Multidetector computed tomography (MDCT) has grown as a three-dimensional modality, increasing its value to evaluate complex multiplanar structures like the LAA. A meta-analysis has shown MDCT as a reliable alternative to TEE (the gold standard technique) [20, 21] for the detection of thrombi in the LAA [22].

In terms of MDCT accuracy for LAA thrombus detection , several studies have shown conflicting results, with sensitivities ranging from 29 to 100 %, specificities from 72 to 98 %, and relatively low positive predictive values from 7 to 31 % [23–30]. Despite such extensive ranges, a recent meta-analysis including 19 studies with 2955 patients identified a mean sensitivity of 96 %, specificity of 92 % and positive predictive value of 41 % [22]. The most consistent finding from several studies has been the negative predictive value of MDCT for thrombus detection, with values ranging from 96 to 100 %, with authors suggesting that patients without filling defects on MDCT do not need a TEE [23–29].

Adaptations to the MDCT protocol have been tried to improve the positive predictive value of this imaging technique for thrombus detection, with delayed imaging (at least 30 s after contrast bolus administration) being one of them, increasing the mean positive predictive value to 92 % or higher [22, 31]. The low positive predictive values were explained in part by the static character of the CT study and the fact that the image capture happens a few seconds after contrast arrives to the left heart (including LAA), which can make it difficult to differentiate thrombus from incomplete contrast mixing due to sluggish flow [equivalent to spontaneous echo contrast] (Fig. 8.2a). Adding delayed imaging improves the ability to differentiate these two situations, since a filling defect persisting 1 min after contrast injection is more likely to represent a thrombus, while the filling defect with sluggish flow should improve with contrast opacification on delayed imaging [22]. However, additional delayed imaging increases radiation exposure, which led to the development of a different protocol, involving only one scan after two separate bolus of contrast, a 50 mL timing bonus first, followed by a 70 mL bolus, with a delay of 180 s between injections [32, 33]. Differentiation between thrombus and sluggish flow is then done by assessing contrast attenuation and shape. Thrombus appears as an oval or round shape, whereas sluggish flow appears more as a triangular shape with homogeneous signal intensity. Despite the lower radiation exposure, the use of a double bolus of contrast is not suited for patients with impaired renal function, and this technique still requires validation. Another approach is obtaining delayed scanning in a prone position, which improves contrast mixing.

Newer CT machines are equipped with dual-energy sources, providing simultaneous acquisition of images from low and high voltage settings, which allows evaluation of tissue characteristics and quantitative analysis of the iodine concentration of LAA filling defects, helping in the differentiation of thrombus from sluggish flow (Fig. 8.2b). A preliminary study showed a positive predictive value of 100 % for thrombus identification with dual-energy [33]. Although encouraging, this technique still needs validation in larger cohorts [33].

The high spatial resolution and three-dimensional data provided by MDCT allows adequate morphologic characterization of the anatomy of the LAA, a crucial aspect of LAA occluder device selection [34]. In the literature, different CT machines were used for LAA evaluation and device preplanning. A study of 197 patients with AF who went MDCT prior to radiofrequency catheter ablation used either a 64-detector-row CT scanner or a volumetric 320-detector-row CT scanner. For the 64-detector the settings were: rotation time 400 ms, collimation 64 × 0.5 mm, tube voltage between 100 and 135 kV (depending on body mass index of the patients) and a tube current of 250–400 mA. For the 320-detector, the settings were: rotation time 350 ms, collimation 320 × 0.5 mm, tube voltages between 100 and 135 kV and currents of 400–580 mA. Patients with heart rates above 65 bpm received beta-blockers. The volume of nonionic contrast media used was dependent on body weight, total scan time, and renal function. For the 64-detector, a flow rate of 5 mL/s and a total amount of 80–110 mL were used. For the 320-detector, a total of 60–100 mL of contrast media was infused in sequential steps: 50–90 mL of contrast media infused at a flow rate of 5.0–6.0 mL/s; 20 mL mixture of 50 % contrast media/saline at the same rate; 25 mL saline at a flow rate of 3.0 mL/s. Automated peak enhancement detection in the left ventricle was used for detection of the contrast bolus, and after achieving +180 HU, craniocaudal scanning was initiated. Image acquisition was acquired during an inspiratory breath-hold of 8–10 s. For the 64-detector, ECG was simultaneously recorded for retrospective gating and images were reconstructed at both 30–35 % and 75–85 % phases of the RR interval for the systole and diastole, respectively. For the 320-detector, prospective ECG-triggered dose modulation was used to visualize one entire cardiac cycle with maximal tube current at 75 %, 65–85 % or 30–80 % of the RR interval in patients with heart rates of <60 bpm, 60–65 bpm or >65 bpm, respectively. The mean effective dose of the 320-detector exams was 3.9 ± 1.8 mSv, and for the 64-detector was 18.1 ± 5.9 mSv. The data was reconstructed with a slice thickness of 0.5 mm and with a reconstruction interval of 0.3 mm (64-detector) and 0.25 mm (320-detector) [34].

As the left atrium is a highly compliant chamber, the patients’ volume status will directly impact sizing. For consistency, if hemodynamic status allows, we recommend infusion of at least 500 mL of saline before the scan so the chamber and LAA are close to their maximum dimensions. This is especially important to guide device sizing for LAA closure. Indeed, we have described that volume status can impact LAA dimensions on TEE; patients are often fasting for TEE pre-procedure, and simply administering 500–1000 mL of normal saline during procedure we observed an average increase of 2 mm in the width and depth of the LAA [35].

Digital post-processing review of the LAA and surrounding structures is important to guide device selection and implantation strategy for LAA closure. Several different workstations are available for image processing, e.g., VitreaWorkstation™ (Vital, Toshiba Medical Systems Group Company, The Netherlands), Aquarius Workstation (TeraRecon Inc, Foster City, CA), Brilliance Workspace (Philips Healthcare, Andover, MA), and 3mensio (Pie Medical Imaging, Maastricht, The Netherlands). These workstations have similar capability to enable manipulation and reconstruction of the LAA and surrounding structures to guide percutaneous LAA closures.

The use of three-dimensional digital reconstruction offers additional structural relational portrayal to conventional axial images. Multiplanar reconstruction (MPR) creates volume images by stacking the axial slices. Maximum-intensity projection (MIP) is another volume rendering technique that projects the voxel with the highest attenuation value on every view throughout the volume onto a 2-dimensional image. Thin 3–5 mm MIP may be used to select the best rotation to best visualize the LAA, but measurements are typically taken with MPR. Three-dimensional volume rendering is also often used for LAA assessment, which creates a 3-dimension illustration of CT volumetric data for display from any desired perspective, enabling selection of optimal correlative fluoroscopic angles. This technique allows choosing of different tissues based upon Hounsfield range, and the colors, transparency, and shading can be altered to better represent the volume shown on the image.

Of note, measurements should be taken at the cardiac phase with the largest left atrial and LAA dimensions, which is typically at late atrial diastole, corresponding to 30–40 % of the R-R cycle [36]. The left atrium and LAA changes with cardiac cycle in all 3-dimensional directions, but this does not occur in a uniform fashion with medial-lateral expansion less prominent than longitudinal and anteroposterior expansions [37]. Thus, 1-dimensional assessment may be insensitive to evaluate such changes in LA size. Similarly, the pulmonary vein orifice measurements on CCTA had been shown to vary with cardiac cycle, with the largest diameter in late atrial diastole, with mean decrease by ~30 % during atrial systole [38].

At Vancouver General Hospital, we perform standard pre-procedural and post-surveillance CCTA with the Toshiba 320-detector, see Table 8.1 for protocol. For pre-procedural scans that have to incorporate “rule out” LAA thrombus, the Siemens Dual Source Flash scanner is used, with a different scanning protocol detailed in Table 8.2. We standardly administer 500–1000 normal saline intravenously before imaging.

To assess for suitability for percutaneous LAA closure with the leading devices (i.e., WATCHMAN, ACP/Amulet), evaluation of the LAA shape and dimensions are crucial. The first step is to clearly delineate the orifice/ostium of the LAA and obtain cross-sectional right-angled images of this point. Conventional axial views alone are often inadequate to assess the LAA orifice/ostium, instead, we routinely utilize MPR for this purpose. We identify a view where the circumflex artery, the pulmonary vein (PV) ridge and the LAA orifice/ostium can be clearly seen in one image. The orifice for the ACP/Amulet is the line that connects from the PV ridge to the circumflex artery (echocardiographic LAA ostium). The cross-section of this plane is then obtained at right-angle projections, to improve the co-axial measurement of the orifice (Fig. 8.3). Then another diameter measurement is taken at the landing zone, which is 10 mm (for ACP) and 12–15 mm (for Amulet) inside the orifice (labeled as the neck of the LAA), making sure that the measurement is co-axial at right-angle projections. For WATCHMAN, the ostium of the LAA is measured from the circumflex artery to 1–2 cm within the PV ridge (anatomic LAA ostium). Again using MPR, right-angles to this plane are viewed and manipulated to obtain the best co-axial plane for measurements (Fig. 8.4). If there are trabeculations at the points of measurements, we err on including the trabeculations for larger measurements. LAA diameters that are suitable for device closures are: 12.6–28.5 mm landing zone for ACP, 12.6–32 mm landing zone for Amulet, and 17–31 mm LAA ostium for WATCHMAN.