Related Experiment Video
Updated: Jan 12, 2026
Author Spotlight: Advancing CBCT and Digital Dental Image Integration with AI-Assisted Digitization
Published on: February 23, 2024
Quantitative evaluation of an artificial intelligence-assisted platform in CT acquisition workflow
Marco Caballo1, Laura McLennan2, Matthew Benbow3
1Computed Tomography, Canon Medical Systems Europe, Bovenkerkerweg 59, 1185 XB Amstelveen, the Netherlands.
Background:
Artificial intelligence (AI) may help reduce the examination time and simplify the scanning process during CT image acquisition. This cross-sectional study aims to quantitatively and objectively assess the potential benefits of an AI-assisted CT acquisition platform in streamlining CT acquisition workflow, by reducing time and user interactions, compared to a non-AI-assisted platform.
Methods:
Twelve certified diagnostic radiographers scanned an anthropomorphic body phantom for four different protocols on two similar CT systems, one equipped with an AI-assisted scanning platform and one without. Scanning sessions were video-recorded, and two primary variables (total examination time and number of user interactions with the platforms) were extracted. Differences in variable outcomes between the two platforms were analyzed statistically with the Mann-Whitney U test (with Bonferroni correction). The influence of radiographers' experience on each variable outcome was quantified with Spearman correlation, and inter-reader agreement among the radiographers with the intra-class correlation coefficient (ICC).
Results:
Acquisition time and number of interactions were both significantly lower on the AI-assisted platform (P < 0.001). The average (±standard deviation) reduction in acquisition time was between 40.2 % (±9.8 %) and 52.8 % (±10.9 %), depending on the protocol. The average (±standard deviation) reduction in interactions was between 35.6 % (±9.7 %) and 45.1 % (±14.0 %), depending on the protocol. No significant correlation was found between radiographer experience and acquisition time or number of interactions for either platform (P ≥ 0.3). Inter-reader agreement was substantial on both platforms (acquisition time: ICC = 0.82 and ICC = 0.85 respectively for the AI-assisted and non-AI-assisted platform; interactions: ICC = 0.76 and ICC = 0.81 respectively for the AI-assisted and non-AI-assisted platform).
Discussion:
AI-assisted CT acquisition platforms may improve CT acquisition workflow, pending further confirmation in future multicenter studies and with large datasets of patient data.
Conclusion:
A reduction in time and interactions required to perform CT image acquisition may have real-world pragmatic implications in reducing radiographer workload and improving departmental throughput.
