Ready for testing artificial intelligence in radiology clinical practice: We would do well to be in the front line leveraging their strengths but also highlighting today weaknesses - Summary - MDSpire
Advertisement
Ready for testing artificial intelligence in radiology clinical practice: We would do well to be in the front line leveraging their strengths but also highlighting today weaknesses
To evaluate AI tools in radiology and their potential to assist radiologists in managing increasing workloads effectively.
Key Findings:
The ANN model outperformed the average radiologist on 144 findings, demonstrating significant potential.
Radiologist performance improved significantly with ANN assistance, particularly in detecting findings.
Some findings showed decreased performance with ANN assistance, highlighting areas for further investigation.
Interpretation:
AI tools can enhance radiological interpretation and confidence but cannot replace radiologists, especially for rare critical findings.
Limitations:
Some findings were excluded from AI output due to insufficient performance, indicating areas for improvement.
There is a risk of automation bias, where reliance on AI may lead to overlooking unreported critical findings.
Lack of reimbursement for AI applications in many regions could hinder their integration into clinical practice.
Conclusion:
AI tools have the potential to enhance radiological reporting quality and must be scientifically validated for clinical use, particularly as workloads increase.
At ARVO, Cecilia S. Lee, MD, MS, and Aaron Y. Lee, MD, MSCI, discuss barriers to AI deployment in ophthalmology, including interoperability and model development.
