Segmentation of X‐ray coronary angiography with an artificial intelligence deep learning model: Impact in operator visual assessment of coronary stenosis severity
See where this sits in the topic map →Summary AI-generated
- TL;DR
- Visual assessment of coronary lesion severity during angiography is crucial, but human estimates often deviate from objective measurements.
- Problem
- Clinicians visually assessing the percentage diameter stenosis (%DS VE) in coronary angiography tend to overestimate lesion severity, particularly for lesions under 70%, which can lead to unwarranted revascularization.
- Method
- Using a previously developed artificial intelligence model for coronary angiography segmentation, operators were asked to estimate percentage diameter stenosis in both standard angiography and AI-segmented images across 123 lesions, comparing their estimates against quantitative coronary analysis (%DS QCA) as a reference.
- Results
- Estimates based on AI-segmented images showed a much smaller absolute difference from the reference measurements compared to standard angiography (59% vs 77% vs an actual 56%). Overall agreement across stenosis strata doubled with segmentation (60.4% vs 30.1%), and inter-operator differences were also smaller.
- Contributions
- Demonstrating that AI-based image segmentation significantly reduces visual overestimation and discrepancies in assessing coronary stenosis severity.
- Limitations
- Not specified in the abstract.
- Takeaways
- Integrating AI segmentation into coronary angiography interpretation can help reduce the overestimation of lesion severity and potentially prevent unwarranted revascularization procedures.
- Applications
- Clinical decision support in cardiology and catheterization laboratories.
- Topics
- Catheterization and Cardiovascular Interventions
- For industry
- Healthcare and Medical Devices
- Why it matters
- Improves the accuracy of coronary angiography interpretation and patient treatment decisions by mitigating operator overestimation of stenosis.
Abstract
Abstract Background Visual assessment of the percentage diameter stenosis (%DS VE ) of lesions is essential in coronary angiography (CAG) interpretation. We have previously developed an artificial intelligence (AI) model capable of accurate CAG segmentation. We aim to compare operators’ %DS VE in angiography versus AI‐segmented images. Methods Quantitative coronary analysis (QCA) %DS (%DS QCA ) was previously performed in our published validation dataset. Operators were asked to estimate %DS VE of lesions in angiography versus AI‐segmented images in separate sessions and differences were assessed using angiography %DS QCA as reference. Results A total of 123 lesions were included. %DS VE was significantly higher in both the angiography (77% ± 20% vs. 56% ± 13%, p < 0.001) and segmentation groups (59% ± 20% vs. 56% ± 13%, p < 0.001), with a much smaller absolute %DS difference in the latter. For lesions with %DS QCA of 50%–70% (60% ± 5%), an even higher discrepancy was found (angiography: 83% ± 13% vs. 60% ± 5%, p < 0.001; segmentation: 63% ± 15% vs. 60% ± 5%, p < 0.001). Similar, less pronounced, findings were observed for %DS QCA < 50% lesions, but not %DS QCA > 70% lesions. Agreement between %DS QCA /%DS VE across %DS QCA strata (<50%, 50%–70%, >70%) was approximately twice in the segmentation group (60.4% vs. 30.1%; p < 0.001). %DS VE inter‐operator differences were smaller with segmentation. Conclusion %DS VE was much less discrepant with segmentation versus angiography. Overestimation of %DS QCA < 70% lesions with angiography was especially common. Segmentation may reduce %DS VE overestimation and thus unwarranted revascularization.