Can incorrect artificial intelligence (AI) results impact radiologists, and if so, what can we do about it? A multi-reader pilot study of lung cancer detection with chest radiography - Summary - MDSpire
Advertisement
Can incorrect artificial intelligence (AI) results impact radiologists, and if so, what can we do about it? A multi-reader pilot study of lung cancer detection with chest radiography
To examine the impact of incorrect AI results on radiologist performance and explore how specific human factors, such as feedback presentation, can be optimized to mitigate this impact.
Key Findings:
AI systems have a sensitivity of 0.87 and specificity of 0.89, indicating potential for false negatives and positives, which can significantly impact diagnostic accuracy.
Incorrect AI results can lead to significant misinterpretation by radiologists, as shown in a study where CAD feedback influenced diagnostic accuracy.
Human factors, such as whether AI results are kept or deleted and whether areas of suspicion are visually outlined, can affect radiologist performance, highlighting the need for careful implementation.
Interpretation:
The study suggests that incorrect AI feedback can negatively influence radiologist decision-making, emphasizing the need for careful implementation and presentation of AI results to enhance clinical outcomes.
Limitations:
The study did not use real AI systems, which may limit the generalizability of the findings to clinical practice.
There is a lack of empirical studies on human factors in AI implementation in radiology, which could provide further insights.
Conclusion:
Optimizing how AI feedback is presented may help mitigate the adverse effects of incorrect AI results on radiologist performance, emphasizing the importance of both algorithm accuracy and effective implementation strategies.
by Michael H. Bernstein, Michael K. Atalay, Elizabeth H. Dibble, Aaron W. P. Maxwell, Adib R. Karam, Saurabh Agarwal, Robert C. Ward, Terrance T. Healey, Grayson L. Baird
