Improving Image Quality of Thin-Slice and Low-keV Images in Dual-Energy CT Angiography for Children With Neuroblastoma Using Deep Learning Image Reconstruction

Neuroblastoma (NB) is a common malignant tumor in children, and the evaluation of vascular involvement image-defined risk factors (IDRFs) using computed tomography angiography (CTA) is crucial for prognostic assessment. To evaluate whether deep learning image reconstruction (DLIR) can improve the image quality of thin-slice, low-keV images in dual-energy CTA (DECTA) and provide a more accurate assessment of IDRFs in children with NB. Forty-three NB patients (median age: 2 years., 6 months to 7 years), who underwent chest or abdominal DECTA, were included. The 0.625 mm slice thickness images at 40 keV were reconstructed using high-strength DLIR (40 keV-DL-0.6 mm) in the study group. The 0.625 mm images at 40 keV and 5 mm images at 68 keV, reconstructed using the adaptive statistical iterative reconstruction-V (ASIR-V) with a strength of 50% (40 keV-AV-0.6 mm,68 keV-AV-5 mm, respectively), served as the control group. Objective measurements included the contrast-to-noise ratio (CNR) and edge-rise slope (ERS) of the aorta, and magnitude of noise power spectrum (NPS) of the liver. Subjective image quality was assessed using a 5-point scale to evaluate overall image noise, image contrast, and the visualization of large and small arteries. The IDRFs were also evaluated across all images. In general, the 0.625-mm images had higher spatial resolution and more confident IDRF assessment compared to the 5-mm images. The 40 keV-DL-0.6-mm images demonstrated the highest CNR and ERS of large vessels, and the best visualization of small arteries among the three image groups (all p < 0.05). Subjective assessments revealed that only the 40 keV-DL-0.6 mm images met diagnostic requirements for overall noise, image contrast, large artery, and small artery visualization simultaneously. DLIR-H significantly improves the image quality of the thin-slice and low-keV images in DECTA for pediatric NB patients, enabling improved visualization of small arteries and more accurate assessment of vascular involvement IDRFs in NB.