The study aimed to evaluate the auxiliary diagnostic performance of an AI system in detecting superficial oesophageal squamous cell carcinoma and precancerous lesions using white light endoscopy (WLE) and non-magnified narrow-band imaging (NBI). Conducted across 12 hospitals in China, the trial involved patients undergoing sedated upper gastrointestinal endoscopy for various reasons. Participants were randomly assigned to either an AI-first or a routine-first group, with the endoscopist performing examinations with and without AI assistance.
Key findings include:
- Per-lesion miss rates were 1.7% in the AI-first group versus 6.7% in the routine-first group.
- Per-patient miss rates were 1.9% in the AI-first group versus 5.1% in the routine-first group.
- There were minimal adverse events, with bleeding after biopsy observed in a small fraction of patients in both groups.
- The study concluded that the AI-assisted endoscopy showed a substantial benefit to a neutral or small negative effect on miss rates of superficial oesophageal squamous cell carcinoma and precancerous lesions. However, the real-world clinical effectiveness and cost-benefit of this AI system require further assessment.
Insights:
- Technological Integration in Medical Diagnostics: The study highlights the growing role of artificial intelligence in enhancing diagnostic accuracy in medical procedures.
- Patient Safety and AI: The minimal adverse events reported suggest that AI integration in medical procedures can be safe and effective.
- Future of AI in Healthcare: The need for further real-world clinical assessments of AI systems in healthcare indicates a cautious but optimistic approach towards fully integrating AI in medical diagnostics.