Biochemistry-based machine learning algorithms in differentiating pleural effusion: current status and perspective

The differential diagnosis of pleural effusion remains challenging. Microbiological and cytopathological examinations are considered the gold standards; however, they are limited by their low sensitivity, subjectivity, invasiveness, and prolonged turnaround times. Pleural fluid and serum biochemical tests offer the advantages of objectivity, short turnaround time, minimal invasiveness, and easy accessibility, which can help pulmonologists estimate the risk of the target disease. However, their effectiveness is often suboptimal when they are used alone. Recent advances suggest that machine learning (ML) algorithms can enhance diagnostic accuracy when combined with multiple parameters. Several studies have applied ML approaches based on biochemical tests to diagnose pleural effusion, with preliminary results indicating an improved diagnostic performance. This article reviews the application of such algorithms in the differential diagnosis of pleural effusion, highlights their current limitations, and provides recommendations for future research.