Objective This study investigated the impact of deep learning image reconstruction (DLIR) combined with metal artifact reduction (MAR) on metal artifacts in CT scans with lumbar spine implants.

Methods A retrospective analysis was conducted on 40 patients with lumbar metal implants who underwent abdominal spectral CT scans at our hospital. After spectral abdominal scanning, images were reconstructed using Adaptive Statistical Iterative Reconstruction-V (ASiR-V) at 50 % (AR50 group) and DLIR at a high level (DH group). MAR was applied to both groups to obtain reconstructions (AR50-MAR group and DH-MAR group). Regions of interest (ROIs) were delineated and measured for metal artifacts and surrounding tissues on the four sets of images. CT values and standard deviations (SD) were recorded, and the signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), and artifact index (AI) of the images were calculated and compared. Two radiologists subjectively evaluated the quality of the images, the severity of metal artifacts, and the area of the artifacts on a 4-point scale.

Results Under the same energy level of 68 keV, statistically significant differences were observed in SD, CNR, and AI values among the four groups (P < 0.05). The SD values in the MAR groups were lower than those in the non-MAR groups. The images reconstructed with DH-MAR had the highest CNR and SNR and the lowest AI values.

Conclusion For patients with lumbar metal implants, the combination of DLIR and MAR in spectral CT can significantly reduce metal artifacts while greatly improving the signal-to-noise ratio of images of the tissue, which is more conducive to clinical diagnosis.