[Evaluation of Low-contrast Detectability of Different Reconstruction Algorithms and Noise Reduction Intensities in the Upper Abdominal Pseudo-human Phantom].

Abstract

Purpose: The effects of reconstruction algorithm and noise reduction intensity on low-contrast detectability in abdominal CT examinations were investigated.

Methods: FBP, hybrid IR, and deep learning-based reconstruction methods (DLR for body, DLR for body sharp) were compared using an upper abdominal pseudo-human phantom. Imaging was performed under four radiation dose conditions, with three noise reduction intensities, and NPS and CNR_LO were used as evaluation indices.

Results: DLR for body sharp showed excellent low-contrast detection performance with strong noise reduction and achieved a higher CNR_LO than the others. Hybrid IR and DLR for body showed equivalent performance regardless of noise reduction intensity, confirming the limitations of low-frequency noise suppression.

Conclusion: It is important to select a reconstruction algorithm and noise reduction intensity according to the purpose of the examination, and DLR for body sharp is useful for improving image quality and reducing exposure at low doses.