Radiation dose reduction in pediatric computed tomography (CT) using deep convolutional neural network denoising

We evaluated the quality of noncontrast chest computed tomography (CT) for pediatric patients at two dose levels with and without denoising using a deep convolutional neural network (CNN). Forty children underwent noncontrast chest CTs for “chronic cough” using a routine dose (RD) protocol. Images were reconstructed using iterative reconstruction (IR). A validated noise insertion method was used to simulate 20% dose (TD) data for each case. A deep CNN model was trained and validated on 10 cases and then applied to the remaining 30 cases. Three certificate of qualification (CAQ)-certified pediatric radiologists evaluated 30 cases under 4 conditions: (1) RD + IR; (2) RD + CNN; (3) TD + IR; and (4) TD + CNN. Likert scales were used to score subjective image quality (1–5, 5 = excellent) and subjective noise artifact (1–4, 4 = no noise). Images were reviewed for specific findings. For the 30 patients evaluated (14 female, mean age: 10.8 years, range: 0.17–17), the mean effective dose was 0.46 ± 0.21 mSv for the original RD exam, with an effective dose of 0.09 mSv for the TD exam. Both RD + CNN (3.6 ± 1.1, p < 0.001) and TD + CNN (3.4 ± 0.9, p = 0.023) had higher image quality than RD + IR (3.1 ± 0.9). Both RD + CNN (3.2 ± 0.9, p-value =