Exploring the value of pleural fluid biomarkers for complementary pleural effusion disease examination

Area of Science:

• Biochemistry
• Medical Diagnostics
• Artificial Intelligence

Background:

• Pleural fluid biomarkers aid in diagnosing pleural effusion.
• Deep learning offers a novel approach to evaluate these biomarkers.
• Distinguishing between malignant pleural effusion (MPE), parapneumonic effusion (PPE), and congestive heart failure (CHF) is clinically important.

Purpose of the Study:

• To evaluate the multidimensional diagnostic utility of deep learning for pleural effusion biomarkers.
• To assess the value of combining biomarkers for distinguishing effusion etiologies.
• To compare deep learning analysis with traditional ROC curve analysis.

Main Methods:

• Pleural effusion samples were categorized into MPE, PPE, and CHF groups.
• Biomarker values were initially analyzed using receiver operating characteristic (ROC) curves.
• Deep learning and entropy weight method (EWM) were employed for multidimensional analysis.

Main Results:

• Significant differences (p < 0.05) were observed in TP, ADA, CEA, CYFRA211, NSE, and MNC% across groups.
• Deep learning-based comprehensive testing of pleural fluid biomarkers demonstrated high accuracy.
• Cytomorphology biomarkers (WBC, MNC%, PNC%, MTC%) exhibited higher clinical value.

Conclusions:

• Multidimensional analysis using deep learning and EWM provides different insights compared to ROC analysis.
• Cell morphology biomarkers warrant greater clinical attention for diagnosing pleural effusions.
• Physical properties of pleural fluid appear to have lesser clinical significance in this context.