In coronary CT angiography subtraction, an image is obtained by registering the precontrast image with the postcontrast image and subtracting the areas recognized as calcification. For registration to be successful, it is important to avoid “misalignment” of the information on precontrast and postcontrast images. Given that the spatial resolution of the image is considered to affect the result of subtraction, we performed a preliminary experiment using a phantom to determine the appropriate imaging conditions.

Two polyvinyl chloride infusion tubes with an inner diameter of 2.2 mm were prepared as simulated blood vessels. One of the two tubes was coated with oil clay containing stone powder and calcium carbonate to simulated a calcified lesion and fixed in a plastic case. To reduce the artifact caused by the difference in CT values between the simulated vessel and its surrounding area and for stability during scanning, the plastic case was filled with water in which gelatin was dissolved and allowed to cool and solidify (Fig. 1).

Phantom used in the preliminary experiment. Polyvinyl chloride infusion tubes were used as the simulated vessel and oil clay containing stone powder and calcium carbonate (arrow) as the simulated calcified lesion

Using a 64-row multislice CT scanner (Supria Grande® Fujifilm Medical, Tokyo, Japan), two scans of the phantom simulating blood vessels were obtained. One scan was obtained after injection of saline and the other after injection of contrast agent to acquire the raw data set. The following settings were used for acquisition of the raw data: voltage 120 kV, current 75 mA, and collimation 0.63 × 32 mm, rotation time 0.75 and helical pitch 0.5938 for the non-contrast images and voltage 120 kV, current 150 mA, and collimation 0.63 × 32 mm, rotation time 0.75 and helical pitch 0.5938 for the contrast images. The contrast agent used was oil-based (iodinated ethyl esters of fatty acids obtained from poppyseed oil; MIRIPLA Suspension Vehicle®, Sumitomo Pharma Co., Ltd., Tokyo, Japan) mixed with liquid paraffin and adjusted to have a CT value of about 450 HU. The raw data obtained from each scan were reconstructed with a slice thickness of 0.625 mm under the following four conditions: field of view (FOV) 120 mm + soft tissue kernel; FOV 120 mm + bone kernel; FOV 100 mm + soft tissue kernel; and FOV 100 mm + bone kernel.

Non-contrast (saline injection) and contrast images were used as the data set, and subtraction was performed on the images created under each of the above four conditions using the Cerebrovascular Subtraction application in the Synapse Vincent 3D image analysis system (Fujifilm Medical, Tokyo, Japan) to create images of subtracted simulated calcified lesions (Fig. 2). This application removes any CT value above a certain threshold (≥ 200 HU in this study) as the target of subtraction, leaving pixels with a difference in signal value before and after contrast (i.e., contrasted pixels). The rigid method (where the moved image is matched to the target image by transformation coefficients that include only translation and rotation elements) and the non-rigid method (where a transformation vector is created to move each element of the transformed image to the corresponding pixel position in the target image, and the image is transformed to match the target image) are used in this application for image registration.

Images of simulated calcified lesions. (a) Image obtained at the time of saline injection (precontrast). (b) Image obtained at the time of contrast injection (postcontrast). (c) Calcification subtraction image created using precontrast and postcontrast images as the data set. The images below show each cross-section of a slice of the calcified area

The lumen diameter of the simulated vessel was measured on the post-subtraction image. Using this method, the Vincent analysis system creates a profile curve of CT values on a specified straight line and automatically measures the half-width in any range on this curve. Using this function, a straight line was placed through the center of the lumen of a simulated post-subtraction vessel, and the lumen diameter was defined as the half-width in the range indicated above 0 HU on the profile curve (Fig. 3). The lumen diameters of two simulated vessels (one with calcified lesions and the other without) in the same six slices were measured and the mean value was defined as its lumen diameter. The lumen diameter under each of the four conditions was compared. To evaluate the intra-observer reliability, the first observer performed the same measurement again after a 2-week washout period. The intra-observer reliability was assessed in terms of intraclass correlation coefficients (ICCs) and 95% confidence intervals (CIs) using a two-way mixed effects model with absolute agreement for a single measurement. ICCs were also calculated to evaluate the inter-observer reliability between the two observers, using a two-way random effects model with absolute agreement for a single rater. Four levels of reliability were defined based on the classification of ICC values proposed in [8]: poor, < 0.50 moderate, 0.50–0.74; good, 0.75–0.89; excellent, ≥ 0.90. All analyses were performed using SPSS Statistics ver. 26 (SPSS software, Armonk, NY).

Profile curve of CT values. A straight line was placed through the center of the simulated postsubtraction vessel lumen on the cross-sectional image, and the lumen diameter was defined as the half-width (double-headed arrow) in the range indicated above 0 HU on the profile curve

Two cases with highly calcified coronary arteries on postmortem CT images (Agatston calcification scores [9] of 1500 and 3570) were included. In both cases, the forensic autopsy was performed at our university with the approval of our ethics committees (approval no. 2987). After the heart was removed, catheters were inserted into the right and left coronary arteries. A Supria Grande 64-row multislice CT scanner was used to obtain precontrast and postcontrast images using a previously reported postmortem coronary angiography method [5] in which a pressurized bag is used to inject contrast medium into the removed heart. When acquiring the precontrast images, saline was injected into the coronary arteries using an infusion solution bag through a catheter, and intravascular gas was removed. A three-way stopcock was connected between the tube for injecting saline and the tube for injecting contrast medium, making it possible to switch between the injections into the catheter without moving the heart. The post-contrast images were acquired at the same position as those recorded during the pre-contrast image scanning. Additionally, the two imaging trajectories before and after administration of contrast were synchronized using the orbitally-synchronized scan function of the Supria Grande scanning system. The conditions for the two scans were the same as in the preliminary experiment. Previous reports have shown that water-soluble contrast medium leaks out of the myocardial interstitium, whereas oil-based contrast medium remain inside the vessel [10, 11]. To ensure visualization of the vessels for successful subtraction, we used the same oil-based contrast medium that was used in the preliminary experiment. In addition, the angiography method used in the present study is to inject the contrast medium for approximately 3 min (approximately 15 mL each in the left and right coronary arteries) and then perform the scan while maintaining the injection. This method can be used to visualize the peripheral vessels as well as the main coronary artery branches.

Reconstructed images were then created for the left and right coronary arteries using a slice thickness of 0.625 mm, the bone reconstruction kernel, and an FOV narrowed to approximately 100 mm. The Supria Grande also performs reconstruction by “iterative approximation processing based on a statistical model (Intelli IP),” which is expected to reduce artifacts and noise and improve image quality.

These precontrast and postcontrast reconstructed images were used as the data set for removal of calcification using the “Cerebrovascular Subtraction” application in the Vincent software, which was also used in the preliminary experiment. Three-dimensional and curved planar reconstruction images were created. Calcified areas were identified on the presubtraction images and stenotic or occlusive lesions on the postsubtraction images. For areas with severe calcification, including suspected stenotic or occlusive lesions, we compared the presubtraction and postsubtraction images of the luminal cross-section of the vessel with the gross findings for the decalcified coronary artery section.