Journal of thoracic disease · 2026
Development and validation of a diagnostic model for malignant pleural effusion based on random forest.
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.
Guo Xu, Wei Xiao-Lei, Yin Shu-Min, Li Yu-Peng, Zhai Kan, Shi Huan-Zhong, Wang Zheng
Аннотация
BACKGROUND: The timely and accurate identification of malignant pleural effusion (MPE) is crucial for prompt treatment. The current diagnostic approaches based on cytology, pathology, or biomarkers have inherent limitations; notably, the drawbacks of single indicators can be overcome by combining various indicators. Accordingly, the present study was designed to construct a machine learning (ML) model integrating pleural effusion (PE) and peripheral blood indicators to identify MPE.
METHODS: This retrospective study enrolled 1,239 adult participants with PE who underwent diagnostic thoracentesis and then eligible participants were randomly assigned to training (70%) and test (30%) sets. Demographic data, laboratory variables, and systemic inflammatory indicators were analyzed. After feature selection by Boruta and the least absolute shrinkage and selection operator, we constructed five ML models, including multivariable logistic regression, support vector machine, random forest (RF), extreme gradient boosting, and k-nearest neighbors. The predictive performance of each model was comprehensively evaluated using multiple metrics, calibration curve, and decision curve analysis. In addition, Shapley additive explanations and Gini importance were applied to interpret each model.
RESULTS: Among the employed ML models, the RF model achieved optimal performance. Following further importance analysis, the eight-feature RF model achieved an area under the receiver operating characteristic curve (AUC) of 0.940, with sensitivity, specificity, positive predictive value, negative predictive value, accuracy, and F1 score of 0.875, 0.835, 0.882, 0.826, 0.858, and 0.878, respectively. These key features included PE-adenosine deaminase, carcinoembryonic antigen, PE-cell count, PE-total protein, PE-percentage of mononuclear cells, age, PE-lactate dehydrogenase, and platelet-to-lymphocyte ratio.
CONCLUSIONS: We developed and validated an interpretable and accurate eight-feature RF model for MPE identification with an AUC of 0.940. This model provides a valuable approach for MPE differential diagnosis.
METHODS: This retrospective study enrolled 1,239 adult participants with PE who underwent diagnostic thoracentesis and then eligible participants were randomly assigned to training (70%) and test (30%) sets. Demographic data, laboratory variables, and systemic inflammatory indicators were analyzed. After feature selection by Boruta and the least absolute shrinkage and selection operator, we constructed five ML models, including multivariable logistic regression, support vector machine, random forest (RF), extreme gradient boosting, and k-nearest neighbors. The predictive performance of each model was comprehensively evaluated using multiple metrics, calibration curve, and decision curve analysis. In addition, Shapley additive explanations and Gini importance were applied to interpret each model.
RESULTS: Among the employed ML models, the RF model achieved optimal performance. Following further importance analysis, the eight-feature RF model achieved an area under the receiver operating characteristic curve (AUC) of 0.940, with sensitivity, specificity, positive predictive value, negative predictive value, accuracy, and F1 score of 0.875, 0.835, 0.882, 0.826, 0.858, and 0.878, respectively. These key features included PE-adenosine deaminase, carcinoembryonic antigen, PE-cell count, PE-total protein, PE-percentage of mononuclear cells, age, PE-lactate dehydrogenase, and platelet-to-lymphocyte ratio.
CONCLUSIONS: We developed and validated an interpretable and accurate eight-feature RF model for MPE identification with an AUC of 0.940. This model provides a valuable approach for MPE differential diagnosis.