospective multicenter study.
СтатьяМногоцентровое исследованиеВалидационное исследованиеНазвание 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.
Dong Shuai, Hou Xinwei, Liu Shengzhe, Cao Yang, Wang Shi, Wang Guiyang, Liu Yucheng, Zhang Yan, Wu Chenglong, Ju Jihui, Wang Kai
Аннотация
INTRODUCTION: Digital necrosis (DN) is a critical postoperative complication following finger replantation surgery. This can necessitate additional surgical interventions that can adversely affect the patient's hand functionality, psychological well-being, and financial standing. The timely identification and management of the risk of post-replantation DN are thus crucial for enhancing patient outcomes. The objective of this study was to create and validate an easily understandable machine learning (ML) model for predicting the risk of DN following finger replantation surgery.
PATIENTS AND METHODS: Data from 1579 patients who underwent finger replantation surgery at Suzhou Ruihua Orthopaedic Hospital between September 2018 and September 2023 were collected and divided into training and internal validation sets. Additionally, 293 data points from two other institutions were employed as independent external validation sets. Ten machine-learning methods, including Gradient Boosting Machine (GBM), were utilized for modeling. The performance of the model was assessed using the area under the receiver operating characteristic curve (AUC) and decision curve analysis (DCA). SHapley Additive exPlanation (SHAP) was utilized to provide both global and local interpretations of the final model.
RESULTS: Nine indices, including the seniority of the doctor and the neutrophil count, were identified as independent predictors of DN. The GBM model showed optimal model with high predictive accuracy for DN risk in both the training set (AUC: 0.995) and the internal validation set (AUC: 0.978), which was confirmed using external validation (AUC: 0.983). The reliability and utility of the GBM model and the web-based computing platform were confirmed by DCA, calibration curve, accuracy, and sensitivity analyses.
CONCLUSION: An interpretable machine-learning model based on complete blood counts and related inflammatory marker levels was constructed and validated to predict the likelihood of developing DN following finger replantation. This model can assist clinicians in the prompt identification of high-risk patients post-replantation, enabling timely intervention.
PATIENTS AND METHODS: Data from 1579 patients who underwent finger replantation surgery at Suzhou Ruihua Orthopaedic Hospital between September 2018 and September 2023 were collected and divided into training and internal validation sets. Additionally, 293 data points from two other institutions were employed as independent external validation sets. Ten machine-learning methods, including Gradient Boosting Machine (GBM), were utilized for modeling. The performance of the model was assessed using the area under the receiver operating characteristic curve (AUC) and decision curve analysis (DCA). SHapley Additive exPlanation (SHAP) was utilized to provide both global and local interpretations of the final model.
RESULTS: Nine indices, including the seniority of the doctor and the neutrophil count, were identified as independent predictors of DN. The GBM model showed optimal model with high predictive accuracy for DN risk in both the training set (AUC: 0.995) and the internal validation set (AUC: 0.978), which was confirmed using external validation (AUC: 0.983). The reliability and utility of the GBM model and the web-based computing platform were confirmed by DCA, calibration curve, accuracy, and sensitivity analyses.
CONCLUSION: An interpretable machine-learning model based on complete blood counts and related inflammatory marker levels was constructed and validated to predict the likelihood of developing DN following finger replantation. This model can assist clinicians in the prompt identification of high-risk patients post-replantation, enabling timely intervention.