Feasibility of deep learning reconstruction algorithm combined with the "ultra triple-low" protocol at 60 kVp in CTPA: An image quality and radiation dose assessment.
Zhichao Liu, Qin Wang, Xianying Ning et al.
✦ AI-curated · Sources linked
In 30 seconds
This randomised controlled trial assessed whether deep learning reconstruction could support pulmonary CT angiography using 60 kVp, 10 ml of iodine contrast, and a 2.5 ml/s injection rate. Seventy patients with BMI < 28 kg/m² were assigned to the ultra triple-low protocol or a standard protocol; the low-dose group was reconstructed using three methods. With deep learning reconstruction, image-quality measures and subjective scores were superior to the other low-dose reconstructions, and all pulmonary artery CT values exceeded 250 HU. The ultra triple-low protocol reduced effective dose by 84.93 % (0.71 ± 0.09 vs. 4.71 ± 1.13 mSv) and contrast volume by 66.7 % (10 ml vs. 30 ml).
Key findings
- The ultra triple-low group had an effective dose of 0.71 ± 0.09 vs. 4.71 ± 1.13 mSv, a reported reduction of 84.93 % (P < 0.001).
- Contrast volume was 10 ml vs. 30 ml, a reported reduction of 66.7 %.
- With deep learning reconstruction, all pulmonary artery CT values exceeded 250 HU, and SNR, CNR, subjective scores, and artifact scores were superior (P < 0.001).
- Inter-observer agreement was strong, with κ = 0.87-0.92.
Why it matters
CTPA requires adequate pulmonary artery enhancement while limiting radiation and contrast exposure. These results suggest that deep learning reconstruction may help maintain diagnostic image quality with substantially lower doses in selected patients. The findings are relevant to protocol development, but the study’s population and conclusion do not establish performance in all patients or clinical settings.
What to keep in mind
The trial included only patients with BMI < 28 kg/m², and the authors state that further validation is warranted.