We performed a retrospective analysis of prospective screened patients admitted to our center over a period of 3 years (2010-2012) showing an acute ischemic stroke treated endovascularly; according to our protocol,
dual-energy CT was performed within the first 12 hours after endovascular treatment to rule out intracerebral bleeding in order to begin the antithrombotic treatment.
We included all dual-energy CT with intracranial hyperdensities.
Thirty-nine patients were included (mean age 67,
range 30 to 85 years),
16 men (mean age 66,
range 49 to 83 years) and 23 women (mean age 67,
range 30 to 85 years).
No patients were lost during the study.
The image acquisition was obtained through a CT Somatom Definition (Siemens Healthcare,
Forchheim,
Germany),
which uses two x-ray tubes (A and B) optimized independently (Kv,
mA) and two detectors in the same gantry.
Both tubes are fired simultaneously therefore the information from both tubes is acquired simultaneously reducing motion or image registration artifacts.
The protocol used was as follows: Tube A at 100 Kv and 250 mA,
Tube B at 140 Kv and 250 mA,
and a 20x0.6 mm collimation (total dose approximately 3 mSv effective,
similar to a conventional head CT).
The use of different energies is based on the attenuation behavior of a substance depends of the energy that is exposed (eg iodine attenuation is greater in 100 Kv than 140 Kv),
enabling create set-images in which some materials are distinguished better than others.
The information obtained was rebuilt in the main console in three different series,
two with a slice thickness of 1.5 mm,
a set corresponding to 100 Kv (Fig. 1 ) and other to 140 Kv (Fig. 2 ); the third set was reconstructed with a slice thickness of 5.0 mm.
This last set of images corresponded to both energy weighted (100 Kv/ 140 Kv) simulating a conventional CT of 120 Kv (Fig. 3 ).
The series is stored in the PACS.
Was carried out post-processing images of 100 Hv and 140 Kv by means of a software (syngo Dual-Energy Brain Hemorrhage,
Siemens) using a 3-material decomposition algorithm based on brain parenchyma,
hemorrhage,
and iodine.
The program separates each voxel and compare the attenuation of two preset material (cerebral parenchymal and hemorrhage) in each set-images,
100 Kv and 140 Kv,
thus the attenuating both substances must be linear in each series,
when this linearity alters can be attributed to the presence of one third substance which is also preselected (iodine).
This allows obtaining an iodine map (for displaying this material,
Fig. 4) and a virtual noncontrast map (to visualize brain parenchymal and hemorrhage,
Fig. 5).
The technique has a limitation; it can not distinguish more materials than 3,
so the presence of calcium or metal materials can not be differentiated.
Extracted studies were randomized into 2 groups of reading (A and B),
21 and 18 patients each one.
Group A was made up of a mixture map (weighted sum of the 100-kV and 140-kV images) simulating a conventional nonenhanced CT.
Group B was made up of a dual-energy CT.
Both groups were analyzed by three blinded observers to the clinical characteristics of each patient,
2 neuroradiologists with over 10 years of experience (NR1 and NR2) and a second-year resident (R2).
The image findings were classified as contrast extravasation (I) or hemorrhage according to ECASS radiological scale of cerebral bleeding (European Cooperative Acute Stroke Study,
Table 1).
The results were dichotomized: Value 0 if correspond to contrast extravasation or haemorrhagic stroke (I/IH1/IH2 = 0),
Value 1 if correspond to a parenchymal hematoma or remote hematoma (PH1/PH2/RH = 1); according with our protocol,
the patients with value 0 (I/IH1/IH2) were treated in the first 24 hours with intravenous heparine.
Therefore,
patients with value 1 (PH1/PH2/RH) were contraindicated to antiacoagulation therapy.
So,
this was a dichotomization according to a treatment decision.
Finally,
we analyzed the correlation of the readings between the 3 observers through the Kappa index value (k,
Table 2) classifying the poor to excellent correlation using SPSS for statistical analysis.