Effects of radiation dose reduction on diagnostic performance of 3rd generation Dual Source CT pulmonary angiography

To evaluate the effects of radiation dose reduction on diagnostic accuracy and image quality of pulmonary angiography CT (CTPA) in adults with suspected pulmonary embolism (PE).

52 consecutive patients received CTPA for suspected PE. Realistic low-Dose CT simulations were generated using an offline software (ReconCT, Siemens Healthineers, Forchheim, Germany), as either filter back projections (FBP) or iterative reconstruction as ADMIRE (strength 3 or 5) with 25 %, 50 % and 75 % of the original dose. To assess image quality (overall image quality, noise, artifacts, and sharpness) and diagnostic confidence, a five-point scale was used. Patient-based and segment-based diagnostic accuracy was calculated for Low-dose computed tomography (LDCT)-reconstruction with original dose CTPA as a standard of reference. Furthermore, effective radiation doses were calculated using a commercially available dose management platform (Radimetrics, Bayer HealthCare, Leverkusen, Germany).

Among 52 patients, a total of 15 patients (28.8 %) had acute pulmonary artery embolism. The median dose-length product and effective dose for all 52 scans were 291.1 ± 210.1 mGy⋅cm and 5.8 ± 3.4 mSv. Overall subjective image quality was highest for ADMIRE 5 with 75 % and lowest for FBP with 25 % of the original dose (median [interquartile range]:5 [5] vs. 3 [2-3], p < 0.001. Patient-based diagnostic accuracy was perfect for all iteratively reconstructed data sets (ADMIRE 3 and 5) (sensitivity: 100 %, negative predictive value [NPV]: 100 %). LDCT data sets with FBP had perfect diagnostic accuracy at 50 % and 75 % of the original dose, which however decreased at 25 % of the original dose (sensitivity: 93 %; [NPV]: 97 %). Segment-based diagnostic accuracy was high for ADMIRE 3 and 5 down to 25 % dose reduction (sensitivity: 90.4 % specificity: 99.5 %) and lowest for FBP with 25 % dose reduction (sensitivity: 84.6 %, specificity: 98.9 %). Inter-class correlation regarding the detection of PE was almost perfect at all doses and recons (ICC: 96.1-1.0). Thus, accurate diagnosis for PE was possible for ADMIRE 3 and 5 datasets with 25 % of the original dose (1.45 mSv) and for FBP with 50 % of the original dose (2.9 mSv).

Our findings indicate that radiation dose reduction down to 25 % (1.45 mSv) of the original data via iterative reconstruction algorithms on a 3rd generation Dual Source CT (DSCT) scanner maintained the diagnostic accuracy and image quality for the assessment of PE in CTPA.