Performance evaluation of a deep learning image reconstruction (DLIR) algorithm in “double low” chest CTA in children: a feasibility study

Chest CT angiography (CTA) is a convenient clinical examination for children with an increasing need to reduce both radiation and contrast medium doses. Iterative Reconstruction algorithms are often used to reduce image noise but encounter limitations under low radiation dose and conventional 100 kVp tube voltage may not provide adequate enhancement under low contrast dose. To evaluate the performance of a deep learning image reconstruction (DLIR) algorithm in conjunction with lower tube voltage in chest CTA in children under reduced radiation and contrast medium (CM) dose. 46 Children (age 5.9 ± 4.2 years) in the study group underwent chest CTA with 70 kVp and CM dose of 0.8–1.2 ml/kg. Images were reconstructed at 0.625 mm using a high setting DLIR (DLIR-H). The control group consisted of 46 age-matching children scanned with 100 kVp, CM dose of 1.3–1.8 ml/kg and images reconstructed with 50% and 100% adaptive statistical iterative reconstruction-V. Two radiologists evaluated images subjectively for overall image noise, vessel contrast and vessel margin clarity separately on a 5-point scale (5, excellent and 1, not acceptable). CT value and image noise of aorta and erector spinae muscle were measured. Compared to the control group, the study group reduced the dose-length-product by 11.2% (p = 0.01) and CM dose by 24% (p < 0.001), improved the enhancement in aorta (416.5 ± 113.1HU vs. 342.0 ± 57.6HU, p < 0.001) and reduced noise (15.1 ± 3.5HU vs. 18.6 ± 4.4HU, p < 0.001). The DLIR-H images provided acceptable scores on all 3 aspects of the qualitative evaluation. “Double low” chest CTA in children using 70 kVp and DLIR provides high image quality with reduced noise and improved vessel enhancement for diagnosis while further reduces radiation and CM dose.