Submillisievert CT using model-based iterative reconstruction with lung-specific setting: An initial phantom study.

Abstract

To assess image quality of filtered back-projection (FBP) and model-based iterative reconstruction (MBIR) with a conventional setting and a new lung-specific setting on submillisievert CT. A lung phantom with artificial nodules was scanned with 10mA at 120 kVp and 80 kVp (0.14mSv and 0.05mSv, respectively); images were reconstructed using FBP and MBIR with conventional setting (MBIRStnd) and lung-specific settings (MBIRRP20/Tx and MBIRRP20). Three observers subjectively scored overall image quality and image findings on a 5-point scale (1 = worst, 5 = best) compared with reference standard images (50mA-FBP at 120, 100, 80 kVp). Image noise was measured objectively. MBIRRP20/Tx performed significantly better than MBIRStnd for overall image quality in 80-kVp images (p < 0.01), blurring of the border between lung and chest wall in 120p-kVp images (p < 0.05) and the ventral area of 80-kVp images (p < 0.001), and clarity of small vessels in the ventral area of 80-kVp images (p = 0.037). At 120 kVp, 10mA-MBIRRP20 and 10mA-MBIRRP20/Tx showed similar performance to 50mA-FBP. MBIRStnd was better for noise reduction. Except for blurring in 120 kVp-MBIRStnd, MBIRs performed better than FBP. Although a conventional setting was advantageous in noise reduction, a lung-specific setting can provide more appropriate image quality, even on submillisievert CT. • Lung-specific submillisievert 10mA-MBIR CT setting has similar performance to 50mA-FBP • The new lung-specific settings improve vessel clarity and blurring of borders • The new settings may provide more appropriate images than conventional settings.