Machine learning-based Diagnostic model for determining the etiology of pleural effusion using Age, ADA and LDH DOI Creative Commons
Qingyu Chen,

Shu-Min Yin,

Ming‐Ming Shao

et al.

Respiratory Research, Journal Year: 2025, Volume and Issue: 26(1)

Published: May 2, 2025

Classification of the etiologies pleural effusion is a critical challenge in clinical practice. Traditional diagnostic methods rely on simple cut-off method based laboratory tests. However, machine learning (ML) offers novel approach artificial intelligence to improving accuracy and capture non-linear relationships. A retrospective study was conducted using data from patients diagnosed with effusion. The dataset divided into training test set ratio 7:3 6 algorithms implemented diagnosis Model performances were assessed by accuracy, precision, recall, F1 scores area under receiver operating characteristic curve (AUC). Feature importance average prediction age, Adenosine (ADA) Lactate dehydrogenase (LDH) analyzed. Decision tree visualized. total 742 included (training cohort: 522, 220), 397 (53.3%) malignant (MPE) 253 (34.1%) tuberculous (TPE) cohort. All models performed well MPE, TPE transudates. Extreme Gradient Boosting Random Forest better above 0.890, while K-Nearest Neighbors Tabular Transformer TPE, 0.870. ADA identified as most important feature. ROC model outperformed those conventional thresholds. This demonstrates that ML ADA, LDH can effectively classify effusion, suggesting ML-based approaches may enhance decision-making.

Language: Английский

Novel secondary ion mass spectrometry identification system for organic materials using random forest DOI Open Access

Tetsuya Masuda,

Miya Fujita,

Tomikazu Ueno

et al.

Journal of Vacuum Science & Technology A Vacuum Surfaces and Films, Journal Year: 2025, Volume and Issue: 43(2)

Published: Feb. 14, 2025

The interpretation of time-of-flight secondary ion mass spectrometry (ToF-SIMS) data is often complicated because ToF-SIMS has a high sensitivity for detecting extremely low amounts molecules and generally produces numerous types fragment ions from each molecule. Although machine learning techniques have been applied to such complex classify the components in sample, identifying unknown difficult, even after classification or segmentation datasets. We developed new (SIMS) identification system based on full spectra by applying supervised method, random forest (RF), with effective teaching information express common organic molecules. automatically extracted chemical structures material string-converted using simplified molecular-input line-entry system. 32 molecules, including peptides, polymers, biomolecules as cellulose, were used training dataset, these correctly predicted SIMS importance RF indicated that peaks representing detected materials identified essential target Moreover, Styrofoam-like Ocean plastic samples polystyrene This study demonstrates potential our accurately identify spectra, offering robust approach expanding molecular samples.

Language: Английский

Citations

0

Machine learning-based Diagnostic model for determining the etiology of pleural effusion using Age, ADA and LDH DOI Creative Commons
Qingyu Chen,

Shu-Min Yin,

Ming‐Ming Shao

et al.

Respiratory Research, Journal Year: 2025, Volume and Issue: 26(1)

Published: May 2, 2025

Classification of the etiologies pleural effusion is a critical challenge in clinical practice. Traditional diagnostic methods rely on simple cut-off method based laboratory tests. However, machine learning (ML) offers novel approach artificial intelligence to improving accuracy and capture non-linear relationships. A retrospective study was conducted using data from patients diagnosed with effusion. The dataset divided into training test set ratio 7:3 6 algorithms implemented diagnosis Model performances were assessed by accuracy, precision, recall, F1 scores area under receiver operating characteristic curve (AUC). Feature importance average prediction age, Adenosine (ADA) Lactate dehydrogenase (LDH) analyzed. Decision tree visualized. total 742 included (training cohort: 522, 220), 397 (53.3%) malignant (MPE) 253 (34.1%) tuberculous (TPE) cohort. All models performed well MPE, TPE transudates. Extreme Gradient Boosting Random Forest better above 0.890, while K-Nearest Neighbors Tabular Transformer TPE, 0.870. ADA identified as most important feature. ROC model outperformed those conventional thresholds. This demonstrates that ML ADA, LDH can effectively classify effusion, suggesting ML-based approaches may enhance decision-making.

Language: Английский

Citations

0