Deep Learning Reconstruction in Pediatric Low-Dose Computed Tomography: A Systematic Review of Image Quality and Radiation Dose Reduction.
Authors/Creators
Description
Background: Radiation exposure remains a critical concern in pediatric computed tomography (CT) due to heightened radiosensitivity and increased lifetime attributable risk of malignancy. Technological advances in image reconstruction have enabled substantial radiation dose optimization. Deep learning reconstruction (DLR), a novel artificial intelligence-based approach, has emerged as a promising method to improve image quality while permitting further dose reduction.
Purpose: To systematically evaluate current evidence regarding the impact of deep learning reconstruction on image quality and radiation dose reduction in pediatric low-dose CT.
Method: A systematic review was conducted in accordance with PRISMA 2020 guidelines. PubMed and Google Scholar were searched for English-language original research article published between January 2019 and March 2026. Studies were eligible if they include pediatric patients (≤18 years), evaluated CT imaging with deep learning reconstruction and reported objective or subjective image quality metrics and/or radiation dose outcomes. Data were extracted using standardized forms, and risk of bias was assessed using an adapted Newcastle-Ottawa framework. Due to heterogeneity in reporting and incomplete availability of variance data, a structured narrative synthesis was performed.
Result: Fifty-three records were identified; for studies met inclusion criteria. All included studies demonstrated significant reductions in image noise and improvements in signal-tonoise ratio (SNR) and contrast-to-noise ratio (CNR) with DLR compared to conventional iterative reconstruction techniques. Two studies reported radiation dose reductions of approximately 50-55% while maintaining diagnostic image quality. Risk of bias was low to moderate across studies.
Conclusion: Available pediatric evidence indicates that deep learning reconstruction improves image quality and enables meaningful radiation dose reduction in low- dose CT protocols. Although current data remain limited, findings consistently support integration of DLR into pediatric CT imaging strategies.