CT dose optimization for the detection of pulmonary arteriovenous malformation (PAVM): A phantom study

Abstract Purpose To determine the lowest suitable dose level for the detection of pulmonary arteriovenous malformation (PAVM) using a task-based image quality assessment. Material and methods A phantom was scanned using the standard chest protocol (STD) and 4 other ultra-low dose protocols (ULD) using various kVp. Raw data were reconstructed using level 5 of the hybrid iterative reconstruction algorithm (iDose4) for the STD protocol, and level 6 of iDose4 and levels 1 to 3 of model-based iterative reconstruction (IMR) for the ULD protocols. Both quantitative criteria and qualitative analysis were used to compare protocols. Noise-power-spectrum and Task-based transfer function were computed using imQuest software. The detectability-index (d’) was computed for the detection of PAVM. A subjective analysis was performed by 2 chest radiologists to validate the image-quality obtained on the anthropomorphic phantom for all protocols. Results Similar d’ values were found for ULD-140 using iDose4 6 compared to STD protocol. Greater d’ values were found for all ULD protocols using IMR compared to STD. Subjective image quality was rated as acceptable to excellent for ULD-140 and ULD-120 for all reconstruction types, for ULD-100 and ULD-80 using IMR2, and for ULD-100 using IMR1. Image smoothing was poor for IMR3 for ULD-100 and ULD-80. Finally, the ULD-80 protocol reconstructed with IMR2 was chosen for the detection of PAVM. With this protocol, the dose (CTDIvol of 0.3 mGy) was reduced by 91% compared with the STD protocol. Conclusion A dose level as low as 0.3 mGy reconstructed with IMR2 provides an image quality suitable for the detection of PAVM.