Clinical and translational gastroenterology · 2025
Machine Learning Model for Predicting Biliary Complications After Liver Transplantation.
СтатьяМногоцентровое 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.
Hu Feng, Li Yuancheng, Zeng Hongfei, Ju Renhua, Jiang Di, Zhang Leida, Li Jun, Liu Xingchao, Liu Guangyi, Zhang Chengcheng
Аннотация
INTRODUCTION: The risk factors of biliary complications (BCs) after liver transplantation are not comprehensively determined. BCs also vary in times of onset. Machine learning (ML) can reveal regularities based on large-scale data to make predictions and have demonstrated good performance in liver transplantation. However, whether ML can be an efficient tool for BC prediction was not determined.
METHODS: Five hundred seventeen patients from 2 centers were enrolled. Patients were randomly divided into training and validation sets at 3:1 ratio. K-fold cross-validation and the synthetic minority oversampling technique were used to debug the models, which were evaluated by receiver operating characteristic curves. SHapley Additive exPlanation values and Sankey diagrams were applied to visualize the results. Seven ML algorithms were administrated to build models for BCs prediction at 3, 6, and 12 months.
RESULTS: Among all the models, support vector machine produced the highest area under curve values in predicting BCs (3-month = 0.916; 6-month = 0.892; 12-month = 0.885). According to the analysis of support vector machine, the 3-month risk factors of BCs and corresponding SHapley Additive exPlanation value ranges were donor age (-0.13, 0.21), Model for End-Stage Liver Disease score (-0.15, 0.18), neoplastic disease (-0.14, 0.28), diabetes (-0.12, 0.27), hypertension (-0.13, 0.21), and intraoperative blood transfusion (-0.09, 0.25), whereas 6-month factors were recipient age (-0.14, 0.16), donor body mass index (-0.10, 0.13), recipient body mass index (-0.13, 0.23), and diabetes (-0.12, 0.43). The 12-month risk factors were recipient age (-0.14, 0.19), diabetes (-0.13, 0.25), and basiliximab (-0.16, 0.24). The Sankey diagram enabled clear visualization of the contribution of individual risk factors to the model in different times of BCs onset.
DISCUSSION: The ML algorithm was able to identify risk factors of BCs in all postoperative periods and this supplied insights in patient management.
METHODS: Five hundred seventeen patients from 2 centers were enrolled. Patients were randomly divided into training and validation sets at 3:1 ratio. K-fold cross-validation and the synthetic minority oversampling technique were used to debug the models, which were evaluated by receiver operating characteristic curves. SHapley Additive exPlanation values and Sankey diagrams were applied to visualize the results. Seven ML algorithms were administrated to build models for BCs prediction at 3, 6, and 12 months.
RESULTS: Among all the models, support vector machine produced the highest area under curve values in predicting BCs (3-month = 0.916; 6-month = 0.892; 12-month = 0.885). According to the analysis of support vector machine, the 3-month risk factors of BCs and corresponding SHapley Additive exPlanation value ranges were donor age (-0.13, 0.21), Model for End-Stage Liver Disease score (-0.15, 0.18), neoplastic disease (-0.14, 0.28), diabetes (-0.12, 0.27), hypertension (-0.13, 0.21), and intraoperative blood transfusion (-0.09, 0.25), whereas 6-month factors were recipient age (-0.14, 0.16), donor body mass index (-0.10, 0.13), recipient body mass index (-0.13, 0.23), and diabetes (-0.12, 0.43). The 12-month risk factors were recipient age (-0.14, 0.19), diabetes (-0.13, 0.25), and basiliximab (-0.16, 0.24). The Sankey diagram enabled clear visualization of the contribution of individual risk factors to the model in different times of BCs onset.
DISCUSSION: The ML algorithm was able to identify risk factors of BCs in all postoperative periods and this supplied insights in patient management.