Conexiant’s news site is now MDSpire News. Learn more
Advertisement
AI cuts lesion measurement time by 34%
Automated lesion measurements increased agreement on treatment response, although expert-generated proposals were accepted more often and required fewer adjustments.
Clinical Report: AI Cuts Lesion Measurement Time by 34%
Overview
Artificial intelligence assistance in lesion measurements significantly reduced reading time by 34% but resulted in greater variability in individual lesion measurements compared to unassisted assessments.
Background
The accurate measurement of lesions in cancer patients is critical for evaluating treatment response and disease progression. Traditional methods can be time-consuming and prone to variability, which may affect clinical decision-making. The integration of artificial intelligence in imaging workflows aims to enhance efficiency and consistency in lesion assessment.
Data Highlights
Condition
Mean Reading Time (seconds)
Interobserver Agreement Change (%)
Mean Absolute Error (mm)
Unassisted
105
-
3.08
AI-Assisted
71
8
4.35
Expert-Assisted
56
13
1.51
Key Findings
AI assistance reduced mean reading time by 34 seconds compared to unassisted assessment.
Interreader agreement on RECIST response classification improved by approximately 8 percentage points with AI assistance.
AI-assisted measurements exhibited greater deviation from expert-derived references than unassisted measurements.
Radiologists accepted 61% of AI-generated proposals, compared to 77% for expert proposals.
Substantial modifications occurred in 23% of AI-assisted measurements among radiologists.
Clinical Implications
Clinicians should be aware of the increased variability in measurements when utilizing AI assistance.
Conclusion
AI assistance in RECIST assessment may enhance efficiency, although expert assistance remains superior in terms of agreement and accuracy.
