Study population
Thirty internal metallic implants in 20 consecutive cadavers (9 males and 11 females,
mean age 74 ± 15 years,
range 43–92 years) submitted to forensic investigation,
including post mortem CT examination,
were examined in the study (Table 1).
Inclusion criterium was: presence of one or more implanted metal device.
Exclusion criteria: cases with major alterations of the soft tissue in proximity of the implant,
for example those caused by putrefaction and lacerations.
CT data acquisition
All examinations were performed on a Dual Source CT (Somatom Definition Flash,
Siemens Medical,
Forchheim,
Germany).
Six scans were performed in each case: three SECT and three DECT,
respectively at three different fixed CT dose index (CTDI): 20.0 mGy (group 1),
10.0 mGy (group 2) and at 5.0 mGy (group 3).
The acquisition parameters are specified in the table 2 (Table 2).
CT image reconstruction and processing
SECT and DECT raw data sets were reconstructed with a slice thickness of 1.5 mm and an increment of 1 mm using a fixed field of view of 200 mm (image matrix 512 × 512),
and two sharp convolution kernels.
Post-processing of DECT data sets was performed using commercially available software (syngo.via,
software VA31,
monoenergetic application) installed on a Leonardo workstation (Siemens HealthCare,
Forchheim,
Germany),
that allows extraction of the monoenergetic images at arbitrary photon energies ranging from 40 keV to 190 keV.
Then,
from DECT axial reconstructions were used to obtain six monoenergetic data sets at 64,
69,
88,
105,
120,
130 keV and at an optimized keV value (OPTkeV) were obtained.
Eight images (7 DECT,
and 1 SECT corresponding to the slice with the thickest area of the metallic implant and the most pronounced streak artifacts was selected in axial plane images) were used for the analysis in each case for each group (total of 720 images).
CT image evaluation
Qualitative analysis (Figure 1) was performed by two radiologists blinded to each other,
using a 4-point Likert scale as follows: (1) absence of artifacts and full diagnostic interpretability; (2) minor artifacts and confident diagnostic interpretability; (3) major artifacts and impaired diagnostic interpretability but still suitable for diagnostic purposes; (4) massive streak artifacts and abolished diagnostic interpretability.
The quantitative analysis of the artifacts was performed by calculating the density (Hounsfield Unit value – HU) of the most marked streak artifact.
For the measurement,
a single observer used a circular region of interest (ROI) that was placed within the hypodense streak area adjacent to the metallic implant,
carefully avoiding partial volume artifacts (Figure 2).
Another ROI providing a reference density measurement was traced in an area outside of the streak artifact.
The artifact intensity was calculated by subtracting the streak density from the reference density.
The quantitative analysis was repeated by the same observer after 1 month to assess intra-observer variability.
Statistical analysis
Qualitative analysis
Inter-reader agreement of categorical variables was analyzed by means of kappa statistics.
Kappa statistics was run on all the 720 images available (8 images of 30 cases in the three groups).
The quality of all images within all the groups was compared by means of a χ2 test.
Finally,
the same test was applied to detect differences between different dose scanning groups.
Quantitative analysis
Intra-reader agreement of quantitative assessment was analyzed by means of Intraclass Correlation Coefficient (ICC).
The quantitative assessment of the monoenergetic OPTkeV and SECT images was compared with respect to the Mann–Whitney test,
whereas in order to analyze SECT and OPTkeV quantitative assessment in relation to different dose scanning protocols the Kruskal–Wallis test was applied.