Evaluation of CT Angiography Image Quality Acquired with Single-Energy Metal Artifact Reduction (SEMAR) Algorithm in Patients After Complex Endovascular Aortic Repair

Between February 2016 and May 2017, we included patients with a routine CT follow-up examination after an F/B-EVAR procedure at our institution. All patients were included only once, and repeated CT examinations were excluded. Our follow-up CT protocol consists of a CT examination (without contrast, arterial, and delayed phase) within 30 days after treatment, followed by a CT examination annually. For this retrospective study, formal informed consent was not deemed necessary by the medical ethical board.

Balloon-expandable Atrium Advanta stents (stainless steel, no radiopaque markers; Maquet, Rastatt, Germany) were deployed in all fenestrations to secure fixation. All branches were connected to the target vessels with Fluency stents (nitinol, tantalum markers; Bard, Tempe, AZ, USA) and fixated with a balloon-expandable Palmaz Genesis stent (stainless steel, no radiopaque markers; Cordis, Baar, Switzerland) in the branch. Self-expandable Cordis S.M.A.R.T. stents (nitinol, tantalum markers; Cordis, Baar, Switzerland) were used for relining. Relining was performed on indication to prevent kinking of the transition between the Fluency or Atrium Advanta stent and the target vessel to ensure a smooth alignment of the vessel with the stent.

All patients were scanned on a 320-detector row CT scanner (Aquillion ONE Genesis edition, Toshiba Medical Systems, Otawara, Japan). The acquisition parameters for the abdominal CTA were as follows: 0.5 s rotation time, 120 kVp, 80-325 mAs (automated exposure control), slice thickness 0.5 mm, collimation of 80 × 0.5 mm, pitch of 65, and the matrix was 512 × 512. The amount of contrast (Ultravist 370; Bayer, Leverkusen, Germany) injected was based on the weight of the patient (1.3 ml/kg) and was injected over a period of 25 s followed by the injection of saline (0.5 ml/kg) over a period of 10 s. Two volumes with 1.0 mm slices were reconstructed for each patient, one using the standard reconstruction technique (Adaptive Iterative Dose Reduction 3D Enhanced (AIDR 3DE), Toshiba Medical Systems), further mentioned as standard CT, and one using the SEMAR algorithm (SEMAR, Toshiba Medical Systems). Reconstruction time of the SEMAR CT images was on average 30 min.

The first step in the SEMAR algorithm is to reconstruct the first-pass image through standard reconstruction and to automatically identify metal traces in this image by means of an HU threshold. The metal segment is identified in the original sinogram through forward projection and removed by linear interpolation using neighboring nonmetal measurements. The interpolated sinogram is reconstructed to create the second-pass image; this image is classified into several tissues (air, water, and bone) through a segmentation process. The second-pass image is forward projected onto the original metal trace sinogram to provide an estimation of the projection attributable to it. The original sinogram is then blended with the tissue-classified image sinogram. The third-pass image is created through a reconstruction of the blended sinogram in which the metal is added to the blended projection to obtain the final image [12].

To objectively evaluate image quality, the images were exported to Philips IntelliSpace Portal (ISP) (Philips Healthcare, Eindhoven, The Netherlands), which is integrated into our PACS (IDS7, Sectra, Linköping, Sweden). Manually drawn circular regions of interest (ROI) were placed in the lumen of peripheral stents in the visceral arteries (celiac, superior mesenteric, and renal arteries) to measure the Hounsfield Units (HU; attenuation) and standard deviation of the HU in the ROI (noise [16]). These ROI were placed in the standard CT images and subsequently copied to the SEMAR CT images in exactly the same location. The ROI were manually placed proximally in the stents of the visceral arteries to include the most severe artifacts (Fig. 3). The inclusion of stent material or calcifications in the ROI was carefully avoided. The average ROI size was 10 mm² for the celiac and superior mesenteric artery (SMA) and 6 mm² for the renal arteries. ROI were drawn in air and in erector spinae muscle to determine the contrast-to-noise ratio [(attenuation of the vessel—attenuation of muscle)/SD air] [7]. CNR is a dimensionless number, it is used as a measure to compare image quality.

To evaluate the subjective image quality, two interventional radiologists (CR: 10 years of experience; RM: 5 years of experience), blinded with regard to the image reconstruction and patient information, independently evaluated the peripheral stents between the standard and SEMAR CT images in the arterial phase presented side by side. The following parameters were assessed: total degree of artifacts 1 = no artifacts; 2 = mild artifacts; 3 = moderate artifacts; 4 = strong artifacts; and 5 = extensive artifacts; and the visualization of the stent lumen 1 = exceptional visualization of the lumen; 2 = good visualization of the lumen; 3 = moderate visualization of the lumen; 4 = poor visualization of the lumen; and 5 = no visualization of the lumen. Finally, the radiologists indicated their preference for either the left or the right CT image.

Statistical analysis was performed with SPSS (version 23.0.0.2, IBM, Armonk, NY, USA). Variables are expressed as median (interquartile range) unless stated otherwise. The quantitative measurements between the SEMAR and the non-SEMAR images and the subjective analysis scale ratings were compared with the Wilcoxon signed-rank test. To assess inter-reader agreement, a linearly weighted k was calculated. A p value of less than 0.05 was considered to represent significant differences.