[Artificial intelligence in medicine: are we ready?]

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Machine Learning Applications

Context:

• Artificial intelligence (AI) and machine learning (ML) are increasingly applied across medical fields.
• Applications include diagnostics, treatment personalization, drug discovery, and risk prediction.
• Existing studies show encouraging results but often have methodological limitations.

Purpose:

• To evaluate the current state of AI and ML in medicine.
• To highlight the need for rigorous scientific validation of AI tools.
• To emphasize the importance of prospective, real-world clinical trials.

Summary:

• Many AI/ML studies rely on retrospective data and internal validation.
• Few studies are prospective, conducted in real clinical settings, or use randomized controlled trials.
• Direct comparison with expert performance using identical datasets is rare.

Impact:

• AI/ML systems require robust validation demonstrating non-inferiority or superiority to conventional methods.
• Safety, reproducibility, cost-effectiveness, and ethical considerations are crucial for clinical integration.
• Future research must focus on prospective, randomized trials in real-world settings.