Healthgrades for ProfessionalsRadiologists Outperform AI in Identifying Lung Diseases on Chest X-rays

Unlocking AI’s Potential in Radiology: Navigating Sensitivity and False Positives

Artificial intelligence (AI) holds promise in revolutionizing medical diagnostics, and a recent study in Radiology investigates its role in interpreting chest X-rays. Examining 72 radiologists and four commercial AI tools, the study reveals nuanced insights that challenge the narrative of AI seamlessly replacing human expertise.

Key Points

• AI tools demonstrated sensitivity in diagnosing airspace disease (72%-91%), pneumothorax (63%-90%), and pleural effusion (62%-95%) in older adult chest X-rays.
• However, the study highlighted a significant challenge – AI generated a notable number of false positives, particularly in complex cases or when dealing with smaller X-ray evidence.
• Positive predictive values for pneumothorax and pleural effusion ranged from 56% to 86%, while for airspace disease, AI’s accuracy dropped to 40%-50%.
• Radiologists outperformed AI, achieving a 96% accuracy rate in diagnosing pneumothorax.
• Lead author Dr. Louis Plesner emphasized the limitations, stating that AI excels in finding diseases but struggles to identify their absence, especially in complex cases.
• The high rate of false positives raises concerns about increased patient testing, time consumption, and unnecessary radiation exposure.

This study doesn’t surprise me and is exactly what would be expected of an AI system. At best, AI augments human skills in a complementary fashion. To view AI and human capability as mutually exclusive will always lead to disappointing results. We are not far along enough in the AI and deep learning space to entirely remove humans from the equation of productivity and patient outcomes. It’s just that simple.
— Zee Rizvi, the co-founder and president of Odesso Health, an AI-assisted service for automating electronic medical records

More in Radiology