Author: Samuel L. Brady, Joseph G. Meier 👨🔬
Affiliation: Cincinnati Childrens Hospital Med Ctr 🌍
Purpose:
To establish noise reference levels for our pediatric hospital.
Methods:
Water equivalent diameter (Dw) and image noise was automatically measured by using a global noise algorithm in soft tissue. Routine head exams (n= 5195) without contrast and routine abdomen – pelvis (AbdPel) exams with contrast (n = 2031) were analyzed of patients aged 0-18 years old (y.o.). Patients were scanned on either a Canon Aquilion Genesis or a GE HealthCare Revolution Apex system using tube current modulation and age or weight-based protocols.
Results:
CTDIvol dose reference levels (DRL) and noise reference levels (NRLs) are reported as the median [Inter Quartile Range]. Head (0-1 y.o.) DRLs for GE (n=661) and Canon (n=227) were 23.7 [23.7 23.7] and 21.7 [21.7 23.4] mGy while NRLs were 2.7 [2.4 2.9] and 2.4 [2.4 2.7] HU respectively. Head (6-18 y.o.) DRLs for GE (n=1699) and Canon (n=814) were 34.0 [34.0 43.3] and 35.0 [28.0 35.0] mGy while NRLs were 3.7 [3.4 3.9] and 3.0 [2.9 3.2] HU respectively. AdbPel (Dw 12 – 16 cm) DRLs for GE (n=128) and Canon (n=65) were 1.3 [0.8 2.2] and 1.1 [0.9 2.2] mGy while NRLs were 10.9 [10.1 11.9] and 6.4 [5.7 7.5] HU respectively. AdbPel (Dw 20 - 24 cm) DRLs for GE (n=346) and Canon (281) were 1.9 [1.6 2.2] and 2.1 [1.8 2.3] mGy while NRLs were 11.5 [11.2 12.1] and 8.0 [7.7 10.9] HU respectively. This analysis discovered that at a certain timepoint the Canon AiCE deep learning (DL) reconstruction was disabled, resulting in elevated image noise. The Canon noise across all AbdPel exams increased from 7.7 [7.1 9.2] to 11.3 [10.2 13.2] HU (p < 0.005) for DL and non-DL reconstructions respectively.
Conclusion:
This work demonstrated the utility of NRLs which identified protocols with elevated noise caused by modified reconstruction parameters.