Generative AI-based low-dose digital subtraction angiography for intra-operative radiation dose reduction: a randomized controlled trial.

Summary

In a multicenter randomized controlled trial, a generative AI-enabled low-dose DSA protocol reduced intra-operative radiation substantially versus standard protocols, with mean air kerma lowered by 306 mGy. Blinding of patients, surgeons, and investigators strengthens the validity, while secondary outcomes assessed efficiency and complications.

Key Findings

• Randomized, multicenter validation in 1,068 patients across 70 centers comparing GenDSA-V2 vs standard protocols.
• Air kerma reduced to 151.3 ± 125.1 mGy vs 457.4 ± 407.4 mGy (mean difference −306.1 mGy; 95% CI −342.3 to −269.9; P < 0.001).
• Dose-area product in the AI group averaged 4009.7 ± 2767.9 μGy·m², with secondary endpoints assessing efficiency, procedure time, and complications.

Clinical Implications

Interventional teams can adopt AI-enabled low-dose DSA workflows to reduce radiation to patients and staff without compromising procedural efficiency. This supports broader radiation stewardship in catheterization labs and may inform dose benchmarks and device procurement.

Limitations

• Technicians were not blinded, which may introduce procedural behavior bias
• The study encompassed diverse indications (neuro, thoracic, hepatic) rather than exclusively cardiology, potentially affecting generalizability to cath labs
• Secondary outcomes (efficiency, time, complications) details not fully reported in the abstract

Future Directions

Head-to-head comparisons across different vendors and cath lab settings, assessment of long-term operator exposure, and evaluation of clinical outcomes and cost-effectiveness will support broader adoption.