Imaging of pulmonary perfusion using subtraction CT angiography is feasible in clinical practice

Imaging techniques have been developed over the years to study pulmonary perfusion, not only as a tool to investigate the sequelae of vascular obstruction, such as acute and chronic pulmonary embolism (PE), but lately also as a potential tool to characterise inflammation and the malignancy potential of lung lesions [1‐3].

As early as 1964, nuclear medicine has been able to assess pulmonary perfusion using isotopes that accumulate in the capillary bed [4]. Although all modern nuclear medicine methods can accurately quantify true perfusion in the pulmonary parenchyma, they also have important drawbacks, such as a low spatial resolution, issues with isotope availability, production and handling, and high cost [5].

Magnetic resonance imaging (MRI) and computed tomography (CT) are more widely available, but are not capable of imaging true substance exchange. Total intravascular movement can be measured with MRI using arterial spin labelling techniques. In addition, volume and speed of vascular contrast distribution are measured with dynamic magnetic resonance angiography [6]. MRI does not use ionising radiation, and has a better spatial resolution than nuclear imaging techniques, but its images are frequently technically inadequate because of a low signal-to-noise ratio, susceptibility and motion artefacts due to air and respiratory movement, and the low amount of tissue in the lungs [2, 7]. Therefore, MRI is not used as a primary imaging tool of the pulmonary vasculature in most hospitals. CT is far more widely accepted as the modality of choice for evaluation of the lungs because of its higher spatial and temporal resolution [8, 9] and its new ability to display iodine distribution, reflecting pulmonary perfusion.

A CT technique commonly used in clinical practice for assessing pulmonary perfusion is dual-energy CT (DECT). It uses material decomposition of iodine from other materials to visualise the regional pulmonary distribution of intravenous contrast in the pulmonary vessels, including the capillaries [10]. This is accomplished by almost simultaneously irradiating the patient with two x-ray beams of different energy, or by using spectral detectors and then processing the data to generate iodine maps, at a radiation dose similar to or moderately higher than CT pulmonary angiography (CTPA) [11]. This technique can show PE-associated perfusion defects (PD) [12, 13] in concordance with ventilation-perfusion single-photon emission computed tomography (V/Q SPECT) findings [14], with increased the sensitivity in detection of PE at CTPA [12, 15]. However, DECT requires dedicated dual-energy hardware.

An alternative, yet far less widely implemented CT technique to evaluate pulmonary perfusion alongside CTPA is subtraction CT (SCT), made possible by a post-processing technique that does not require special hardware. This technical note will introduce the concept of SCT, along with its clinical interpretation, benefits, limitations, and future perspectives.

SCT involves the subtraction of an unenhanced, pre-injection CT image from an enhanced, post-injection CT image to obtain information on iodine distribution. Since every CT scanner is capable of making an unenhanced and enhanced CT scan, the technique would be more widely applicable than DECT. SCT was introduced by Screaton in 2003 and Wildberger in 2005 [16, 17]. Both studies showed clear visualisation of perfusion defects due to vascular obstruction in anesthetised animals, which guaranteed almost no motion between the two scans.

In phantom experiments with different iodine densities, SCT showed a higher contrast-to-noise ratio between soft tissue and iodine compared to DECT. In DECT, the signal difference between the low- and high-energy images is roughly linearly associated with the local iodine concentration. This implies that the signal of the high-energy image is actually subtracted from the low-energy image, reducing the signal in DECT compared to SCT, where the iodine signal is fully exploited [18, 19]. However, subject motion between the two acquisitions in SCT must be absent or compensated for to result in only the iodine signal. This means that motion correction is the biggest challenge in SCT. With increasing performance and speed of registration algorithms, achieving adequate motion correction is now feasible (Fig. 1). In our practice, we use motion correction software for SCT (^SURESubtraction Lung, Canon Medical Systems) that employs an iterative, non-rigid registration framework to register the unenhanced to the enhanced CT images [20]. This means that both linear and non-linear voxel displacements can be registered. Experiments with dynamic digital phantoms demonstrated that this software can correct motion adequately. Specifically, in simulated scans involving a caudo-cranial diaphragm position difference of 20 mm between the unenhanced and enhanced scans, the 75^th percentile of the lung voxel-to-voxel residual error distance was 1.6 mm [21]. This seems sufficient for evaluating clinically relevant perfusion defects caused by segmental and potentially first-order subsegmental vascular obstructions, as perfusion defects caused by subsegmental embolism are in the centimetre range [16], and the median diaphragm difference between unenhanced and enhanced CT scans in our clinical practice is 5.7 mm in CTPA scans.

SCT is a simple, inexpensive and fast method for imaging and relative quantification of pulmonary perfusion at an acceptable radiation dose. The technique has been implemented in clinical practice in patients with suspicion of pulmonary embolism for several years, while having potential for other clinical applications that evaluate enhancement of the pulmonary parenchyma.

Iodine maps created with SCT are similar to those created with DECT. Artefact behavior is different, however; DECT images suffer from motion artefacts if some of the anatomy in the field of view moves during or between the high- or low-energy scan. The latter effect is non-existent for energy-discriminating detectors, and negligible for a rapid kV-switching system, but could play a role in dual-source systems or systems with dual rotation [18]. Motion artefacts in SCT occur if the patient moves during one of the two scans or if the inspirational difference between scans is major and is not adequately addressed. Streak artefacts due to contrast inflow are less severe with SCT compared to dual-source DECT [24].

SCT- and DECT-derived iodine maps reflect the iodine enhancement at the moment of image acquisition. However, absolute quantification of pulmonary perfusion parameters is impossible to achieve from a single acquisition, be it SCT or DECT, and requires sequential scanning (CT perfusion). This is because the variability in contrast enhancement among patients is high when a single acquisition is evaluated. The pulmonary contrast enhancement not only depends on the presence of pathology and contrast administration parameters, but also on parameters such as respiratory cycle dynamics, cardiac output, and bronchial physiology [27, 28]. In addition, the enhancement pattern in the pulmonary parenchyma may change with time after contrast arrival, especially if there is bronchial arterial collateral supply to areas of the lung that are hypo-perfused by the pulmonary artery [29]. A second scan in the systemic arterial phase can help distinguish these two perfusion components. If sequential scanning were performed at a reasonable dose, absolute quantification might be possible. This might result in the ability to obtain functional (perfusion) information of structures in the pulmonary parenchyma. Other possibilities could be to characterise small structures or lesions, e.g. nodules, with the potential benefit of distinguishing between benign and malignant nodules [30].

Improvements in SCT could be achieved by reducing motion artefacts due to cardiac motion, pulsation and breathing in the pre- and post-contrast images using electrocardiogram (ECG) synchronisation or advanced motion-correction techniques. Improved registration of small vessels would provide not only subtraction imaging of the lung parenchyma, but of the vasculature itself [31]. This can be helpful for direct clot detection (lack of iodine enhancement causes a defect in the vessel on iodine maps). Improved beam hardening would help eliminate streak artefacts due to high contrast concentration in inflow veins [32].

In conclusion, software-based motion correction combined with temporal subtraction enables imaging of contrast enhancement in the lungs similar to DECT. The technique is widely applicable at a radiation dose equivalent to DECT. Similar to DECT, SCT depicts perfusion abnormalities in patients with vascular, bronchial or other pathology. Future advances in pulmonary SCT include dynamic acquisition and reduction of motion artefacts in small structures.