Technique, pitfalls, quality, radiation dose and findings of dynamic 4-dimensional computed tomography for airway imaging in infants and children

Some extended detector row multi-detector CT scanners allow for volumetric imaging of the entire airway during free breathing without having to move the patient through the gantry. This avoids misregistration of the airway position due to normal craniocaudal movement of the airway during respiration and allows 3-D reconstruction and viewing of cine-loops during breathing [1]. This technique is ideal for evaluating tracheobronchomalacia in children.

Tracheobronchomalacia is defined as excessive collapsibility of the airway, which is either idiopathic or secondary to extrinsic compression [2, 3]. Bronchoscopic diagnosis is subjective and defined as proportional airway collapse of more than 50% of the lumen, compared to the normal airway, under self-ventilation [4, 5]. For a classical imaging diagnosis, a proportional collapsibility of the airway of greater than 50% must be demonstrated [2]. In meeting this definition, tracheograms, bronchograms and fluoroscopy cannot quantify the airway cross-sectional area but instead quantify anteroposterior diameter [6]. Dynamic four-dimensional (4-D) computed tomography (CT) provides images for subjective evaluation and more objective measurement of airway collapse, noninvasively, through volume imaging of the whole airway during free breathing, without moving the patient through the gantry and by providing information on airway dynamics in a cine-loop format [1, 3]. For free-breathing dynamic airway assessment using dynamic 4-D CT, tracheobronchomalacia has been defined as airway diameter reduction of more than 28% [1, 7].

Limited availability of 320-detector row CT scanners with volume scanning capabilities and the perception that dynamic 4-D CT imparts a higher radiation dose than bronchography have restricted its use [8]. We aim to present our early experiences with dynamic 4-D CT in children to assist those wishing to start such a program.

Tracheobronchomalacia is the excessive collapse of the trachea and/or bronchi during expiration [5, 11]. Congenital tracheomalacia is the most common congenital tracheal abnormality while acquired tracheomalacia results from an insult, e.g., trauma, external compression, positive pressure ventilation, infection or inflammation [6]. Focal tracheomalacia in children is seen with congenital compression of the trachea or through prolonged intubation and in these patients the trachea is often stenosed on inspiration and collapses further on expiration [3]. Severe tracheobronchomalacia requires treatment [4] dependent on the site, cause and comorbidities [4]. Treatment options include conservative management such as artificial ventilation with high post-expiratory pressure and interventions including aortopexy, tracheostomy, splinting and stent placement [4, 12]. The leading cause of tracheobronchomalacia is vascular compression (48%) [4] and, therefore, aortopexy is frequently used [12]. This technique is also utilised with oesophageal atresia and idiopathic tracheomalacia [4]. Intraluminal stents are less desirable because they can become dislodged or obstructed [12]. Treatment depends on imaging for demonstrating a stenosis, diagnosing tracheobronchomalacia, distinguishing focal tracheobronchomalacia from diffuse tracheobronchomalacia and identifying vascular anomalies.

Traditionally, diagnosis of tracheobronchomalacia is by bronchoscopy, which has significant disadvantages: it is invasive [7]; requires general anaesthesia and positive-pressure ventilation (masking tracheobronchomalacia) [5, 13]; it risks complications (including death) [5]; is operator dependent; is subjective, lacking airway measurements (and underestimating collapse), and small airways in infants are not well seen [5, 7, 13]. Fluoroscopy, tracheobronchography, multi-detector CT and dynamic magnetic resonance scanning are also used for diagnosis [6]. Fluoroscopy deserves attention because it is noninvasive, quick, does not require contrast or patient cooperation and yields dynamic information with high specificity (reported to be 94–100%) [13], but it lacks sensitivity (23.8%–62%) [1, 6], is poor at demonstrating anatomical detail, is two-dimensional, is subjective and also underestimates collapse [3, 13]. Contrast bronchography is preferred by many because it shows tracheobronchomalacia can be repeated at varying pressures of ventilation [8, 14] and is more sensitive than bronchoscopy. However, the airway can only be assessed in one plane at a time with overlap of bronchi on the lateral view, again resulting in underestimation of tracheobronchomalacia [13], it is invasive, requires anaesthetic support with intubation and involves injection of contrast material into the airways, risking desaturation from alveolar flooding [13]. Despite these risks, bronchography is considered safe in experienced units [14]. Radiation effective dose from bronchography ranges between 0.26 and 2.47 mSv [13], which is similar to our bronchographic studies (0.3–3.5 mSv). Spirometer-controlled cine magnetic resonance imaging (MRI) is not widely used because it can take from 9 to 20 min to perform, requires patient cooperation and lacks the spatial resolution of CT; thus, it is only feasible for children older than 8 years [5]. CT is a noninvasive alternative, but concerns regarding radiation [5] from imaging during inspiration and expiration (double the dose of radiation) have hampered its widespread use in children [6, 15, 16]. Dynamic 4-D CT has many advantages because it is noninvasive, dynamic (providing airway inspiratory and expiratory information during physiological breathing) [17], fast; high-quality [2, 15]; objective for measurement of airway collapse [3] and craniocaudal extent (i.e. distinguishes focal from diffuse tracheobronchomalacia) [11, 18], demonstrates adjacent structures [3] and it allows 3-D reconstructions [11] (Fig. 9). CT is most useful for simultaneously demonstrating tracheobronchomalacia and any cardiovascular cause [3].

Short z-axis coverage of narrow-array CT scanners requires helical scanning, resulting in different airway segments being imaged at varying phases of respiration [18], because expiration occurs earlier near the carina than in the more proximal trachea [1, 6]. In contrast, volume scanning allows the entire length of the airway to be scanned simultaneously [1]. Wide-detector scanners such as the 320-detector row CT volumetric scanner used by us, Kroft et al. [19] and Greenberg and Dyamenahalli [7] provide coverage up to 16 cm allowing inclusion of the whole paediatric chest without moving the table [1] [19]. The more modern version of this scanner (Aquilion ONE Vision edition, Otawara, Japan) can acquire at 0.275 s per full rotation if at maximum speed providing isometric, isophasic and isovolumetric 4-D imaging in a real-time respiratory cycle [2], i.e. in non-sedated and non-intubated children [2, 7]. The acquisition includes multiple gantry rotations, divided into separate temporal dynamics (usually 4–5 gantry rotations). Half-scan reconstruction usually allows for the creation of eight dynamic phases from the four rotations, for viewing dynamic airway changes in cine mode [2, 7] as 3-D and multiplanar reconstructions [1]. Reconstructions improve on axial scans when findings are subtle, when determining craniocaudal extent of focal tracheobronchomalacia, for viewing oblique and complex airway anatomy and when communicating information [2] (Supplementary material 6).

Greenberg (in 2012) [1] performed dynamic 4-D CT in 24 infants and small children including intubated patients, much like we did. We based our definition of tracheobronchomalacia of greater than 28% reduction in airway area during expiration when not using forced respiration on previous reports [1, 7]. In 2014, Greenberg and Dyamenahalli [7] reported dynamic 4-D CT in 17 children: 2 of 12 with tracheomalacia had vascular compression and 4 of 8 with left bronchomalacia had atrial or vascular compression. In their patients, dynamic 4-D CT allowed detailed simultaneous evaluation of airway and vascular abnormalities resulting in management changes in 70% [7]. We also demonstrated cardiovascular abnormalities with contrast-enhanced dynamic 4-D CT in 12 patients (36%); of these, 9 had tracheobronchomalacia. Our findings changed management significantly in 5 children (25%) who underwent surgical interventions. Long-term follow-up of these cases is not yet available. In another 14 children (70%), conservative management, including optimisation of ventilation in intensive care patients or watchful waiting in ambulant patients, resulted in good short-term outcomes. One patient died from their underlying cardiac condition.

The pitfalls and poor quality dynamic 4-D CT encountered at the start of our program detract from the success of later studies. At first, we failed to recognize the value of routine intravenous contrast administration in demonstrating unsuspected vascular anomalies, causing secondary tracheobronchomalacia (reported in 20%-81%) [4, 13]; we now administer intravenous contrast routinely. Lee et al. [15] showed 53% prevalence of tracheobronchomalacia in symptomatic children with mediastinal aortic vascular anomalies -- the relevance being that respiratory symptoms from tracheobronchomalacia may persist after surgical correction of the vascular anomaly. The presence of an endotracheal tube in the group referred from neonatal intensive care is also an avoidable pitfall, which can distort the trachea, affect dynamic changes [6] and result in major artefact (Fig. 4). We overcame this pitfall by repositioning the endotracheal tube tip just above the thoracic inlet, using the initial scout view as a guide at the start [11, 20] (Fig. 2). A third avoidable pitfall is an in situ nasogastric tube (18% of our patients) (Fig. 3) (Supplementary material 4), which causes artefact and which we now routinely remove before dynamic 4-D CT. The unavoidable presence of oesophageal air is a prominent feature in crying babies and affects volume-rendered 3-D reconstructions in awake children (Figs. 2 and 5).

Motion artefact, exaggerated by indwelling tubes, affected quality. Even though volumetric CT is up to 24 times faster than helical CT [19], continuous scanning over 1.4 s allowed for motion artefact in 55% of our patients, yet rendered only 1/3 of studies nondiagnostic. The narrative is important again, in that at the start of the service, care of the child during scanning was left to the clinical team and was without anaesthesia in 82% of cases. Anaesthesia is undesirable because it minimizes changes in transmural pressure masking tracheomalacia [6] and is dependent on anaesthetic services. Even with the use of anaesthesia, Lee et al. [17] encountered motion artefact in 20% of their combined inspiratory and expiratory dynamic CT in children. Some movement artefact is unavoidable because the position of the tracheobronchial tree changes during the respiratory cycle in the craniocaudal direction. Dynamic 4-D CT is not affected by the craniocaudal movement of the airway [13], but anteroposterior movement can cause the airway to be out of plane for minimum intensity coronal slabs. Presence of a gowned radiologist within the scanner room until immediately preceding the scan and use of a vacuum immobilisation device (RedVac VMR433X01, VMR438X01; Kohlbrat and Bunz, Austria) improved our quality. We also improved quality by reconstructing a half rotation of projection data rather than a full one, improving temporal resolution and creating smoother cine-loops.

Effective radiation doses from paired inspiratory and expiratory multi-detector CT range between 3.5 and 7.5 mSv [11], but recent studies have shown reductions of up to 23% [6, 13, 16]. Low-dose studies are diagnostic, despite increased image noise due to the natural high contrast of the airway and lungs [1, 11, 13, 16]. Volume CT imparts up to 40% lower radiation doses because there is none of the z-overranging associated with helical CT and less overbeaming (penumbra effect) [1, 19]. We also recommend limiting the craniocaudal acquisition from just below the vocal cords to 3 cm below the level of the carina [11, 13, 19] except where whole-lung demonstration is needed (Supplementary material 7); limiting continuous scan time to a single breath [1], using iterative reconstruction [7] and switching to an Adaptive Iterative Dose Reduction reconstruction algorithm [7] for further radiation dose reduction. Greenberg [1] initially (2012) reported a mean effective dose of dynamic 4-D CT in children of 1.7 mSv (standard deviation [SD], 1.1 mSv), which improved to 1.1 mSv (range: 0.4–1.9 mSv) in 2014 [7]. This matches our experience and compares favourably against bronchography [13] but contrasts with a 2005 report by Mok et al. [14] where bronchography (0.27–2.47 mSv) performed better than helical CT (0.86–10.67 mSv). Our benchmarking against bronchogram doses (mean: 1.4 mSv; range: 0.3–3.5 mSv) compares well with the report by Mok et al. [14] but exceeds our dynamic 4-D CT doses. Volume scan mode is only available with specific vendors, but no additional kit is required and there is no extra cost once the scanner is purchased. Our scanner was installed for imaging major trauma and the possibilities from volume scanning were only recognised later. Therefore, no additional funding was required. Since the study is performed with free breathing, no anaesthetic support is required for setting up and continuing the service. Apart from radiographer and radiologist buy-in, referral departments need to learn about the new technique, its advantages, limitations and possible indications. Our experience of a gentle start performing dynamic 4-D CT only for patients in whom one of the traditional techniques is either not possible or has failed allows the team to perfect technical aspects and develop skill. The impressive visuals from dynamic 4-D CT sell themselves. Standard of care changes when referrals for the new technique push old techniques into obsoleteness.

As acknowledged in previous publications [1, 7], not all patients had correlative studies such as bronchoscopy, which is a limitation of this study, but the findings are accepted because of the established accuracy of CT.

We recommend dynamic 4-D CT as an achievable, low-dose, one-stop-shop imaging technique for diagnosing tracheobronchomalacia and its vascular causes through several referral patterns because it impacts management decisions for surgery or for optimising conservative management. The routine use of intravenous contrast, removal of nasogastric tubes, withdrawal of indwelling endotracheal tubes into the upper trachea and use of vacuum restraining devices will help novice users avoid many of the pitfalls and improve quality.