Development and Validation of a Machine Learning Diagnostic Model Using Routine Clinical Data to Differentiate Malignant from Benign Pericardial Effusion

Objectives: This study aimed to develop and validate a machine learning model based on routine clinical data to provide real-time, accurate decision support for distinguishing malignant pericardial effusion (MPE) from benign pericardial effusion (BPE). Methods: In total, 296 patients were retrospectively enrolled, comprising 168 with BPE and 128 with MPE. The dataset was randomly divided into training and test sets in a 7:3 ratio. Six machine learning algorithms were trained to construct predictive models, with their predictive accuracy assessed using the Area Under the Curve (AUC) and Decision Curve Analysis (DCA). Furthermore, SHapley Additive exPlanations (SHAP) analysis was utilized to interpret model outputs and identify key predictive variables. Results: All six machine learning models achieved favorable diagnostic performance. Among them, the LightGBM (LGB) model demonstrated the highest predictive accuracy; the LGB test set under the receiver operating characteristic curve (AUC) was 0.930 (95% CI 0.871–0.979), with a sensitivity of 86.8%, a specificity of 94.0%, and an accuracy of 90.9%. XGBoost (XGB) had a lower score than LGB (0.111 versus 0.142). LGB calibration intercept and slope were 0.246 and 3.328. The combined serum D-dimer/carbohydrate antigen 125 (CA125) model had AUC 0.876; its difference from LGB was uncertain (ΔAUC 0.054, 95% CI −0.011 to 0.125; Holm-adjusted p = 0.382). Mean LGB out-of-fold AUC was 0.888 (SD 0.008). Fluid lactate dehydrogenase had the highest SHAP importance. Conclusions: This study developed and validated an interpretable machine learning model incorporating five readily available clinical indicators that enables accurate and rapid differentiation between MPE and BPE, enhancing diagnostic decision-making for pericardial effusion.

Authors

Institutions

Publication Details

Journal
Journal of Clinical Medicine
Published
2026-11-17
DOI
https://doi.org/10.3390/jcm15229012
Primary Topic
Pericarditis and Cardiac Tamponade
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Development and Validation of a Machine Learning Diagnostic Model Using Routine Clinical Data to Differentiate Malignant from Benign Pericardial Effusion

Lin Pei, Shuang Sun, Jing Song, Ziqi Song et al.
2 links
Journal of Clinical Medicine
Pericarditis and Cardiac Tamponade
article

Development and Validation of a Machine Learning Diagnostic Model Using Routine Clinical Data to Differentiate Malignant from Benign Pericardial Effusion

Lin Pei, Shuang Sun, Jing Song, Ziqi Song, Chunying Zhang, Junxian Song, Mei Rong Zhao, Jian Liu
article en

Abstract

Objectives: This study aimed to develop and validate a machine learning model based on routine clinical data to provide real-time, accurate decision support for distinguishing malignant pericardial effusion (MPE) from benign pericardial effusion (BPE). Methods: In total, 296 patients were retrospectively enrolled, comprising 168 with BPE and 128 with MPE. The dataset was randomly divided into training and test sets in a 7:3 ratio. Six machine learning algorithms were trained to construct predictive models, with their predictive accuracy assessed using the Area Under the Curve (AUC) and Decision Curve Analysis (DCA). Furthermore, SHapley Additive exPlanations (SHAP) analysis was utilized to interpret model outputs and identify key predictive variables. Results: All six machine learning models achieved favorable diagnostic performance. Among them, the LightGBM (LGB) model demonstrated the highest predictive accuracy; the LGB test set under the receiver operating characteristic curve (AUC) was 0.930 (95% CI 0.871–0.979), with a sensitivity of 86.8%, a specificity of 94.0%, and an accuracy of 90.9%. XGBoost (XGB) had a lower score than LGB (0.111 versus 0.142). LGB calibration intercept and slope were 0.246 and 3.328. The combined serum D-dimer/carbohydrate antigen 125 (CA125) model had AUC 0.876; its difference from LGB was uncertain (ΔAUC 0.054, 95% CI −0.011 to 0.125; Holm-adjusted p = 0.382). Mean LGB out-of-fold AUC was 0.888 (SD 0.008). Fluid lactate dehydrogenase had the highest SHAP importance. Conclusions: This study developed and validated an interpretable machine learning model incorporating five readily available clinical indicators that enables accurate and rapid differentiation between MPE and BPE, enhancing diagnostic decision-making for pericardial effusion.

Journal of Clinical MedicineVol. 15(22)
Peking University People's Hospital (CN)
Good health and well-being
Pericarditis and Cardiac Tamponade
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon