Keywords:
Image verification, Artifacts, Computer Applications-Virtual imaging, Computer Applications-Detection, diagnosis, CT-High Resolution, CT-Angiography, Vascular, Cardiovascular system, Arteries / Aorta
Authors:
V. De Stasio1, P. C. Douek2, S. A. Si-mohamed3, L. Hanquier2, L. Boussel3, R. P. J. Budde4, S. Boccalini5; 1Rome/IT, 2Lyon/FR, 3Bron/FR, 4Rotterdam/NL, 5Genova/IT
Results
Patients’ characteristics
The mean age was 54,3 ± 19,7 (23-80) and 82% of patients were males.
Manual measurements
We found significantly higher values of aortic diameter measurements between conventional and spectral 40keV images for both observers (mean difference: 0.6±0.6 and 0.7±0.6mm; max difference: 2.4 and 1.8mm; p<0.01) ( Fig. 1 and Fig. 2 ). Differences between conventional and 70keV images were lower (mean difference: 0.36±0.5 and 0.24±0.42mm; max difference 1.8 and 1.1mm; p<0.01).
Significant differences between the two expert observers for each type of images were found (p<0.01).
Automatic measurements
The automatic software was affected by a significant measurement variability between conventional and 60keV images (maximum difference of 1.5mm; p=0.006) ( Fig. 3 ). This was the lowest spectral keV value on which the software was able to identify acceptable vessel contours for at least some of the datasets (20 patients (73%)). Otherwise, the automatic analysis software constantly produced evident errors in the identification of contours for 40keV and 50 keV monoenergetic reconstructions. No significant difference was demonstrated between automatic measurements on conventional and 70keV images (p=0.16).