Medical AI shows workflow benefits despite diagnostic accuracy limits

Recent developments in medical artificial intelligence highlight both the workflow benefits and the diagnostic limitations of AI in radiology.

A study by researchers at the Technical University of Munich evaluated four commercial AI systems used to interpret chest X-rays. The research found that while AI assistance helped radiologists interpret images faster and increased their diagnostic confidence, it did not consistently improve overall diagnostic accuracy. In some instances, the use of AI actually reduced accuracy for specific findings, such as pulmonary nodules and pleural effusions, due to an increase in false-positive results.

In a separate advancement, the Berlin-based company Vara has received Class IIb CE certification for its autonomous triage system. This allows the AI to independently report certain mammograms as clearly normal without requiring a radiologist’s review. This technology aims to alleviate the workload in breast-cancer screening programs by allowing the AI to handle low-risk, normal cases while directing more complex examinations to human specialists.